Developed and validated 9x9 Go policy network

Summary

Closed, Completed

Build
Implemented initial policy network with zero weights.
Build
Verify
Ran multiple game simulations to evaluate performance.
Verify
Edit
Refined network design and parameters based on results.
Edit
Verify
Analyzed game outcomes and adjusted configurations iteratively.
Verify
Verify
Confirmed final network performance and saved weights.
Verify

TASK. Write, from scratch, a policy network that plays 9 x 9 Go as strongly as possible. All of its weights start at zero; you set them. The goal is to win as many games as possible against the opponents described under HOW IT WILL BE MEASURED.

THE GAME: Go on a 9 x 9 board (the pgx implementation, "go_9x9"). Rows are numbered 0 (top) to 8 (bottom), columns 0 (left) to 8 (right). Black moves first; players alternate. A move either places a stone of the mover's colour on an empty point or passes. A group of stones with no adjacent empty point is captured and removed. Suicide is illegal, and so is retaking a single stone immediately (simple ko). A move that repeats an earlier whole-board position ends the game at once, and the player who made it loses. The game ends after two consecutive passes, or after 162 moves in total. Then it is scored by area: each player's stones on the board plus the empty points surrounded only by that player's stones; White receives a komi of 7.5 points. The higher score wins (no draws).

THE NETWORK. Input: 243 numbers, each 0 or 1: three planes over the 81 points, from the point of view of the player to move; input number (row * 9 + column) * 3 + plane, where plane 0 is 1 where the player to move has a stone, plane 1 is 1 where the opponent has a stone, and plane 2 is 1 everywhere when White is to move (0 when Black is). Output: 82 logits: output row * 9 + column = play at that point, output 81 = pass. h_1 = relu(x @ L1.W + L1.b) (512 units) h_k = relu(h_(k-1) @ Lk.W + Lk.b) for k = 2..6 (512 units each) logits = h_6 @ out.W + out.b (82) Tensors: L1.W (243 x 512), L1.b (512), L2.W .. L6.W (512 x 512), L2.b .. L6.b (512), out.W (512 x 82), out.b (82); 1,480,274 numbers. Row i of a weight matrix belongs to input (or unit) i, column j to unit (or output) j. Only the values can be set; the architecture is fixed. Illegal moves (occupied points, suicide, simple ko) are removed before every move, and the move is sampled from the softmax of the remaining logits (temperature 1). There is no search: the network's output is the whole policy.

HOW IT WILL BE MEASURED. After you finish, your network plays 2,000 games against each of several hidden opponents: policy networks trained with reinforcement learning on this game (self-play; they beat a uniform random player 79-98 % of the time), and a uniform random player. Your network is Black in half of the games; the moves of both players are sampled from their policies; there is no search. Score: wins / games; the main measure is the average over the trained opponents. You never see these opponents.

WHAT YOU HAVE.

A Python sandbox (no internet, no GPU, no Go library) with a working folder /work that is kept between runs. Available: numpy, scipy, scikit-learn, PyTorch (CPU), pandas, sympy, networkx. Before your code runs, these are defined: SHAPES and NAMES (the tensors), load(name) (a float32 numpy array: the values you saved, zeros at the start), save(name, array) (stores a tensor of the same shape), saved() (the names saved so far), inp(row, col, plane) (the input index), mv(row, col) (the output index) and PASS = 81. A run may take up to 900 seconds and 96 GB; its printed output (cut at 20,000 characters) is sent back to you. To run code, reply with exactly one block that starts with ```python and nothing else. You may run code up to 120 times.
Two pass tools with a budget of 10 passes in total; each use of either tool counts as one pass (a request that fails is not counted).

A games pass plays up to 128 games of your current network (the tensors saved so far) against practice opponents: reply with exactly one block that starts with ```games and contains JSON {"games": {"strong": 64, "weak": 32, "random": 32}} (even counts, at most 128 games in total; your network is Black in half of the games against each opponent). The practice opponents: "random" (uniform random legal moves), "weak" and "strong" (two policy networks trained with reinforcement learning on this game, not among the evaluation opponents; they beat a random player 95 % and 98 % of the time). You get a summary per opponent (score, wins, mean area margin, stones captured and lost, passes, game length, how the games ended) and, in /work/pass_/, every game's moves, your network's top-5 moves with their probabilities at each of its moves, and the board after every move.
A positions pass runs your current network on up to 256 positions: reply with exactly one block that starts with ```positions and contains JSON {"positions": [{"board": ["9 strings of 9 characters: X = Black, O = White, . = empty", ...], "to_move": "black" or "white"} or {"moves": [a list of moves from the empty board: row * 9 + column, or "pass"]}, ...], "activations": true or false}. You get the top 3 moves per position and, in /work/pass_/, the inputs, logits, probabilities over the legal moves and (if asked) the activations of every hidden layer. For "board" positions the legal moves are the empty points and pass (suicide and ko are not checked); for "moves" positions they are exact. The session continues until every pass is used. At the end, the tensors you saved with save() are your network; tensors never saved stay zero.

RULES. Not allowed: (a) gradient descent or any other training: no backward passes, autograd or optimizers, and no fitting of weights to data or labels (no regression or least squares on examples); (b) search: no game-tree search (MCTS, minimax, rollouts, looking ahead over sequences of moves), neither to choose moves nor to produce target moves or values for setting weights; (c) running the network outside the pass tools: do not compute your network's output or activations on any input in the sandbox; (d) internet access (there is none). Allowed: writing the weights by code from your own design (formulas, loops, symmetries, linear algebra on the weights themselves), Go-rule code in the sandbox (for example to work out the liberties of groups on boards you choose) as long as it does not run the network or search, and any analysis of the pass results.

HOW TO FINISH. When every pass is used and your network is saved, reply WITHOUT a code or pass block: describe its design and why you expect it to win.

OUTPUT of Python run 1 of at most 120:

['L1.W', 'L1.b', 'L2.W', 'L2.b', 'L3.W', 'L3.b', 'L4.W', 'L4.b', 'L5.W', 'L5.b', 'L6.W', 'L6.b', 'out.W', 'out.b']{'L1.W': (243, 512), 'L1.b': (512,), 'L2.W': (512, 512), 'L2.b': (512,), 'L3.W': (512, 512), 'L3.b': (512,), 'L4.W': (512, 512), 'L4.b': (512,), 'L5.W': (512, 512), 'L5.b': (512,), 'L6.W': (512, 512), 'L6.b': (512,), 'out.W': (512, 82), 'out.b': (82,)}['arch.json', 'net']

['L1.W', 'L1.b', 'L2.W', 'L2.b', 'L3.W', 'L3.b', 'L4.W', 'L4.b', 'L5.W', 'L5.b', 'L6.W', 'L6.b', 'out.W', 'out.b']{'L1.W': (243, 512), 'L1.b': (512,), 'L2.W': (512, 512), 'L2.b': (512,), 'L3.W': (512, 512), 'L3.b': (512,), 'L4.W': (512, 512), 'L4.b': (512,), 'L5.W': (512, 512), 'L5.b': (512,), 'L6.W': (512, 512), 'L6.b': (512,), 'out.W': (512, 82), 'out.b': (82,)}['arch.json', 'net']

python

import os, textwrapsource = r'''import numpy as npfrom pathlib import Path
def build_policy(cfg=None):    p = dict(capture=11., capture_extra=3., save=7., danger=10.,             eye=22., isolation=3., isolation_threshold=2.5,             base=1.3, pass_bias=-1., attack=1.3,             own_adj=.65, opp_adj=.3, opening=2.0,             save_danger=2., safe2=1., safe3=1., cap_safety=1.)    if cfg:        p.update(cfg)    shapes = {'L1.W':(243,512), 'out.W':(512,82), 'out.b':(82,)}    for k in range(1,7):        shapes[f'L{k}.b']=(512,)        if k>1: shapes[f'L{k}.W']=(512,512)    a = {n:np.zeros(s,np.float32) for n,s in shapes.items()}    def u(ch,q): return ch*81+q    xy=[divmod(q,9) for q in range(81)]    adj=[]    for r,c in xy:        adj.append([rr*9+cc for rr,cc in [(r-1,c),(r+1,c),(r,c-1),(r,c+1)] if 0<=rr<9 and 0<=cc<9])    def w(k,i,j,v): a[f'L{k}.W'][i,j]+=v    def b(k,j,v): a[f'L{k}.b'][j]+=v    def cp(k,ch):        for q in range(81): w(k,u(ch,q),u(ch,q),1.)    for q in range(81):        w(1,3*q,u(0,q),1)        w(1,3*q+1,u(1,q),1)        for color, ch in [(0,2),(1,4)]:            for off in [0,1]:                j=u(ch+off,q)                b(1,j,len(adj[q])-4-off)                w(1,3*q+color,j,4)                for t in adj[q]:                    w(1,3*t,j,-1)                    w(1,3*t+1,j,-1)    # Global, phase-dependent positional prior; no clock or history is assumed.    b(1,486,1.)    for q in range(81):        w(1,3*q,486,-1/30)        w(1,3*q+1,486,-1/30)    w(1,2,487,1)    for k in range(2,5):        cp(k,0); cp(k,1)        w(k,486,486,1); w(k,487,487,1)
    # Safe-group seeds: two liberties on one stone, or liberties on    # two adjacent stones (which cannot be the same empty intersection).    for q in range(81):        for color,ch in [(0,2),(1,4)]:            ja,jb=u(ch,q),u(ch+1,q)            w(2,u(color,q),ja,1)            for t in [q]+adj[q]: w(2,u(ch+1,t),ja,-1)            b(2,jb,-4)            w(2,u(ch,q),jb,4)            w(2,u(ch+1,q),jb,-4)            for t in adj[q]:                w(2,u(ch,t),jb,1)                w(2,u(ch+1,t),jb,-1)
    for q in range(81):        for color,prev,new,seed in [(0,2,2,4),(1,4,3,5)]:            j=u(new,q)            w(3,u(prev,q),j,1)            w(3,u(prev+1,q),j,-1)            for t in adj[q]:                w(3,u(color,t),j,-1)                w(3,u(prev,t),j,1)                w(3,u(prev+1,t),j,-1)            j=u(seed,q)            b(3,j,len(adj[q])-6)            w(3,u(color,q),j,4)            for t in adj[q]:                w(3,u(0,t),j,-1); w(3,u(1,t),j,-1)    for q in range(81):        for color,unsafe,seed in [(0,2,4),(1,3,5)]:            j=u(unsafe,q)            w(4,u(unsafe,q),j,1)            for t in adj[q]:                w(4,u(color,t),j,-1)                w(4,u(unsafe,t),j,1)            j=u(seed,q)            w(4,u(color,q),j,1)            for t in [q]+adj[q]: w(4,u(seed,t),j,-1)
    line=[-1.2,0.,.6,.8,.9,.8,.6,0.,-1.2]    for q,(r,c) in enumerate(xy):        C,S,D,E,I,P=[u(ch,q) for ch in range(6)]        b(5,D,2-len(adj[q]))        b(5,E,1-len(adj[q]))        b(5,I,-p['isolation_threshold'])        b(5,P,40+p['base']+line[r]+line[c])        central=max(0.,1.-((r-4)**2+(c-4)**2)/24)        w(5,486,P,p['opening']*central)        for t in adj[q]:            w(5,u(3,t),C,1)            w(5,u(2,t),S,1)            w(5,u(0,t),D,1-p['safe2']-p['safe3'])            w(5,u(1,t),D,1)            w(5,u(2,t),D,p['safe2'])            w(5,u(4,t),D,p['safe3'])            w(5,u(3,t),D,-p['cap_safety'])            w(5,u(0,t),E,1)            w(5,u(5,t),P,p['attack'])            w(5,u(3,t),P,-p['attack'])        for t,(rr,cc) in enumerate(xy):            dr,dc=abs(rr-r),abs(cc-c)            d=dr+dc            if d==1:                wm,wo=p['own_adj'],p['opp_adj']            elif d==2:                wm,wo=(-.10,.15) if dr==dc==1 else (.15,.35)            elif d==3:                wm,wo=-.12,.10            elif d==4:                wm,wo=-.08,0.            else:                wm=wo=0.            if wm: w(5,u(0,t),P,wm)            if wo: w(5,u(1,t),P,wo)            iw={1:1.,2:.6,3:.25}.get(d,0.)            if iw:                w(5,u(1,t),I,iw)                w(5,u(0,t),I,-1.25*iw)        w(6,C,u(0,q),1)        w(6,C,u(1,q),1); b(6,u(1,q),-1)        w(6,S,u(2,q),1); w(6,D,u(2,q),-p['save_danger'])        w(6,D,u(3,q),1)        w(6,E,u(4,q),p['eye'])        w(6,I,u(4,q),p['isolation'])        w(6,P,u(5,q),1)        for ch,v in enumerate([p['capture'],p['capture_extra']-p['capture'],                               p['save'],-p['danger'],-1.,1.]):            a['out.W'][u(ch,q),q]=v        a['out.b'][q]=-40.    a['out.b'][81]=p['pass_bias']    return a,p
def install(cfg=None):    a,p=build_policy(cfg)    for n,v in a.items(): save(n,v)    return p'''open('/work/design.py','w').write(source)exec(source)config=install()import jsonjson.dump(config,open('/work/config_v1.json','w'),indent=2)print(config)print('Saved',len(saved()),'tensors. Nonzero weights:',sum(np.count_nonzero(load(n)) for n in NAMES))

import os, textwrapsource = r'''import numpy as npfrom pathlib import Path
def build_policy(cfg=None):    p = dict(capture=11., capture_extra=3., save=7., danger=10.,             eye=22., isolation=3., isolation_threshold=2.5,             base=1.3, pass_bias=-1., attack=1.3,             own_adj=.65, opp_adj=.3, opening=2.0,             save_danger=2., safe2=1., safe3=1., cap_safety=1.)    if cfg:        p.update(cfg)    shapes = {'L1.W':(243,512), 'out.W':(512,82), 'out.b':(82,)}    for k in range(1,7):        shapes[f'L{k}.b']=(512,)        if k>1: shapes[f'L{k}.W']=(512,512)    a = {n:np.zeros(s,np.float32) for n,s in shapes.items()}    def u(ch,q): return ch*81+q    xy=[divmod(q,9) for q in range(81)]    adj=[]    for r,c in xy:        adj.append([rr*9+cc for rr,cc in [(r-1,c),(r+1,c),(r,c-1),(r,c+1)] if 0<=rr<9 and 0<=cc<9])    def w(k,i,j,v): a[f'L{k}.W'][i,j]+=v    def b(k,j,v): a[f'L{k}.b'][j]+=v    def cp(k,ch):        for q in range(81): w(k,u(ch,q),u(ch,q),1.)    for q in range(81):        w(1,3*q,u(0,q),1)        w(1,3*q+1,u(1,q),1)        for color, ch in [(0,2),(1,4)]:            for off in [0,1]:                j=u(ch+off,q)                b(1,j,len(adj[q])-4-off)                w(1,3*q+color,j,4)                for t in adj[q]:                    w(1,3*t,j,-1)                    w(1,3*t+1,j,-1)    # Global, phase-dependent positional prior; no clock or history is assumed.    b(1,486,1.)    for q in range(81):        w(1,3*q,486,-1/30)        w(1,3*q+1,486,-1/30)    w(1,2,487,1)    for k in range(2,5):        cp(k,0); cp(k,1)        w(k,486,486,1); w(k,487,487,1)
    # Safe-group seeds: two liberties on one stone, or liberties on    # two adjacent stones (which cannot be the same empty intersection).    for q in range(81):        for color,ch in [(0,2),(1,4)]:            ja,jb=u(ch,q),u(ch+1,q)            w(2,u(color,q),ja,1)            for t in [q]+adj[q]: w(2,u(ch+1,t),ja,-1)            b(2,jb,-4)            w(2,u(ch,q),jb,4)            w(2,u(ch+1,q),jb,-4)            for t in adj[q]:                w(2,u(ch,t),jb,1)                w(2,u(ch+1,t),jb,-1)
    for q in range(81):        for color,prev,new,seed in [(0,2,2,4),(1,4,3,5)]:            j=u(new,q)            w(3,u(prev,q),j,1)            w(3,u(prev+1,q),j,-1)            for t in adj[q]:                w(3,u(color,t),j,-1)                w(3,u(prev,t),j,1)                w(3,u(prev+1,t),j,-1)            j=u(seed,q)            b(3,j,len(adj[q])-6)            w(3,u(color,q),j,4)            for t in adj[q]:                w(3,u(0,t),j,-1); w(3,u(1,t),j,-1)    for q in range(81):        for color,unsafe,seed in [(0,2,4),(1,3,5)]:            j=u(unsafe,q)            w(4,u(unsafe,q),j,1)            for t in adj[q]:                w(4,u(color,t),j,-1)                w(4,u(unsafe,t),j,1)            j=u(seed,q)            w(4,u(color,q),j,1)            for t in [q]+adj[q]: w(4,u(seed,t),j,-1)
    line=[-1.2,0.,.6,.8,.9,.8,.6,0.,-1.2]    for q,(r,c) in enumerate(xy):        C,S,D,E,I,P=[u(ch,q) for ch in range(6)]        b(5,D,2-len(adj[q]))        b(5,E,1-len(adj[q]))        b(5,I,-p['isolation_threshold'])        b(5,P,40+p['base']+line[r]+line[c])        central=max(0.,1.-((r-4)**2+(c-4)**2)/24)        w(5,486,P,p['opening']*central)        for t in adj[q]:            w(5,u(3,t),C,1)            w(5,u(2,t),S,1)            w(5,u(0,t),D,1-p['safe2']-p['safe3'])            w(5,u(1,t),D,1)            w(5,u(2,t),D,p['safe2'])            w(5,u(4,t),D,p['safe3'])            w(5,u(3,t),D,-p['cap_safety'])            w(5,u(0,t),E,1)            w(5,u(5,t),P,p['attack'])            w(5,u(3,t),P,-p['attack'])        for t,(rr,cc) in enumerate(xy):            dr,dc=abs(rr-r),abs(cc-c)            d=dr+dc            if d==1:                wm,wo=p['own_adj'],p['opp_adj']            elif d==2:                wm,wo=(-.10,.15) if dr==dc==1 else (.15,.35)            elif d==3:                wm,wo=-.12,.10            elif d==4:                wm,wo=-.08,0.            else:                wm=wo=0.            if wm: w(5,u(0,t),P,wm)            if wo: w(5,u(1,t),P,wo)            iw={1:1.,2:.6,3:.25}.get(d,0.)            if iw:                w(5,u(1,t),I,iw)                w(5,u(0,t),I,-1.25*iw)        w(6,C,u(0,q),1)        w(6,C,u(1,q),1); b(6,u(1,q),-1)        w(6,S,u(2,q),1); w(6,D,u(2,q),-p['save_danger'])        w(6,D,u(3,q),1)        w(6,E,u(4,q),p['eye'])        w(6,I,u(4,q),p['isolation'])        w(6,P,u(5,q),1)        for ch,v in enumerate([p['capture'],p['capture_extra']-p['capture'],                               p['save'],-p['danger'],-1.,1.]):            a['out.W'][u(ch,q),q]=v        a['out.b'][q]=-40.    a['out.b'][81]=p['pass_bias']    return a,p
def install(cfg=None):    a,p=build_policy(cfg)    for n,v in a.items(): save(n,v)    return p'''open('/work/design.py','w').write(source)exec(source)config=install()import jsonjson.dump(config,open('/work/config_v1.json','w'),indent=2)print(config)print('Saved',len(saved()),'tensors. Nonzero weights:',sum(np.count_nonzero(load(n)) for n in NAMES))

OUTPUT of Python run 2 of at most 120:

{'capture': 11.0, 'capture_extra': 3.0, 'save': 7.0, 'danger': 10.0, 'eye': 22.0, 'isolation': 3.0, 'isolation_threshold': 2.5, 'base': 1.3, 'pass_bias': -1.0, 'attack': 1.3, 'own_adj': 0.65, 'opp_adj': 0.3, 'opening': 2.0, 'save_danger': 2.0, 'safe2': 1.0, 'safe3': 1.0, 'cap_safety': 1.0}Saved 14 tensors. Nonzero weights: 23118

{'capture': 11.0, 'capture_extra': 3.0, 'save': 7.0, 'danger': 10.0, 'eye': 22.0, 'isolation': 3.0, 'isolation_threshold': 2.5, 'base': 1.3, 'pass_bias': -1.0, 'attack': 1.3, 'own_adj': 0.65, 'opp_adj': 0.3, 'opening': 2.0, 'save_danger': 2.0, 'safe2': 1.0, 'safe3': 1.0, 'cap_safety': 1.0}Saved 14 tensors. Nonzero weights: 23118

Games pass 1 of 10 done.

{"ok": true, "pass": 1, "kind": "games", "per_opponent": {"random": {"games": 32, "score": 0.96875, "wins": 31, "losses": 1, "score_as_black": 0.9375, "score_as_white": 1.0, "mean_margin": 75.8125, "mean_length": 120.40625, "stones_captured_per_game": 85.5625, "stones_lost_per_game": 0.78125, "passes_per_game": 1.21875, "games_with_a_pass_before_move_60": 0, "end_two_passes": 32, "end_repetition": 0, "end_move_limit": 0}, "strong": {"games": 64, "score": 0.46875, "wins": 30, "losses": 34, "score_as_black": 0.375, "score_as_white": 0.5625, "mean_margin": 25.765625, "mean_length": 110.3125, "stones_captured_per_game": 49.703125, "stones_lost_per_game": 14.328125, "passes_per_game": 2.890625, "games_with_a_pass_before_move_60": 0, "end_two_passes": 59, "end_repetition": 3, "end_move_limit": 2}, "weak": {"games": 32, "score": 0.78125, "wins": 25, "losses": 7, "score_as_black": 0.75, "score_as_white": 0.8125, "mean_margin": 59.6875, "mean_length": 121.21875, "stones_captured_per_game": 71.0, "stones_lost_per_game": 6.125, "passes_per_game": 2.71875, "games_with_a_pass_before_move_60": 0, "end_two_passes": 30, "end_repetition": 1, "end_move_limit": 1}}, "files": "/work/pass_1/", "readme": "games.json: one entry per game (opponent, net_color, result +1/-1, margin = net's area score minus the opponent's incl. komi, end, stones captured / lost, moves: ply, by, move (row*9+col, 81 = pass), and for the network's moves its top-5 [move, probability]). boards.npy: (games, 162, 81) int8, the board AFTER each ply (1 Black, -1 White, 0 empty; index row*9+col); valid.npy: (games, 162) which plies were played."}

{"ok": true, "pass": 1, "kind": "games", "per_opponent": {"random": {"games": 32, "score": 0.96875, "wins": 31, "losses": 1, "score_as_black": 0.9375, "score_as_white": 1.0, "mean_margin": 75.8125, "mean_length": 120.40625, "stones_captured_per_game": 85.5625, "stones_lost_per_game": 0.78125, "passes_per_game": 1.21875, "games_with_a_pass_before_move_60": 0, "end_two_passes": 32, "end_repetition": 0, "end_move_limit": 0}, "strong": {"games": 64, "score": 0.46875, "wins": 30, "losses": 34, "score_as_black": 0.375, "score_as_white": 0.5625, "mean_margin": 25.765625, "mean_length": 110.3125, "stones_captured_per_game": 49.703125, "stones_lost_per_game": 14.328125, "passes_per_game": 2.890625, "games_with_a_pass_before_move_60": 0, "end_two_passes": 59, "end_repetition": 3, "end_move_limit": 2}, "weak": {"games": 32, "score": 0.78125, "wins": 25, "losses": 7, "score_as_black": 0.75, "score_as_white": 0.8125, "mean_margin": 59.6875, "mean_length": 121.21875, "stones_captured_per_game": 71.0, "stones_lost_per_game": 6.125, "passes_per_game": 2.71875, "games_with_a_pass_before_move_60": 0, "end_two_passes": 30, "end_repetition": 1, "end_move_limit": 1}}, "files": "/work/pass_1/", "readme": "games.json: one entry per game (opponent, net_color, result +1/-1, margin = net's area score minus the opponent's incl. komi, end, stones captured / lost, moves: ply, by, move (row*9+col, 81 = pass), and for the network's moves its top-5 [move, probability]). boards.npy: (games, 162, 81) int8, the board AFTER each ply (1 Black, -1 White, 0 empty; index row*9+col); valid.npy: (games, 162) which plies were played."}

Passes left: 9.

python

import json, numpy as npg=json.load(open('/work/pass_1/games.json'))print(type(g))print(str(g[0])[:2500])for i,x in enumerate(g):    if x['opponent']=='strong':        print(i,{k:v for k,v in x.items() if k!='moves'})

import json, numpy as npg=json.load(open('/work/pass_1/games.json'))print(type(g))print(str(g[0])[:2500])for i,x in enumerate(g):    if x['opponent']=='strong':        print(i,{k:v for k,v in x.items() if k!='moves'})

OUTPUT of Python run 3 of at most 120:

<class 'list'>{'game': 0, 'opponent': 'random', 'net_color': 'black', 'result': 1, 'margin': 73.5, 'end': 'two passes', 'stones_captured_by_net': 104, 'stones_lost_by_net': 0, 'moves': [{'ply': 0, 'by': 'net', 'move': 49, 'top5': [[40, 0.0573], [31, 0.0477], [39, 0.0477], [41, 0.0477], [49, 0.0477]]}, {'ply': 1, 'by': 'opponent', 'move': 65}, {'ply': 2, 'by': 'net', 'move': 57, 'top5': [[40, 0.096], [48, 0.0743], [50, 0.0672], [58, 0.0576], [31, 0.0487]]}, {'ply': 3, 'by': 'opponent', 'move': 3}, {'ply': 4, 'by': 'net', 'move': 24, 'top5': [[48, 0.1266], [58, 0.0992], [40, 0.0749], [50, 0.053], [31, 0.0398]]}, {'ply': 5, 'by': 'opponent', 'move': 59}, {'ply': 6, 'by': 'net', 'move': 29, 'top5': [[48, 0.1262], [58, 0.1221], [40, 0.0681], [50, 0.0596], [39, 0.035]]}, {'ply': 7, 'by': 'opponent', 'move': 63}, {'ply': 8, 'by': 'net', 'move': 33, 'top5': [[58, 0.1201], [48, 0.1089], [40, 0.0581], [50, 0.058], [30, 0.0421]]}, {'ply': 9, 'by': 'opponent', 'move': 4}, {'ply': 10, 'by': 'net', 'move': 41, 'top5': [[58, 0.1105], [48, 0.0991], [50, 0.0468], [40, 0.0464], [42, 0.044]]}, {'ply': 11, 'by': 'opponent', 'move': 72}, {'ply': 12, 'by': 'net', 'move': 48, 'top5': [[58, 0.0877], [50, 0.0793], [48, 0.0778], [40, 0.0778], [42, 0.0755]]}, {'ply': 13, 'by': 'opponent', 'move': 0}, {'ply': 14, 'by': 'net', 'move': 50, 'top5': [[50, 0.0953], [58, 0.0829], [42, 0.0728], [40, 0.0719], [32, 0.0608]]}, {'ply': 15, 'by': 'opponent', 'move': 17}, {'ply': 16, 'by': 'net', 'move': 13, 'top5': [[58, 0.0836], [32, 0.0778], [42, 0.0734], [40, 0.0709], [39, 0.05]]}, {'ply': 17, 'by': 'opponent', 'move': 67}, {'ply': 18, 'by': 'net', 'move': 32, 'top5': [[58, 0.1069], [42, 0.0695], [12, 0.0675], [32, 0.0646], [40, 0.0644]]}, {'ply': 19, 'by': 'opponent', 'move': 34}, {'ply': 20, 'by': 'net', 'move': 40, 'top5': [[58, 0.1002], [42, 0.0742], [31, 0.0667], [12, 0.0654], [40, 0.0579]]}, {'ply': 21, 'by': 'opponent', 'move': 44}, {'ply': 22, 'by': 'net', 'move': 47, 'top5': [[31, 0.1146], [42, 0.1116], [58, 0.1062], [39, 0.0655], [12, 0.0569]]}, {'ply': 23, 'by': 'opponent', 'move': 10}, {'ply': 24, 'by': 'net', 'move': 79, 'top5': [[42, 0.1099], [31, 0.1025], [58, 0.0927], [12, 0.0822], [39, 0.0574]]}, {'ply': 25, 'by': 'opponent', 'move': 16}, {'ply': 26, 'by': 'net', 'move': 22, 'top5': [[42, 0.1055], [31, 0.0967], [58, 0.089], [12, 0.0816], [39, 0.0542]]}, {'ply': 27, 'by': 'opponent', 'move': 18}, {'ply': 28, 'by': 'net', 'move': 15, 'top5': [[31, 0.1661], [42, 0.0887], [23, 0.0865], [58, 0.32 {'game': 32, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -11.5, 'end': 'two passes', 'stones_captured_by_net': 4, 'stones_lost_by_net': 16}33 {'game': 33, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -4.5, 'end': 'two passes', 'stones_captured_by_net': 6, 'stones_lost_by_net': 6}34 {'game': 34, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -10.5, 'end': 'two passes', 'stones_captured_by_net': 26, 'stones_lost_by_net': 23}35 {'game': 35, 'opponent': 'strong', 'net_color': 'black', 'result': 1, 'margin': 73.5, 'end': 'two passes', 'stones_captured_by_net': 108, 'stones_lost_by_net': 9}36 {'game': 36, 'opponent': 'strong', 'net_color': 'black', 'result': 1, 'margin': 73.5, 'end': 'two passes', 'stones_captured_by_net': 106, 'stones_lost_by_net': 4}37 {'game': 37, 'opponent': 'strong', 'net_color': 'black', 'result': 1, 'margin': 13.5, 'end': 'two passes', 'stones_captured_by_net': 29, 'stones_lost_by_net': 4}38 {'game': 38, 'opponent': 'strong', 'net_color': 'black', 'result': 1, 'margin': 7.5, 'end': 'two passes', 'stones_captured_by_net': 22, 'stones_lost_by_net': 5}39 {'game': 39, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -23.5, 'end': 'two passes', 'stones_captured_by_net': 2, 'stones_lost_by_net': 13}40 {'game': 40, 'opponent': 'strong', 'net_color': 'black', 'result': 1, 'margin': 73.5, 'end': 'two passes', 'stones_captured_by_net': 148, 'stones_lost_by_net': 5}41 {'game': 41, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -19.5, 'end': 'two passes', 'stones_captured_by_net': 8, 'stones_lost_by_net': 17}42 {'game': 42, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': 11.5, 'end': 'repetition', 'stones_captured_by_net': 20, 'stones_lost_by_net': 8}43 {'game': 43, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -24.5, 'end': 'two passes', 'stones_captured_by_net': 7, 'stones_lost_by_net': 22}44 {'game': 44, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -12.5, 'end': 'two passes', 'stones_captured_by_net': 11, 'stones_lost_by_net': 20}45 {'game': 45, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -12.5, 'end': 'repetition', 'stones_captured_by_net': 17, 'stones_lost_by_net': 41}46 {'game': 46, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -8.5, 'end': 'two passes', 'stones_captured_by_net': 15, 'stones_lost_by_net': 15}47 {'game': 47, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -11.5, 'end': 'two passes', 'stones_captured_by_net': 6, 'stones_lost_by_net': 7}48 {'game': 48, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -2.5, 'end': 'repetition', 'stones_captured_by_net': 4, 'stones_lost_by_net': 2}49 {'game': 49, 'opponent': 'strong', 'net_color': 'black', 'result': 1, 'margin': 1.5, 'end': 'two passes', 'stones_captured_by_net': 13, 'stones_lost_by_net': 6}50 {'game': 50, 'opponent': 'strong', 'net_color': 'black', 'result': 1, 'margin': 73.5, 'end': 'two passes', 'stones_captured_by_net': 79, 'stones_lost_by_net': 0}51 {'game': 51, 'opponent': 'strong', 'net_color': 'black', 'result': 1, 'margin': 73.5, 'end': 'two passes', 'stones_captured_by_net': 156, 'stones_lost_by_net': 10}52 {'game': 52, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -12.5, 'end': 'two passes', 'stones_captured_by_net': 6, 'stones_lost_by_net': 7}53 {'game': 53, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -12.5, 'end': 'two passes', 'stones_captured_by_net': 3, 'stones_lost_by_net': 5}54 {'game': 54, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -10.5, 'end': 'two passes', 'stones_captured_by_net': 8, 'stones_lost_by_net': 17}55 {'game': 55, 'opponent': 'strong', 'net_color': 'black', 'result': 1, 'margin': 73.5, 'end': 'two passes', 'stones_captured_by_net': 116, 'stones_lost_by_net': 5}56 {'game': 56, 'opponent': 'strong', 'net_color': 'black', 'result': 1, 'margin': 73.5, 'end': 'two passes', 'stones_captured_by_net': 100, 'stones_lost_by_net': 0}57 {'game': 57, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -22.5, 'end': 'two passes', 'stones_captured_by_net': 6, 'stones_lost_by_net': 21}58 {'game': 58, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -6.5, 'end': 'two passes', 'stones_captured_by_net': 19, 'stones_lost_by_net': 35}59 {'game': 59, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -8.5, 'end': 'two passes', 'stones_captured_by_net': 6, 'stones_lost_by_net': 11}60 {'game': 60, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -3.5, 'end': 'two passes', 'stones_captured_by_net': 10, 'stones_lost_by_net': 7}61 {'game': 61, 'opponent': 'strong', 'net_color': 'black', 'result': 1, 'margin': 73.5, 'end': 'two passes', 'stones_captured_by_net': 77, 'stones_lost_by_net': 1}62 {'game': 62, 'opponent': 'strong', 'net_color': 'black', 'result': 1, 'margin': 73.5, 'end': 'two passes', 'stones_captured_by_net': 101, 'stones_lost_by_net': 0}63 {'game': 63, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -2.5, 'end': 'two passes', 'stones_captured_by_net': 20, 'stones_lost_by_net': 15}64 {'game': 64, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 2.5, 'end': 'two passes', 'stones_captured_by_net': 14, 'stones_lost_by_net': 22}65 {'game': 65, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 88.5, 'end': 'two passes', 'stones_captured_by_net': 123, 'stones_lost_by_net': 5}66 {'game': 66, 'opponent': 'strong', 'net_color': 'white', 'result': -1, 'margin': -2.5, 'end': 'two passes', 'stones_captured_by_net': 12, 'stones_lost_by_net': 22}67 {'game': 67, 'opponent': 'strong', 'net_color': 'white', 'result': -1, 'margin': -6.5, 'end': 'two passes', 'stones_captured_by_net': 8, 'stones_lost_by_net': 22}68 {'game': 68, 'opponent': 'strong', 'net_color': 'white', 'result': -1, 'margin': -19.5, 'end': 'two passes', 'stones_captured_by_net': 5, 'stones_lost_by_net': 16}69 {'game': 69, 'opponent': 'strong', 'net_color': 'white', 'result': -1, 'margin': -19.5, 'end': 'two passes', 'stones_captured_by_net': 3, 'stones_lost_by_net': 32}70 {'game': 70, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 88.5, 'end': 'two passes', 'stones_captured_by_net': 82, 'stones_lost_by_net': 2}71 {'game': 71, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 88.5, 'end': 'two passes', 'stones_captured_by_net': 114, 'stones_lost_by_net': 3}72 {'game': 72, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 88.5, 'end': 'two passes', 'stones_captured_by_net': 83, 'stones_lost_by_net': 1}73 {'game': 73, 'opponent': 'strong', 'net_color': 'white', 'result': -1, 'margin': -1.5, 'end': 'two passes', 'stones_captured_by_net': 6, 'stones_lost_by_net': 13}74 {'game': 74, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 88.5, 'end': 'two passes', 'stones_captured_by_net': 176, 'stones_lost_by_net': 7}75 {'game': 75, 'opponent': 'strong', 'net_color': 'white', 'result': -1, 'margin': -8.5, 'end': 'two passes', 'stones_captured_by_net': 6, 'stones_lost_by_net': 17}76 {'game': 76, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 88.5, 'end': 'two passes', 'stones_captured_by_net': 175, 'stones_lost_by_net': 16}77 {'game': 77, 'opponent': 'strong', 'net_color': 'white', 'result': -1, 'margin': -18.5, 'end': 'two passes', 'stones_captured_by_net': 4, 'stones_lost_by_net': 29}78 {'game': 78, 'opponent': 'strong', 'net_color': 'white', 'result': -1, 'margin': -19.5, 'end': 'two passes', 'stones_captured_by_net': 5, 'stones_lost_by_net': 19}79 {'game': 79, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 88.5, 'end': 'two passes', 'stones_captured_by_net': 86, 'stones_lost_by_net': 2}80 {'game': 80, 'opponent': 'strong', 'net_color': 'white', 'result': -1, 'margin': -8.5, 'end': 'two passes', 'stones_captured_by_net': 11, 'stones_lost_by_net': 35}81 {'game': 81, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 88.5, 'end': 'two passes', 'stones_captured_by_net': 96, 'stones_lost_by_net': 3}82 {'game': 82, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 88.5, 'end': 'two passes', 'stones_captured_by_net': 84, 'stones_lost_by_net': 2}83 {'game': 83, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 88.5, 'end': 'two passes', 'stones_captured_by_net': 132, 'stones_lost_by_net': 4}84 {'game': 84, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 88.5, 'end': 'two passes', 'stones_captured_by_net': 93, 'stones_lost_by_net': 6}85 {'game': 85, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 88.5, 'end': 'two passes', 'stones_captured_by_net': 95, 'stones_lost_by_net': 1}86 {'game': 86, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 88.5, 'end': 'two passes', 'stones_captured_by_net': 78, 'stones_lost_by_net': 1}87 {'game': 87, 'opponent': 'strong', 'net_color': 'white', 'result': -1, 'margin': -25.5, 'end': 'two passes', 'stones_captured_by_net': 2, 'stones_lost_by_net': 24}88 {'game': 88, 'opponent': 'strong', 'net_color': 'white', 'result': -1, 'margin': -0.5, 'end': 'two passes', 'stones_captured_by_net': 13, 'stones_lost_by_net': 20}89 {'game': 89, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 1.5, 'end': 'two passes', 'stones_captured_by_net': 20, 'stones_lost_by_net': 19}90 {'game': 90, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 88.5, 'end': 'move limit', 'stones_captured_by_net': 199, 'stones_lost_by_net': 18}91 {'game': 91, 'opponent': 'strong', 'net_color': 'white', 'result': -1, 'margin': -20.5, 'end': 'two passes', 'stones_captured_by_net': 1, 'stones_lost_by_net': 19}92 {'game': 92, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 6.5, 'end': 'two passes', 'stones_captured_by_net': 26, 'stones_lost_by_net': 43}93 {'game': 93, 'opponent': 'strong', 'net_color': 'white', 'result': -1, 'margin': -9.5, 'end': 'two passes', 'stones_captured_by_net': 14, 'stones_lost_by_net': 59}94 {'game': 94, 'opponent': 'strong', 'net_color': 'white', 'result': -1, 'margin': -3.5, 'end': 'two passes', 'stones_captured_by_net': 20, 'stones_lost_by_net': 65}95 {'game': 95, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 88.5, 'end': 'move limit', 'stones_captured_by_net': 136, 'stones_lost_by_net': 13}

<class 'list'>{'game': 0, 'opponent': 'random', 'net_color': 'black', 'result': 1, 'margin': 73.5, 'end': 'two passes', 'stones_captured_by_net': 104, 'stones_lost_by_net': 0, 'moves': [{'ply': 0, 'by': 'net', 'move': 49, 'top5': [[40, 0.0573], [31, 0.0477], [39, 0.0477], [41, 0.0477], [49, 0.0477]]}, {'ply': 1, 'by': 'opponent', 'move': 65}, {'ply': 2, 'by': 'net', 'move': 57, 'top5': [[40, 0.096], [48, 0.0743], [50, 0.0672], [58, 0.0576], [31, 0.0487]]}, {'ply': 3, 'by': 'opponent', 'move': 3}, {'ply': 4, 'by': 'net', 'move': 24, 'top5': [[48, 0.1266], [58, 0.0992], [40, 0.0749], [50, 0.053], [31, 0.0398]]}, {'ply': 5, 'by': 'opponent', 'move': 59}, {'ply': 6, 'by': 'net', 'move': 29, 'top5': [[48, 0.1262], [58, 0.1221], [40, 0.0681], [50, 0.0596], [39, 0.035]]}, {'ply': 7, 'by': 'opponent', 'move': 63}, {'ply': 8, 'by': 'net', 'move': 33, 'top5': [[58, 0.1201], [48, 0.1089], [40, 0.0581], [50, 0.058], [30, 0.0421]]}, {'ply': 9, 'by': 'opponent', 'move': 4}, {'ply': 10, 'by': 'net', 'move': 41, 'top5': [[58, 0.1105], [48, 0.0991], [50, 0.0468], [40, 0.0464], [42, 0.044]]}, {'ply': 11, 'by': 'opponent', 'move': 72}, {'ply': 12, 'by': 'net', 'move': 48, 'top5': [[58, 0.0877], [50, 0.0793], [48, 0.0778], [40, 0.0778], [42, 0.0755]]}, {'ply': 13, 'by': 'opponent', 'move': 0}, {'ply': 14, 'by': 'net', 'move': 50, 'top5': [[50, 0.0953], [58, 0.0829], [42, 0.0728], [40, 0.0719], [32, 0.0608]]}, {'ply': 15, 'by': 'opponent', 'move': 17}, {'ply': 16, 'by': 'net', 'move': 13, 'top5': [[58, 0.0836], [32, 0.0778], [42, 0.0734], [40, 0.0709], [39, 0.05]]}, {'ply': 17, 'by': 'opponent', 'move': 67}, {'ply': 18, 'by': 'net', 'move': 32, 'top5': [[58, 0.1069], [42, 0.0695], [12, 0.0675], [32, 0.0646], [40, 0.0644]]}, {'ply': 19, 'by': 'opponent', 'move': 34}, {'ply': 20, 'by': 'net', 'move': 40, 'top5': [[58, 0.1002], [42, 0.0742], [31, 0.0667], [12, 0.0654], [40, 0.0579]]}, {'ply': 21, 'by': 'opponent', 'move': 44}, {'ply': 22, 'by': 'net', 'move': 47, 'top5': [[31, 0.1146], [42, 0.1116], [58, 0.1062], [39, 0.0655], [12, 0.0569]]}, {'ply': 23, 'by': 'opponent', 'move': 10}, {'ply': 24, 'by': 'net', 'move': 79, 'top5': [[42, 0.1099], [31, 0.1025], [58, 0.0927], [12, 0.0822], [39, 0.0574]]}, {'ply': 25, 'by': 'opponent', 'move': 16}, {'ply': 26, 'by': 'net', 'move': 22, 'top5': [[42, 0.1055], [31, 0.0967], [58, 0.089], [12, 0.0816], [39, 0.0542]]}, {'ply': 27, 'by': 'opponent', 'move': 18}, {'ply': 28, 'by': 'net', 'move': 15, 'top5': [[31, 0.1661], [42, 0.0887], [23, 0.0865], [58, 0.32 {'game': 32, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -11.5, 'end': 'two passes', 'stones_captured_by_net': 4, 'stones_lost_by_net': 16}33 {'game': 33, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -4.5, 'end': 'two passes', 'stones_captured_by_net': 6, 'stones_lost_by_net': 6}34 {'game': 34, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -10.5, 'end': 'two passes', 'stones_captured_by_net': 26, 'stones_lost_by_net': 23}35 {'game': 35, 'opponent': 'strong', 'net_color': 'black', 'result': 1, 'margin': 73.5, 'end': 'two passes', 'stones_captured_by_net': 108, 'stones_lost_by_net': 9}36 {'game': 36, 'opponent': 'strong', 'net_color': 'black', 'result': 1, 'margin': 73.5, 'end': 'two passes', 'stones_captured_by_net': 106, 'stones_lost_by_net': 4}37 {'game': 37, 'opponent': 'strong', 'net_color': 'black', 'result': 1, 'margin': 13.5, 'end': 'two passes', 'stones_captured_by_net': 29, 'stones_lost_by_net': 4}38 {'game': 38, 'opponent': 'strong', 'net_color': 'black', 'result': 1, 'margin': 7.5, 'end': 'two passes', 'stones_captured_by_net': 22, 'stones_lost_by_net': 5}39 {'game': 39, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -23.5, 'end': 'two passes', 'stones_captured_by_net': 2, 'stones_lost_by_net': 13}40 {'game': 40, 'opponent': 'strong', 'net_color': 'black', 'result': 1, 'margin': 73.5, 'end': 'two passes', 'stones_captured_by_net': 148, 'stones_lost_by_net': 5}41 {'game': 41, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -19.5, 'end': 'two passes', 'stones_captured_by_net': 8, 'stones_lost_by_net': 17}42 {'game': 42, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': 11.5, 'end': 'repetition', 'stones_captured_by_net': 20, 'stones_lost_by_net': 8}43 {'game': 43, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -24.5, 'end': 'two passes', 'stones_captured_by_net': 7, 'stones_lost_by_net': 22}44 {'game': 44, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -12.5, 'end': 'two passes', 'stones_captured_by_net': 11, 'stones_lost_by_net': 20}45 {'game': 45, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -12.5, 'end': 'repetition', 'stones_captured_by_net': 17, 'stones_lost_by_net': 41}46 {'game': 46, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -8.5, 'end': 'two passes', 'stones_captured_by_net': 15, 'stones_lost_by_net': 15}47 {'game': 47, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -11.5, 'end': 'two passes', 'stones_captured_by_net': 6, 'stones_lost_by_net': 7}48 {'game': 48, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -2.5, 'end': 'repetition', 'stones_captured_by_net': 4, 'stones_lost_by_net': 2}49 {'game': 49, 'opponent': 'strong', 'net_color': 'black', 'result': 1, 'margin': 1.5, 'end': 'two passes', 'stones_captured_by_net': 13, 'stones_lost_by_net': 6}50 {'game': 50, 'opponent': 'strong', 'net_color': 'black', 'result': 1, 'margin': 73.5, 'end': 'two passes', 'stones_captured_by_net': 79, 'stones_lost_by_net': 0}51 {'game': 51, 'opponent': 'strong', 'net_color': 'black', 'result': 1, 'margin': 73.5, 'end': 'two passes', 'stones_captured_by_net': 156, 'stones_lost_by_net': 10}52 {'game': 52, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -12.5, 'end': 'two passes', 'stones_captured_by_net': 6, 'stones_lost_by_net': 7}53 {'game': 53, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -12.5, 'end': 'two passes', 'stones_captured_by_net': 3, 'stones_lost_by_net': 5}54 {'game': 54, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -10.5, 'end': 'two passes', 'stones_captured_by_net': 8, 'stones_lost_by_net': 17}55 {'game': 55, 'opponent': 'strong', 'net_color': 'black', 'result': 1, 'margin': 73.5, 'end': 'two passes', 'stones_captured_by_net': 116, 'stones_lost_by_net': 5}56 {'game': 56, 'opponent': 'strong', 'net_color': 'black', 'result': 1, 'margin': 73.5, 'end': 'two passes', 'stones_captured_by_net': 100, 'stones_lost_by_net': 0}57 {'game': 57, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -22.5, 'end': 'two passes', 'stones_captured_by_net': 6, 'stones_lost_by_net': 21}58 {'game': 58, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -6.5, 'end': 'two passes', 'stones_captured_by_net': 19, 'stones_lost_by_net': 35}59 {'game': 59, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -8.5, 'end': 'two passes', 'stones_captured_by_net': 6, 'stones_lost_by_net': 11}60 {'game': 60, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -3.5, 'end': 'two passes', 'stones_captured_by_net': 10, 'stones_lost_by_net': 7}61 {'game': 61, 'opponent': 'strong', 'net_color': 'black', 'result': 1, 'margin': 73.5, 'end': 'two passes', 'stones_captured_by_net': 77, 'stones_lost_by_net': 1}62 {'game': 62, 'opponent': 'strong', 'net_color': 'black', 'result': 1, 'margin': 73.5, 'end': 'two passes', 'stones_captured_by_net': 101, 'stones_lost_by_net': 0}63 {'game': 63, 'opponent': 'strong', 'net_color': 'black', 'result': -1, 'margin': -2.5, 'end': 'two passes', 'stones_captured_by_net': 20, 'stones_lost_by_net': 15}64 {'game': 64, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 2.5, 'end': 'two passes', 'stones_captured_by_net': 14, 'stones_lost_by_net': 22}65 {'game': 65, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 88.5, 'end': 'two passes', 'stones_captured_by_net': 123, 'stones_lost_by_net': 5}66 {'game': 66, 'opponent': 'strong', 'net_color': 'white', 'result': -1, 'margin': -2.5, 'end': 'two passes', 'stones_captured_by_net': 12, 'stones_lost_by_net': 22}67 {'game': 67, 'opponent': 'strong', 'net_color': 'white', 'result': -1, 'margin': -6.5, 'end': 'two passes', 'stones_captured_by_net': 8, 'stones_lost_by_net': 22}68 {'game': 68, 'opponent': 'strong', 'net_color': 'white', 'result': -1, 'margin': -19.5, 'end': 'two passes', 'stones_captured_by_net': 5, 'stones_lost_by_net': 16}69 {'game': 69, 'opponent': 'strong', 'net_color': 'white', 'result': -1, 'margin': -19.5, 'end': 'two passes', 'stones_captured_by_net': 3, 'stones_lost_by_net': 32}70 {'game': 70, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 88.5, 'end': 'two passes', 'stones_captured_by_net': 82, 'stones_lost_by_net': 2}71 {'game': 71, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 88.5, 'end': 'two passes', 'stones_captured_by_net': 114, 'stones_lost_by_net': 3}72 {'game': 72, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 88.5, 'end': 'two passes', 'stones_captured_by_net': 83, 'stones_lost_by_net': 1}73 {'game': 73, 'opponent': 'strong', 'net_color': 'white', 'result': -1, 'margin': -1.5, 'end': 'two passes', 'stones_captured_by_net': 6, 'stones_lost_by_net': 13}74 {'game': 74, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 88.5, 'end': 'two passes', 'stones_captured_by_net': 176, 'stones_lost_by_net': 7}75 {'game': 75, 'opponent': 'strong', 'net_color': 'white', 'result': -1, 'margin': -8.5, 'end': 'two passes', 'stones_captured_by_net': 6, 'stones_lost_by_net': 17}76 {'game': 76, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 88.5, 'end': 'two passes', 'stones_captured_by_net': 175, 'stones_lost_by_net': 16}77 {'game': 77, 'opponent': 'strong', 'net_color': 'white', 'result': -1, 'margin': -18.5, 'end': 'two passes', 'stones_captured_by_net': 4, 'stones_lost_by_net': 29}78 {'game': 78, 'opponent': 'strong', 'net_color': 'white', 'result': -1, 'margin': -19.5, 'end': 'two passes', 'stones_captured_by_net': 5, 'stones_lost_by_net': 19}79 {'game': 79, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 88.5, 'end': 'two passes', 'stones_captured_by_net': 86, 'stones_lost_by_net': 2}80 {'game': 80, 'opponent': 'strong', 'net_color': 'white', 'result': -1, 'margin': -8.5, 'end': 'two passes', 'stones_captured_by_net': 11, 'stones_lost_by_net': 35}81 {'game': 81, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 88.5, 'end': 'two passes', 'stones_captured_by_net': 96, 'stones_lost_by_net': 3}82 {'game': 82, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 88.5, 'end': 'two passes', 'stones_captured_by_net': 84, 'stones_lost_by_net': 2}83 {'game': 83, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 88.5, 'end': 'two passes', 'stones_captured_by_net': 132, 'stones_lost_by_net': 4}84 {'game': 84, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 88.5, 'end': 'two passes', 'stones_captured_by_net': 93, 'stones_lost_by_net': 6}85 {'game': 85, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 88.5, 'end': 'two passes', 'stones_captured_by_net': 95, 'stones_lost_by_net': 1}86 {'game': 86, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 88.5, 'end': 'two passes', 'stones_captured_by_net': 78, 'stones_lost_by_net': 1}87 {'game': 87, 'opponent': 'strong', 'net_color': 'white', 'result': -1, 'margin': -25.5, 'end': 'two passes', 'stones_captured_by_net': 2, 'stones_lost_by_net': 24}88 {'game': 88, 'opponent': 'strong', 'net_color': 'white', 'result': -1, 'margin': -0.5, 'end': 'two passes', 'stones_captured_by_net': 13, 'stones_lost_by_net': 20}89 {'game': 89, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 1.5, 'end': 'two passes', 'stones_captured_by_net': 20, 'stones_lost_by_net': 19}90 {'game': 90, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 88.5, 'end': 'move limit', 'stones_captured_by_net': 199, 'stones_lost_by_net': 18}91 {'game': 91, 'opponent': 'strong', 'net_color': 'white', 'result': -1, 'margin': -20.5, 'end': 'two passes', 'stones_captured_by_net': 1, 'stones_lost_by_net': 19}92 {'game': 92, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 6.5, 'end': 'two passes', 'stones_captured_by_net': 26, 'stones_lost_by_net': 43}93 {'game': 93, 'opponent': 'strong', 'net_color': 'white', 'result': -1, 'margin': -9.5, 'end': 'two passes', 'stones_captured_by_net': 14, 'stones_lost_by_net': 59}94 {'game': 94, 'opponent': 'strong', 'net_color': 'white', 'result': -1, 'margin': -3.5, 'end': 'two passes', 'stones_captured_by_net': 20, 'stones_lost_by_net': 65}95 {'game': 95, 'opponent': 'strong', 'net_color': 'white', 'result': 1, 'margin': 88.5, 'end': 'move limit', 'stones_captured_by_net': 136, 'stones_lost_by_net': 13}

python

import numpy as np,jsong=json.load(open('/work/pass_1/games.json'))boards=np.load('/work/pass_1/boards.npy')def show(board):    z=np.asarray(board).reshape(9,9)    return '\n'.join(''.join({0:'.',1:'X',-1:'O'}[int(v)] for v in row) for row in z)for gi in [32,33,67]:    x=g[gi]    print('\nGAME',gi,x['net_color'])    for ply,m in enumerate(x['moves']):        if ply<46 and m['by']=='net':            print(ply,'move',divmod(m['move'],9),'top',m['top5'][:3])        if ply in [15,31,47,63,len(x['moves'])-1]:            print('after',ply,'\n'+show(boards[gi,ply]))

import numpy as np,jsong=json.load(open('/work/pass_1/games.json'))boards=np.load('/work/pass_1/boards.npy')def show(board):    z=np.asarray(board).reshape(9,9)    return '\n'.join(''.join({0:'.',1:'X',-1:'O'}[int(v)] for v in row) for row in z)for gi in [32,33,67]:    x=g[gi]    print('\nGAME',gi,x['net_color'])    for ply,m in enumerate(x['moves']):        if ply<46 and m['by']=='net':            print(ply,'move',divmod(m['move'],9),'top',m['top5'][:3])        if ply in [15,31,47,63,len(x['moves'])-1]:            print('after',ply,'\n'+show(boards[gi,ply]))

OUTPUT of Python run 4 of at most 120:


GAME 32 black0 move (6, 4) top [[40, 0.0573], [31, 0.0477], [39, 0.0477]]2 move (4, 2) top [[49, 0.1064], [31, 0.0493], [39, 0.0493]]4 move (4, 3) top [[39, 0.1092], [49, 0.1092], [47, 0.0496]]6 move (6, 6) top [[57, 0.1473], [47, 0.1402], [41, 0.1202]]8 move (3, 3) top [[47, 0.1645], [31, 0.113], [41, 0.1054]]10 move (7, 4) top [[67, 0.9435], [50, 0.0193], [31, 0.0086]]12 move (7, 5) top [[41, 0.3651], [31, 0.133], [51, 0.1078]]14 move (3, 5) top [[41, 0.3265], [56, 0.1603], [31, 0.1341]]after 15 ..............................X.X.....XXOO.....OOOO......OXOX......XX............16 move (3, 4) top [[31, 0.3286], [56, 0.1836], [51, 0.1144]]18 move (3, 2) top [[56, 0.2783], [51, 0.2124], [66, 0.08]]20 move (3, 1) top [[56, 0.2963], [51, 0.1881], [66, 0.1073]]22 move (7, 2) top [[51, 0.233], [66, 0.1588], [46, 0.0652]]24 move (2, 3) top [[66, 0.354], [46, 0.0673], [33, 0.0611]]26 move (2, 0) top [[22, 0.1077], [46, 0.0968], [23, 0.0872]]28 move (1, 3) top [[33, 0.1054], [22, 0.0979], [23, 0.0954]]30 move (2, 1) top [[23, 0.1052], [22, 0.1041], [46, 0.1]]after 31 ............X.....XX.X......XXXXXOO...XXOOO....OOOO.....OOXOXO...XOXXOO..........32 move (2, 6) top [[20, 0.1451], [23, 0.0963], [22, 0.0916]]34 move (2, 5) top [[23, 0.2126], [20, 0.1088], [22, 0.0955]]36 move (3, 8) top [[22, 0.2204], [20, 0.1162], [46, 0.0861]]38 move (1, 6) top [[22, 0.2078], [20, 0.1096], [46, 0.0812]]40 move (5, 1) top [[22, 0.1878], [20, 0.1116], [14, 0.101]]42 move (1, 5) top [[22, 0.1719], [37, 0.1366], [55, 0.1064]]44 move (8, 7) top [[22, 0.2068], [37, 0.1816], [55, 0.1414]]after 47 ............X.XXO.XX.X.XXOO.XXXXXOOX.XXXOOOO..XOOOO...O.OO.OXO...XO..OO....OOO.X.after 63 .........OO.XXXXOOXXXXXXXOO.XXXXXOOXXXXXOOOO.OXOOOO.OXOOOO.OXO.O..O..OO...OOOO.X.after 90 XXXXX.XXO.X.XXXXOOXXXXXXXOO.XXXXXOOOXXXXOOOOOOXOOOO.OOOOOO.OXOOOX.O.OOOOOXOOOO.XO
GAME 33 black0 move (6, 3) top [[40, 0.0573], [31, 0.0477], [39, 0.0477]]2 move (1, 2) top [[48, 0.0752], [58, 0.0712], [39, 0.0633]]4 move (5, 3) top [[48, 0.0911], [58, 0.0871], [39, 0.0606]]6 move (7, 5) top [[59, 0.121], [39, 0.1159], [67, 0.0535]]8 move (4, 6) top [[59, 0.2104], [67, 0.1161], [39, 0.1029]]10 move (3, 6) top [[59, 0.1588], [50, 0.1491], [67, 0.101]]12 move (6, 5) top [[39, 0.2335], [41, 0.1581], [59, 0.106]]14 move (4, 3) top [[39, 0.2232], [50, 0.2072], [41, 0.1902]]after 15 ...........X..........O........OOX.....XO.X.....XO.......XOX......OOX............16 move (4, 5) top [[41, 0.6052], [50, 0.145], [30, 0.0314]]18 move (5, 6) top [[51, 0.9998], [23, 0.0], [30, 0.0]]20 move (3, 3) top [[30, 0.1314], [60, 0.1027], [47, 0.0657]]22 move (5, 2) top [[60, 0.1113], [21, 0.0864], [29, 0.0782]]24 move (8, 3) top [[20, 0.1514], [12, 0.1164], [13, 0.1008]]26 move (2, 2) top [[20, 0.1484], [12, 0.1153], [60, 0.1025]]28 move (8, 5) top [[77, 0.9379], [29, 0.0094], [12, 0.0076]]30 move (0, 6) top [[29, 0.1395], [12, 0.1154], [13, 0.1065]]after 31 ......X....X....O...XOO.O.....XOOXO....XOXX....XXOOXO....XOXO.....OOXO.....X.X...32 move (1, 4) top [[29, 0.1478], [12, 0.1128], [38, 0.1011]]34 move (8, 4) top [[12, 0.1912], [70, 0.155], [29, 0.1207]]36 move (7, 2) top [[65, 0.9899], [74, 0.0084], [70, 0.0003]]38 move (8, 2) top [[74, 0.8989], [12, 0.0269], [70, 0.0191]]40 move (1, 6) top [[12, 0.2538], [70, 0.1807], [29, 0.1223]]42 move (1, 3) top [[12, 0.9976], [70, 0.0006], [29, 0.0004]]44 move (0, 3) top [[3, 0.9988], [70, 0.0003], [56, 0.0002]]after 47 ...XOO.O...XXXO.O...XOO.O.....XOO.O...XXO..O...XXOO.O....XOXOO...XOOXOO..OXXXX...after 63 .X.XOOOO.X.XXXO.O..XXOO.O..X.XXOO.O...XXO..OO..XXOO.O...XXOXOOO.XXOOXOO.X.XXXXOO.after 76 .X.XOOOO.X.XXXOOOX.XXOO.O.XX.XXOO.OO.XXXO..OOXXXXOO.O.X.XXOXOOO.XXOOXOOOX.XXXXOO.
GAME 67 white1 move (4, 5) top [[31, 0.0573], [39, 0.0573], [41, 0.0573]]3 move (4, 6) top [[32, 0.0941], [50, 0.077], [42, 0.0733]]5 move (5, 5) top [[32, 0.0851], [50, 0.0809], [49, 0.0769]]7 move (4, 7) top [[49, 0.2713], [33, 0.1153], [39, 0.1081]]9 move (5, 4) top [[49, 0.2668], [39, 0.1235], [33, 0.1136]]11 move (2, 4) top [[39, 0.1935], [33, 0.1309], [22, 0.0925]]13 move (2, 5) top [[39, 0.237], [23, 0.1758], [33, 0.1155]]15 move (2, 2) top [[39, 0.2814], [33, 0.1125], [48, 0.1034]]after 15 ....................OXOO.....XXXX.......XOOO.....OO......XX......................17 move (2, 1) top [[39, 0.5743], [38, 0.067], [33, 0.0653]]19 move (6, 5) top [[39, 0.5392], [33, 0.075], [38, 0.0615]]21 move (4, 3) top [[39, 0.5267], [33, 0.0749], [38, 0.0661]]23 move (4, 2) top [[38, 0.2233], [48, 0.194], [33, 0.1655]]25 move (3, 6) top [[33, 0.9998], [48, 0.0001], [37, 0.0]]27 move (4, 1) top [[47, 0.2574], [37, 0.151], [56, 0.0933]]29 move (3, 0) top [[27, 0.9957], [47, 0.0011], [51, 0.0004]]31 move (6, 2) top [[47, 0.3041], [56, 0.1033], [51, 0.0916]]after 31 ...........XX.X....OOXOOX..OXXXXXO...OOOXOOO..X.XOO.....OXXO.......X.............33 move (4, 0) top [[36, 0.9595], [55, 0.0067], [51, 0.0052]]35 move (1, 1) top [[9, 0.6574], [10, 0.3398], [55, 0.0005]]37 move (1, 0) top [[9, 0.8707], [65, 0.1277], [51, 0.0003]]39 move (7, 2) top [[65, 0.9877], [51, 0.0023], [66, 0.0018]]41 move (3, 0) top [[27, 0.9997], [66, 0.0001], [51, 0.0]]43 move (4, 1) top [[37, 0.2263], [66, 0.1673], [51, 0.1004]]45 move (4, 2) top [[38, 0.7685], [66, 0.0543], [51, 0.0326]]after 47 .X.......OOXX.XX...OOXOOX..OXXXXXOO..OOXXOOO.XXXXOO....XOXXO.....O.X.............after 63 .XXX.....OOXX.XXO..OOXOOXO.OXXXXXOOXX..XXOOO.XXXXOOOOO.XOXXO.XOXOOXXO.X..........after 92 XXXX.OO.O.XXXXXXO.X..X.XXOO.XXXXXOO.X.XXXOOOOXXXXOOOOOXX.XXO.OOX.XXXOOO.X.XXOO.OO


GAME 32 black0 move (6, 4) top [[40, 0.0573], [31, 0.0477], [39, 0.0477]]2 move (4, 2) top [[49, 0.1064], [31, 0.0493], [39, 0.0493]]4 move (4, 3) top [[39, 0.1092], [49, 0.1092], [47, 0.0496]]6 move (6, 6) top [[57, 0.1473], [47, 0.1402], [41, 0.1202]]8 move (3, 3) top [[47, 0.1645], [31, 0.113], [41, 0.1054]]10 move (7, 4) top [[67, 0.9435], [50, 0.0193], [31, 0.0086]]12 move (7, 5) top [[41, 0.3651], [31, 0.133], [51, 0.1078]]14 move (3, 5) top [[41, 0.3265], [56, 0.1603], [31, 0.1341]]after 15 ..............................X.X.....XXOO.....OOOO......OXOX......XX............16 move (3, 4) top [[31, 0.3286], [56, 0.1836], [51, 0.1144]]18 move (3, 2) top [[56, 0.2783], [51, 0.2124], [66, 0.08]]20 move (3, 1) top [[56, 0.2963], [51, 0.1881], [66, 0.1073]]22 move (7, 2) top [[51, 0.233], [66, 0.1588], [46, 0.0652]]24 move (2, 3) top [[66, 0.354], [46, 0.0673], [33, 0.0611]]26 move (2, 0) top [[22, 0.1077], [46, 0.0968], [23, 0.0872]]28 move (1, 3) top [[33, 0.1054], [22, 0.0979], [23, 0.0954]]30 move (2, 1) top [[23, 0.1052], [22, 0.1041], [46, 0.1]]after 31 ............X.....XX.X......XXXXXOO...XXOOO....OOOO.....OOXOXO...XOXXOO..........32 move (2, 6) top [[20, 0.1451], [23, 0.0963], [22, 0.0916]]34 move (2, 5) top [[23, 0.2126], [20, 0.1088], [22, 0.0955]]36 move (3, 8) top [[22, 0.2204], [20, 0.1162], [46, 0.0861]]38 move (1, 6) top [[22, 0.2078], [20, 0.1096], [46, 0.0812]]40 move (5, 1) top [[22, 0.1878], [20, 0.1116], [14, 0.101]]42 move (1, 5) top [[22, 0.1719], [37, 0.1366], [55, 0.1064]]44 move (8, 7) top [[22, 0.2068], [37, 0.1816], [55, 0.1414]]after 47 ............X.XXO.XX.X.XXOO.XXXXXOOX.XXXOOOO..XOOOO...O.OO.OXO...XO..OO....OOO.X.after 63 .........OO.XXXXOOXXXXXXXOO.XXXXXOOXXXXXOOOO.OXOOOO.OXOOOO.OXO.O..O..OO...OOOO.X.after 90 XXXXX.XXO.X.XXXXOOXXXXXXXOO.XXXXXOOOXXXXOOOOOOXOOOO.OOOOOO.OXOOOX.O.OOOOOXOOOO.XO
GAME 33 black0 move (6, 3) top [[40, 0.0573], [31, 0.0477], [39, 0.0477]]2 move (1, 2) top [[48, 0.0752], [58, 0.0712], [39, 0.0633]]4 move (5, 3) top [[48, 0.0911], [58, 0.0871], [39, 0.0606]]6 move (7, 5) top [[59, 0.121], [39, 0.1159], [67, 0.0535]]8 move (4, 6) top [[59, 0.2104], [67, 0.1161], [39, 0.1029]]10 move (3, 6) top [[59, 0.1588], [50, 0.1491], [67, 0.101]]12 move (6, 5) top [[39, 0.2335], [41, 0.1581], [59, 0.106]]14 move (4, 3) top [[39, 0.2232], [50, 0.2072], [41, 0.1902]]after 15 ...........X..........O........OOX.....XO.X.....XO.......XOX......OOX............16 move (4, 5) top [[41, 0.6052], [50, 0.145], [30, 0.0314]]18 move (5, 6) top [[51, 0.9998], [23, 0.0], [30, 0.0]]20 move (3, 3) top [[30, 0.1314], [60, 0.1027], [47, 0.0657]]22 move (5, 2) top [[60, 0.1113], [21, 0.0864], [29, 0.0782]]24 move (8, 3) top [[20, 0.1514], [12, 0.1164], [13, 0.1008]]26 move (2, 2) top [[20, 0.1484], [12, 0.1153], [60, 0.1025]]28 move (8, 5) top [[77, 0.9379], [29, 0.0094], [12, 0.0076]]30 move (0, 6) top [[29, 0.1395], [12, 0.1154], [13, 0.1065]]after 31 ......X....X....O...XOO.O.....XOOXO....XOXX....XXOOXO....XOXO.....OOXO.....X.X...32 move (1, 4) top [[29, 0.1478], [12, 0.1128], [38, 0.1011]]34 move (8, 4) top [[12, 0.1912], [70, 0.155], [29, 0.1207]]36 move (7, 2) top [[65, 0.9899], [74, 0.0084], [70, 0.0003]]38 move (8, 2) top [[74, 0.8989], [12, 0.0269], [70, 0.0191]]40 move (1, 6) top [[12, 0.2538], [70, 0.1807], [29, 0.1223]]42 move (1, 3) top [[12, 0.9976], [70, 0.0006], [29, 0.0004]]44 move (0, 3) top [[3, 0.9988], [70, 0.0003], [56, 0.0002]]after 47 ...XOO.O...XXXO.O...XOO.O.....XOO.O...XXO..O...XXOO.O....XOXOO...XOOXOO..OXXXX...after 63 .X.XOOOO.X.XXXO.O..XXOO.O..X.XXOO.O...XXO..OO..XXOO.O...XXOXOOO.XXOOXOO.X.XXXXOO.after 76 .X.XOOOO.X.XXXOOOX.XXOO.O.XX.XXOO.OO.XXXO..OOXXXXOO.O.X.XXOXOOO.XXOOXOOOX.XXXXOO.
GAME 67 white1 move (4, 5) top [[31, 0.0573], [39, 0.0573], [41, 0.0573]]3 move (4, 6) top [[32, 0.0941], [50, 0.077], [42, 0.0733]]5 move (5, 5) top [[32, 0.0851], [50, 0.0809], [49, 0.0769]]7 move (4, 7) top [[49, 0.2713], [33, 0.1153], [39, 0.1081]]9 move (5, 4) top [[49, 0.2668], [39, 0.1235], [33, 0.1136]]11 move (2, 4) top [[39, 0.1935], [33, 0.1309], [22, 0.0925]]13 move (2, 5) top [[39, 0.237], [23, 0.1758], [33, 0.1155]]15 move (2, 2) top [[39, 0.2814], [33, 0.1125], [48, 0.1034]]after 15 ....................OXOO.....XXXX.......XOOO.....OO......XX......................17 move (2, 1) top [[39, 0.5743], [38, 0.067], [33, 0.0653]]19 move (6, 5) top [[39, 0.5392], [33, 0.075], [38, 0.0615]]21 move (4, 3) top [[39, 0.5267], [33, 0.0749], [38, 0.0661]]23 move (4, 2) top [[38, 0.2233], [48, 0.194], [33, 0.1655]]25 move (3, 6) top [[33, 0.9998], [48, 0.0001], [37, 0.0]]27 move (4, 1) top [[47, 0.2574], [37, 0.151], [56, 0.0933]]29 move (3, 0) top [[27, 0.9957], [47, 0.0011], [51, 0.0004]]31 move (6, 2) top [[47, 0.3041], [56, 0.1033], [51, 0.0916]]after 31 ...........XX.X....OOXOOX..OXXXXXO...OOOXOOO..X.XOO.....OXXO.......X.............33 move (4, 0) top [[36, 0.9595], [55, 0.0067], [51, 0.0052]]35 move (1, 1) top [[9, 0.6574], [10, 0.3398], [55, 0.0005]]37 move (1, 0) top [[9, 0.8707], [65, 0.1277], [51, 0.0003]]39 move (7, 2) top [[65, 0.9877], [51, 0.0023], [66, 0.0018]]41 move (3, 0) top [[27, 0.9997], [66, 0.0001], [51, 0.0]]43 move (4, 1) top [[37, 0.2263], [66, 0.1673], [51, 0.1004]]45 move (4, 2) top [[38, 0.7685], [66, 0.0543], [51, 0.0326]]after 47 .X.......OOXX.XX...OOXOOX..OXXXXXOO..OOXXOOO.XXXXOO....XOXXO.....O.X.............after 63 .XXX.....OOXX.XXO..OOXOOXO.OXXXXXOOXX..XXOOO.XXXXOOOOO.XOXXO.XOXOOXXO.X..........after 92 XXXX.OO.O.XXXXXXO.X..X.XXOO.XXXXXOO.X.XXXOOOOXXXXOOOOOXX.XXO.OOX.XXXOOO.X.XXOO.OO

python

import jsonsrc=open('/work/design.py').read()src=src.replace("safe2=1., safe3=1., cap_safety=1.)",                "safe2=1., safe3=1., cap_safety=1., crowd=0., temperature_scale=1.)")src=src.replace("b(5,E,1-len(adj[q]))","b(5,E,2-len(adj[q]))")src=src.replace("""w(6,E,u(4,q),p['eye'])        w(6,I,u(4,q),p['isolation'])        w(6,P,u(5,q),1)""","""w(6,E,u(4,q),1); b(6,u(4,q),-1)        w(6,P,u(5,q),1)        b(6,u(5,q),100)        w(6,E,u(5,q),-p['crowd'])        w(6,I,u(5,q),-p['isolation'])""")src=src.replace("p['save'],-p['danger'],-1.,1.]",                "p['save'],-p['danger'],-(p['eye']-2*p['crowd']),1.]")src=src.replace("a['out.b'][q]=-40.","a['out.b'][q]=-140.")src=src.replace("return a,p","a['out.W'] *= p['temperature_scale']\n    a['out.b'] *= p['temperature_scale']\n    return a,p")open('/work/design_v2.py','w').write(src)exec(src)config=install(dict(own_adj=2.,opp_adj=.6,crowd=3.,temperature_scale=1.5,opening=3.))json.dump(config,open('/work/config_v2.json','w'),indent=2)print(config)

import jsonsrc=open('/work/design.py').read()src=src.replace("safe2=1., safe3=1., cap_safety=1.)",                "safe2=1., safe3=1., cap_safety=1., crowd=0., temperature_scale=1.)")src=src.replace("b(5,E,1-len(adj[q]))","b(5,E,2-len(adj[q]))")src=src.replace("""w(6,E,u(4,q),p['eye'])        w(6,I,u(4,q),p['isolation'])        w(6,P,u(5,q),1)""","""w(6,E,u(4,q),1); b(6,u(4,q),-1)        w(6,P,u(5,q),1)        b(6,u(5,q),100)        w(6,E,u(5,q),-p['crowd'])        w(6,I,u(5,q),-p['isolation'])""")src=src.replace("p['save'],-p['danger'],-1.,1.]",                "p['save'],-p['danger'],-(p['eye']-2*p['crowd']),1.]")src=src.replace("a['out.b'][q]=-40.","a['out.b'][q]=-140.")src=src.replace("return a,p","a['out.W'] *= p['temperature_scale']\n    a['out.b'] *= p['temperature_scale']\n    return a,p")open('/work/design_v2.py','w').write(src)exec(src)config=install(dict(own_adj=2.,opp_adj=.6,crowd=3.,temperature_scale=1.5,opening=3.))json.dump(config,open('/work/config_v2.json','w'),indent=2)print(config)

OUTPUT of Python run 5 of at most 120:

{'capture': 11.0, 'capture_extra': 3.0, 'save': 7.0, 'danger': 10.0, 'eye': 22.0, 'isolation': 3.0, 'isolation_threshold': 2.5, 'base': 1.3, 'pass_bias': -1.0, 'attack': 1.3, 'own_adj': 2.0, 'opp_adj': 0.6, 'opening': 3.0, 'save_danger': 2.0, 'safe2': 1.0, 'safe3': 1.0, 'cap_safety': 1.0, 'crowd': 3.0, 'temperature_scale': 1.5}

{'capture': 11.0, 'capture_extra': 3.0, 'save': 7.0, 'danger': 10.0, 'eye': 22.0, 'isolation': 3.0, 'isolation_threshold': 2.5, 'base': 1.3, 'pass_bias': -1.0, 'attack': 1.3, 'own_adj': 2.0, 'opp_adj': 0.6, 'opening': 3.0, 'save_danger': 2.0, 'safe2': 1.0, 'safe3': 1.0, 'cap_safety': 1.0, 'crowd': 3.0, 'temperature_scale': 1.5}

Games pass 2 of 10 done.

{"ok": true, "pass": 2, "kind": "games", "per_opponent": {"random": {"games": 8, "score": 1.0, "wins": 8, "losses": 0, "score_as_black": 1.0, "score_as_white": 1.0, "mean_margin": 81.0, "mean_length": 128.0, "stones_captured_per_game": 101.375, "stones_lost_per_game": 0.5, "passes_per_game": 1.125, "games_with_a_pass_before_move_60": 0, "end_two_passes": 8, "end_repetition": 0, "end_move_limit": 0}, "strong": {"games": 96, "score": 0.5729166865348816, "wins": 55, "losses": 41, "score_as_black": 0.4166666567325592, "score_as_white": 0.7291666865348816, "mean_margin": 34.52083206176758, "mean_length": 117.09375, "stones_captured_per_game": 58.21875, "stones_lost_per_game": 12.854166666666666, "passes_per_game": 3.4583333333333335, "games_with_a_pass_before_move_60": 0, "end_two_passes": 87, "end_repetition": 6, "end_move_limit": 3}, "weak": {"games": 24, "score": 0.9583333134651184, "wins": 23, "losses": 1, "score_as_black": 1.0, "score_as_white": 0.9166666865348816, "mean_margin": 77.91666412353516, "mean_length": 131.16666666666666, "stones_captured_per_game": 92.45833333333333, "stones_lost_per_game": 3.0, "passes_per_game": 3.25, "games_with_a_pass_before_move_60": 0, "end_two_passes": 22, "end_repetition": 1, "end_move_limit": 1}}, "files": "/work/pass_2/", "readme": "games.json: one entry per game (opponent, net_color, result +1/-1, margin = net's area score minus the opponent's incl. komi, end, stones captured / lost, moves: ply, by, move (row*9+col, 81 = pass), and for the network's moves its top-5 [move, probability]). boards.npy: (games, 162, 81) int8, the board AFTER each ply (1 Black, -1 White, 0 empty; index row*9+col); valid.npy: (games, 162) which plies were played."}

{"ok": true, "pass": 2, "kind": "games", "per_opponent": {"random": {"games": 8, "score": 1.0, "wins": 8, "losses": 0, "score_as_black": 1.0, "score_as_white": 1.0, "mean_margin": 81.0, "mean_length": 128.0, "stones_captured_per_game": 101.375, "stones_lost_per_game": 0.5, "passes_per_game": 1.125, "games_with_a_pass_before_move_60": 0, "end_two_passes": 8, "end_repetition": 0, "end_move_limit": 0}, "strong": {"games": 96, "score": 0.5729166865348816, "wins": 55, "losses": 41, "score_as_black": 0.4166666567325592, "score_as_white": 0.7291666865348816, "mean_margin": 34.52083206176758, "mean_length": 117.09375, "stones_captured_per_game": 58.21875, "stones_lost_per_game": 12.854166666666666, "passes_per_game": 3.4583333333333335, "games_with_a_pass_before_move_60": 0, "end_two_passes": 87, "end_repetition": 6, "end_move_limit": 3}, "weak": {"games": 24, "score": 0.9583333134651184, "wins": 23, "losses": 1, "score_as_black": 1.0, "score_as_white": 0.9166666865348816, "mean_margin": 77.91666412353516, "mean_length": 131.16666666666666, "stones_captured_per_game": 92.45833333333333, "stones_lost_per_game": 3.0, "passes_per_game": 3.25, "games_with_a_pass_before_move_60": 0, "end_two_passes": 22, "end_repetition": 1, "end_move_limit": 1}}, "files": "/work/pass_2/", "readme": "games.json: one entry per game (opponent, net_color, result +1/-1, margin = net's area score minus the opponent's incl. komi, end, stones captured / lost, moves: ply, by, move (row*9+col, 81 = pass), and for the network's moves its top-5 [move, probability]). boards.npy: (games, 162, 81) int8, the board AFTER each ply (1 Black, -1 White, 0 empty; index row*9+col); valid.npy: (games, 162) which plies were played."}

Passes left: 8.

python

import json,numpy as npg=json.load(open('/work/pass_2/games.json')); boards=np.load('/work/pass_2/boards.npy')def show(b): return '\n'.join(''.join({0:'.',1:'X',-1:'O'}[int(v)] for v in row) for row in np.asarray(b).reshape(9,9))loss=[i for i,x in enumerate(g) if x['opponent']=='strong' and x['result']<0]print('Strong losses',[(i,g[i]['margin'],g[i]['end'],g[i]['stones_lost_by_net']) for i in loss])for gi in loss[:3]:    x=g[gi]    print('\nGAME',gi,x['net_color'])    for ply,m in enumerate(x['moves']):        if ply<38 and m['by']=='net': print(ply,m['move'],m['top5'][:3])        if ply in [15,31,47,63,len(x['moves'])-1]: print('after',ply,'\n'+show(boards[gi,ply]))for gi,x in enumerate(g):    if x['end']=='repetition':        print('\nREPETITION',gi,x['result'],x['net_color'],x['margin'])        for m in x['moves'][-9:]: print(m)        print(show(boards[gi,len(x['moves'])-4]))        break

import json,numpy as npg=json.load(open('/work/pass_2/games.json')); boards=np.load('/work/pass_2/boards.npy')def show(b): return '\n'.join(''.join({0:'.',1:'X',-1:'O'}[int(v)] for v in row) for row in np.asarray(b).reshape(9,9))loss=[i for i,x in enumerate(g) if x['opponent']=='strong' and x['result']<0]print('Strong losses',[(i,g[i]['margin'],g[i]['end'],g[i]['stones_lost_by_net']) for i in loss])for gi in loss[:3]:    x=g[gi]    print('\nGAME',gi,x['net_color'])    for ply,m in enumerate(x['moves']):        if ply<38 and m['by']=='net': print(ply,m['move'],m['top5'][:3])        if ply in [15,31,47,63,len(x['moves'])-1]: print('after',ply,'\n'+show(boards[gi,ply]))for gi,x in enumerate(g):    if x['end']=='repetition':        print('\nREPETITION',gi,x['result'],x['net_color'],x['margin'])        for m in x['moves'][-9:]: print(m)        print(show(boards[gi,len(x['moves'])-4]))        break

OUTPUT of Python run 6 of at most 120:

Strong losses [(8, -1.5, 'two passes', 22), (10, -18.5, 'two passes', 16), (13, -1.5, 'two passes', 17), (14, -34.5, 'two passes', 30), (16, -23.5, 'two passes', 41), (18, -25.5, 'two passes', 25), (19, -20.5, 'two passes', 11), (22, -24.5, 'two passes', 19), (23, -20.5, 'two passes', 28), (25, -32.5, 'two passes', 18), (26, -4.5, 'repetition', 14), (30, -34.5, 'two passes', 23), (33, -15.5, 'two passes', 15), (34, -21.5, 'two passes', 18), (35, -23.5, 'two passes', 14), (36, -23.5, 'two passes', 11), (40, -20.5, 'two passes', 9), (41, 26.5, 'repetition', 4), (43, -8.5, 'two passes', 12), (44, -13.5, 'two passes', 8), (45, -26.5, 'two passes', 15), (47, -2.5, 'two passes', 31), (48, -16.5, 'two passes', 28), (49, -18.5, 'two passes', 8), (50, -19.5, 'two passes', 9), (52, 10.5, 'repetition', 9), (54, -24.5, 'two passes', 36), (55, -14.5, 'two passes', 9), (58, 46.5, 'repetition', 3), (62, -1.5, 'two passes', 20), (65, -25.5, 'two passes', 32), (68, -29.5, 'two passes', 42), (74, -8.5, 'two passes', 19), (78, -17.5, 'two passes', 32), (83, -4.5, 'two passes', 16), (84, -20.5, 'two passes', 25), (85, -3.5, 'two passes', 26), (88, -49.5, 'two passes', 73), (91, -14.5, 'repetition', 24), (96, -17.5, 'repetition', 42), (101, -11.5, 'two passes', 23)]
GAME 8 black0 41 [[40, 0.1021], [31, 0.0728], [39, 0.0728]]2 32 [[32, 0.2743], [50, 0.2743], [42, 0.2246]]4 39 [[50, 0.1892], [39, 0.16], [49, 0.1485]]6 48 [[30, 0.7037], [49, 0.0929], [48, 0.0566]]8 38 [[30, 0.8021], [42, 0.029], [38, 0.0232]]10 30 [[30, 0.4964], [47, 0.3622], [57, 0.0344]]12 47 [[47, 0.4183], [59, 0.279], [29, 0.2067]]14 42 [[29, 0.3382], [59, 0.3233], [56, 0.1913]]after 15 .......................O......XOXO....XXOXX....XXOO......OO......................16 51 [[51, 0.3109], [29, 0.2597], [59, 0.2074]]18 29 [[29, 0.2898], [21, 0.2314], [59, 0.1714]]20 21 [[21, 0.2857], [59, 0.2459], [24, 0.1891]]22 56 [[59, 0.318], [20, 0.304], [56, 0.2539]]24 52 [[52, 0.9996], [20, 0.0003], [34, 0.0]]26 66 [[20, 0.4276], [61, 0.2713], [34, 0.195]]28 67 [[67, 0.3197], [61, 0.2657], [34, 0.184]]30 34 [[61, 0.322], [65, 0.2148], [34, 0.2084]]after 31 ....................OXOOOO...XXOXOX...XXOXXO...XXOOXX...XOOOO.....XXO............32 44 [[44, 0.969], [35, 0.0308], [61, 0.0001]]34 61 [[65, 0.4443], [61, 0.3443], [13, 0.0285]]36 13 [[65, 0.6775], [13, 0.0435], [12, 0.0404]]after 47 ...X.O.....X.XXO....OXOOOOX..XXOXOX...XXOXX.X..XXOOXX...XOOOOX....XXOOOX.........after 63 ...XX.X....X.XXOOO.X.XOOOO..XXXOXO.O.XXXOXXO..XXXOOXXO..XOOOOXX..XXXOOOX.....O.O.after 113 XX.XXOOO.XXX.XXOOO.X.XOOOOOXXXXOXOOO.XXXOX.OOXXXXOO..OXXXOOOOOO.XXXXOOOOX.X.XOOOO
GAME 10 black0 30 [[40, 0.1021], [31, 0.0728], [39, 0.0728]]2 39 [[31, 0.3896], [39, 0.3896], [21, 0.0583]]4 48 [[31, 0.4146], [48, 0.3269], [38, 0.0891]]6 31 [[31, 0.8345], [57, 0.0448], [41, 0.0374]]8 22 [[32, 0.1506], [41, 0.1401], [50, 0.1356]]10 41 [[21, 0.5675], [41, 0.1163], [50, 0.0832]]12 50 [[50, 0.6894], [21, 0.1855], [42, 0.0328]]14 21 [[21, 0.5772], [42, 0.1008], [51, 0.0846]]after 15 .....................XX.......XXOO.....XOX......XOX.....OOOO.....................16 67 [[42, 0.2071], [51, 0.176], [23, 0.1658]]18 29 [[42, 0.1886], [51, 0.1672], [23, 0.1529]]20 38 [[38, 0.503], [20, 0.1272], [51, 0.1243]]22 20 [[47, 0.6433], [66, 0.1299], [20, 0.0836]]24 46 [[47, 0.634], [23, 0.1669], [66, 0.1631]]26 66 [[47, 0.684], [66, 0.1231], [23, 0.1148]]28 47 [[47, 0.821], [23, 0.0976], [37, 0.0487]]30 24 [[24, 0.4564], [37, 0.3137], [64, 0.0876]]after 31 ....................XXXOXO...XXXOO....XXOXO...XXXOX....OOOOOO....OXXOO...........32 15 [[15, 0.9994], [37, 0.0004], [64, 0.0001]]34 37 [[37, 0.5084], [14, 0.2588], [43, 0.0441]]36 28 [[34, 0.3574], [14, 0.3316], [28, 0.1714]]after 47 .....XXO.....OOXOO.XXXXOXO..XXXXOOXX.XXXOXOO..XXXOX.O..OOOOOO...OOXXOO...........after 63 .O..O..O.OXXXOO.OOOXXXXO.O.XXXXXOOXX.XXXOXOOOXXXXOX.OO.OOOOOO...OOXXOO....X.XX...after 102 X.XXO..O.XXXXOOOOO.XXXXOOO.XXXXXOOXX.XXXO.OOOXXXXOOOOOXOOOOOOOOXOOOXOOOOOOOO.XXXO
GAME 13 black0 42 [[40, 0.1021], [31, 0.0728], [39, 0.0728]]2 33 [[41, 0.6113], [33, 0.0914], [51, 0.0914]]4 41 [[41, 0.4722], [32, 0.1632], [51, 0.1458]]6 32 [[32, 0.9234], [50, 0.0384], [51, 0.0149]]8 30 [[50, 0.405], [39, 0.1626], [51, 0.1011]]10 22 [[50, 0.3237], [48, 0.2471], [51, 0.0814]]12 29 [[29, 0.9991], [23, 0.0006], [50, 0.0001]]14 13 [[13, 0.9958], [38, 0.0017], [20, 0.0008]]after 15 .............X......OOXO.....XXOXX.....OOXX......O........O......................16 51 [[38, 0.5617], [24, 0.1266], [50, 0.0857]]18 14 [[50, 0.6594], [38, 0.2206], [24, 0.0437]]20 24 [[24, 1.0], [50, 0.0], [38, 0.0]]22 50 [[50, 0.6677], [38, 0.2948], [15, 0.0137]]24 38 [[38, 0.7332], [60, 0.1572], [15, 0.0429]]26 47 [[47, 0.6784], [60, 0.186], [15, 0.054]]28 28 [[46, 0.4999], [28, 0.4999], [60, 0.0001]]30 55 [[60, 0.4565], [19, 0.2951], [15, 0.1502]]after 31 ...........OOXX.....OOX.X...XXXOXX...OXOOXX...OXOOXX...XOOOO.....................32 15 [[60, 0.5083], [19, 0.2588], [15, 0.1486]]34 60 [[60, 0.5681], [19, 0.2893], [54, 0.0531]]36 34 [[19, 0.5592], [54, 0.1494], [27, 0.1248]]after 47 ...........OOXXXX..OOOX.XX.O...OXXX.OO.OOXXXOOO.OOXXX.XXOOOOXO.......OO..........after 63 ...X.OX....OOXXXXO.OOOX.XXOO...OXXX.OO.OOXXXOOO.OOXXX.XXOOOOXO.XXO.OXOO..X.O.XXO.after 103 OXXXX.X.X.OOOXXXXXOOOOX.XX.OO..OXXXXOO.OOXXX.OO.OOXXXX.OOOOOXX.XOOOOX.XXOOOOXXX.X
REPETITION 26 -1 black -4.5{'ply': 82, 'by': 'net', 'move': 1, 'top5': [[1, 0.8154], [81, 0.141], [67, 0.0174], [70, 0.0116], [0, 0.0105]]}{'ply': 83, 'by': 'opponent', 'move': 71}{'ply': 84, 'by': 'net', 'move': 81, 'top5': [[81, 0.873], [67, 0.1077], [32, 0.0157], [68, 0.0036], [72, 0.0]]}{'ply': 85, 'by': 'opponent', 'move': 53}{'ply': 86, 'by': 'net', 'move': 67, 'top5': [[81, 0.873], [67, 0.1077], [32, 0.0157], [68, 0.0036], [72, 0.0]]}{'ply': 87, 'by': 'opponent', 'move': 76}{'ply': 88, 'by': 'net', 'move': 72, 'top5': [[72, 1.0], [81, 0.0], [32, 0.0], [68, 0.0], [46, 0.0]]}{'ply': 89, 'by': 'opponent', 'move': 54}{'ply': 90, 'by': 'net', 'move': 63, 'top5': [[63, 1.0], [81, 0.0], [32, 0.0], [46, 0.0], [72, 0.0]]}.XX.XXOO.XXXX.XOOO.XXXXXXOOXXXXX.XOO.XXXOXOOOX.XXOXOOO.XOOOXOOOXOOOX.O.O.OOOOOO.O

Strong losses [(8, -1.5, 'two passes', 22), (10, -18.5, 'two passes', 16), (13, -1.5, 'two passes', 17), (14, -34.5, 'two passes', 30), (16, -23.5, 'two passes', 41), (18, -25.5, 'two passes', 25), (19, -20.5, 'two passes', 11), (22, -24.5, 'two passes', 19), (23, -20.5, 'two passes', 28), (25, -32.5, 'two passes', 18), (26, -4.5, 'repetition', 14), (30, -34.5, 'two passes', 23), (33, -15.5, 'two passes', 15), (34, -21.5, 'two passes', 18), (35, -23.5, 'two passes', 14), (36, -23.5, 'two passes', 11), (40, -20.5, 'two passes', 9), (41, 26.5, 'repetition', 4), (43, -8.5, 'two passes', 12), (44, -13.5, 'two passes', 8), (45, -26.5, 'two passes', 15), (47, -2.5, 'two passes', 31), (48, -16.5, 'two passes', 28), (49, -18.5, 'two passes', 8), (50, -19.5, 'two passes', 9), (52, 10.5, 'repetition', 9), (54, -24.5, 'two passes', 36), (55, -14.5, 'two passes', 9), (58, 46.5, 'repetition', 3), (62, -1.5, 'two passes', 20), (65, -25.5, 'two passes', 32), (68, -29.5, 'two passes', 42), (74, -8.5, 'two passes', 19), (78, -17.5, 'two passes', 32), (83, -4.5, 'two passes', 16), (84, -20.5, 'two passes', 25), (85, -3.5, 'two passes', 26), (88, -49.5, 'two passes', 73), (91, -14.5, 'repetition', 24), (96, -17.5, 'repetition', 42), (101, -11.5, 'two passes', 23)]
GAME 8 black0 41 [[40, 0.1021], [31, 0.0728], [39, 0.0728]]2 32 [[32, 0.2743], [50, 0.2743], [42, 0.2246]]4 39 [[50, 0.1892], [39, 0.16], [49, 0.1485]]6 48 [[30, 0.7037], [49, 0.0929], [48, 0.0566]]8 38 [[30, 0.8021], [42, 0.029], [38, 0.0232]]10 30 [[30, 0.4964], [47, 0.3622], [57, 0.0344]]12 47 [[47, 0.4183], [59, 0.279], [29, 0.2067]]14 42 [[29, 0.3382], [59, 0.3233], [56, 0.1913]]after 15 .......................O......XOXO....XXOXX....XXOO......OO......................16 51 [[51, 0.3109], [29, 0.2597], [59, 0.2074]]18 29 [[29, 0.2898], [21, 0.2314], [59, 0.1714]]20 21 [[21, 0.2857], [59, 0.2459], [24, 0.1891]]22 56 [[59, 0.318], [20, 0.304], [56, 0.2539]]24 52 [[52, 0.9996], [20, 0.0003], [34, 0.0]]26 66 [[20, 0.4276], [61, 0.2713], [34, 0.195]]28 67 [[67, 0.3197], [61, 0.2657], [34, 0.184]]30 34 [[61, 0.322], [65, 0.2148], [34, 0.2084]]after 31 ....................OXOOOO...XXOXOX...XXOXXO...XXOOXX...XOOOO.....XXO............32 44 [[44, 0.969], [35, 0.0308], [61, 0.0001]]34 61 [[65, 0.4443], [61, 0.3443], [13, 0.0285]]36 13 [[65, 0.6775], [13, 0.0435], [12, 0.0404]]after 47 ...X.O.....X.XXO....OXOOOOX..XXOXOX...XXOXX.X..XXOOXX...XOOOOX....XXOOOX.........after 63 ...XX.X....X.XXOOO.X.XOOOO..XXXOXO.O.XXXOXXO..XXXOOXXO..XOOOOXX..XXXOOOX.....O.O.after 113 XX.XXOOO.XXX.XXOOO.X.XOOOOOXXXXOXOOO.XXXOX.OOXXXXOO..OXXXOOOOOO.XXXXOOOOX.X.XOOOO
GAME 10 black0 30 [[40, 0.1021], [31, 0.0728], [39, 0.0728]]2 39 [[31, 0.3896], [39, 0.3896], [21, 0.0583]]4 48 [[31, 0.4146], [48, 0.3269], [38, 0.0891]]6 31 [[31, 0.8345], [57, 0.0448], [41, 0.0374]]8 22 [[32, 0.1506], [41, 0.1401], [50, 0.1356]]10 41 [[21, 0.5675], [41, 0.1163], [50, 0.0832]]12 50 [[50, 0.6894], [21, 0.1855], [42, 0.0328]]14 21 [[21, 0.5772], [42, 0.1008], [51, 0.0846]]after 15 .....................XX.......XXOO.....XOX......XOX.....OOOO.....................16 67 [[42, 0.2071], [51, 0.176], [23, 0.1658]]18 29 [[42, 0.1886], [51, 0.1672], [23, 0.1529]]20 38 [[38, 0.503], [20, 0.1272], [51, 0.1243]]22 20 [[47, 0.6433], [66, 0.1299], [20, 0.0836]]24 46 [[47, 0.634], [23, 0.1669], [66, 0.1631]]26 66 [[47, 0.684], [66, 0.1231], [23, 0.1148]]28 47 [[47, 0.821], [23, 0.0976], [37, 0.0487]]30 24 [[24, 0.4564], [37, 0.3137], [64, 0.0876]]after 31 ....................XXXOXO...XXXOO....XXOXO...XXXOX....OOOOOO....OXXOO...........32 15 [[15, 0.9994], [37, 0.0004], [64, 0.0001]]34 37 [[37, 0.5084], [14, 0.2588], [43, 0.0441]]36 28 [[34, 0.3574], [14, 0.3316], [28, 0.1714]]after 47 .....XXO.....OOXOO.XXXXOXO..XXXXOOXX.XXXOXOO..XXXOX.O..OOOOOO...OOXXOO...........after 63 .O..O..O.OXXXOO.OOOXXXXO.O.XXXXXOOXX.XXXOXOOOXXXXOX.OO.OOOOOO...OOXXOO....X.XX...after 102 X.XXO..O.XXXXOOOOO.XXXXOOO.XXXXXOOXX.XXXO.OOOXXXXOOOOOXOOOOOOOOXOOOXOOOOOOOO.XXXO
GAME 13 black0 42 [[40, 0.1021], [31, 0.0728], [39, 0.0728]]2 33 [[41, 0.6113], [33, 0.0914], [51, 0.0914]]4 41 [[41, 0.4722], [32, 0.1632], [51, 0.1458]]6 32 [[32, 0.9234], [50, 0.0384], [51, 0.0149]]8 30 [[50, 0.405], [39, 0.1626], [51, 0.1011]]10 22 [[50, 0.3237], [48, 0.2471], [51, 0.0814]]12 29 [[29, 0.9991], [23, 0.0006], [50, 0.0001]]14 13 [[13, 0.9958], [38, 0.0017], [20, 0.0008]]after 15 .............X......OOXO.....XXOXX.....OOXX......O........O......................16 51 [[38, 0.5617], [24, 0.1266], [50, 0.0857]]18 14 [[50, 0.6594], [38, 0.2206], [24, 0.0437]]20 24 [[24, 1.0], [50, 0.0], [38, 0.0]]22 50 [[50, 0.6677], [38, 0.2948], [15, 0.0137]]24 38 [[38, 0.7332], [60, 0.1572], [15, 0.0429]]26 47 [[47, 0.6784], [60, 0.186], [15, 0.054]]28 28 [[46, 0.4999], [28, 0.4999], [60, 0.0001]]30 55 [[60, 0.4565], [19, 0.2951], [15, 0.1502]]after 31 ...........OOXX.....OOX.X...XXXOXX...OXOOXX...OXOOXX...XOOOO.....................32 15 [[60, 0.5083], [19, 0.2588], [15, 0.1486]]34 60 [[60, 0.5681], [19, 0.2893], [54, 0.0531]]36 34 [[19, 0.5592], [54, 0.1494], [27, 0.1248]]after 47 ...........OOXXXX..OOOX.XX.O...OXXX.OO.OOXXXOOO.OOXXX.XXOOOOXO.......OO..........after 63 ...X.OX....OOXXXXO.OOOX.XXOO...OXXX.OO.OOXXXOOO.OOXXX.XXOOOOXO.XXO.OXOO..X.O.XXO.after 103 OXXXX.X.X.OOOXXXXXOOOOX.XX.OO..OXXXXOO.OOXXX.OO.OOXXXX.OOOOOXX.XOOOOX.XXOOOOXXX.X
REPETITION 26 -1 black -4.5{'ply': 82, 'by': 'net', 'move': 1, 'top5': [[1, 0.8154], [81, 0.141], [67, 0.0174], [70, 0.0116], [0, 0.0105]]}{'ply': 83, 'by': 'opponent', 'move': 71}{'ply': 84, 'by': 'net', 'move': 81, 'top5': [[81, 0.873], [67, 0.1077], [32, 0.0157], [68, 0.0036], [72, 0.0]]}{'ply': 85, 'by': 'opponent', 'move': 53}{'ply': 86, 'by': 'net', 'move': 67, 'top5': [[81, 0.873], [67, 0.1077], [32, 0.0157], [68, 0.0036], [72, 0.0]]}{'ply': 87, 'by': 'opponent', 'move': 76}{'ply': 88, 'by': 'net', 'move': 72, 'top5': [[72, 1.0], [81, 0.0], [32, 0.0], [68, 0.0], [46, 0.0]]}{'ply': 89, 'by': 'opponent', 'move': 54}{'ply': 90, 'by': 'net', 'move': 63, 'top5': [[63, 1.0], [81, 0.0], [32, 0.0], [46, 0.0], [72, 0.0]]}.XX.XXOO.XXXX.XOOO.XXXXXXOOXXXXX.XOO.XXXOXOOOX.XXOXOOO.XOOOXOOOXOOOX.O.O.OOOOOO.O

python

import jsonsrc=open('/work/design_v2.py').read()src=src.replace("temperature_scale=1.)","temperature_scale=1., defend=0., first_center=0.)")src=src.replace("w(1,2,487,1)","""w(1,2,487,1)    b(1,488,1)    for q in range(81):        w(1,3*q,488,-1)        w(1,3*q+1,488,-1)""")src=src.replace("w(k,486,486,1); w(k,487,487,1)",                "w(k,486,486,1); w(k,487,487,1); w(k,488,488,1)")src=src.replace("w(5,486,P,p['opening']*central)","""w(5,486,P,p['opening']*central)        if q==40: w(5,488,P,p['first_center'])""")src=src.replace("w(5,u(3,t),P,-p['attack'])","""w(5,u(3,t),P,-p['attack'])            w(5,u(4,t),P,p['defend'])            w(5,u(2,t),P,-p['defend'])""")open('/work/design_v3.py','w').write(src)exec(src)config=install(dict(own_adj=2.,opp_adj=1.5,crowd=3.,temperature_scale=1.5,                    opening=3.,defend=1.,first_center=20.))json.dump(config,open('/work/config_v3.json','w'),indent=2)print(config)

import jsonsrc=open('/work/design_v2.py').read()src=src.replace("temperature_scale=1.)","temperature_scale=1., defend=0., first_center=0.)")src=src.replace("w(1,2,487,1)","""w(1,2,487,1)    b(1,488,1)    for q in range(81):        w(1,3*q,488,-1)        w(1,3*q+1,488,-1)""")src=src.replace("w(k,486,486,1); w(k,487,487,1)",                "w(k,486,486,1); w(k,487,487,1); w(k,488,488,1)")src=src.replace("w(5,486,P,p['opening']*central)","""w(5,486,P,p['opening']*central)        if q==40: w(5,488,P,p['first_center'])""")src=src.replace("w(5,u(3,t),P,-p['attack'])","""w(5,u(3,t),P,-p['attack'])            w(5,u(4,t),P,p['defend'])            w(5,u(2,t),P,-p['defend'])""")open('/work/design_v3.py','w').write(src)exec(src)config=install(dict(own_adj=2.,opp_adj=1.5,crowd=3.,temperature_scale=1.5,                    opening=3.,defend=1.,first_center=20.))json.dump(config,open('/work/config_v3.json','w'),indent=2)print(config)

OUTPUT of Python run 7 of at most 120:

{'capture': 11.0, 'capture_extra': 3.0, 'save': 7.0, 'danger': 10.0, 'eye': 22.0, 'isolation': 3.0, 'isolation_threshold': 2.5, 'base': 1.3, 'pass_bias': -1.0, 'attack': 1.3, 'own_adj': 2.0, 'opp_adj': 1.5, 'opening': 3.0, 'save_danger': 2.0, 'safe2': 1.0, 'safe3': 1.0, 'cap_safety': 1.0, 'crowd': 3.0, 'temperature_scale': 1.5, 'defend': 1.0, 'first_center': 20.0}

{'capture': 11.0, 'capture_extra': 3.0, 'save': 7.0, 'danger': 10.0, 'eye': 22.0, 'isolation': 3.0, 'isolation_threshold': 2.5, 'base': 1.3, 'pass_bias': -1.0, 'attack': 1.3, 'own_adj': 2.0, 'opp_adj': 1.5, 'opening': 3.0, 'save_danger': 2.0, 'safe2': 1.0, 'safe3': 1.0, 'cap_safety': 1.0, 'crowd': 3.0, 'temperature_scale': 1.5, 'defend': 1.0, 'first_center': 20.0}

Games pass 3 of 10 done.

{"ok": true, "pass": 3, "kind": "games", "per_opponent": {"random": {"games": 8, "score": 1.0, "wins": 8, "losses": 0, "score_as_black": 1.0, "score_as_white": 1.0, "mean_margin": 79.375, "mean_length": 130.25, "stones_captured_per_game": 100.375, "stones_lost_per_game": 1.0, "passes_per_game": 1.875, "games_with_a_pass_before_move_60": 0, "end_two_passes": 8, "end_repetition": 0, "end_move_limit": 0}, "strong": {"games": 96, "score": 0.5520833134651184, "wins": 53, "losses": 43, "score_as_black": 0.4583333432674408, "score_as_white": 0.6458333134651184, "mean_margin": 35.35416793823242, "mean_length": 110.35416666666667, "stones_captured_per_game": 47.5, "stones_lost_per_game": 11.322916666666666, "passes_per_game": 1.875, "games_with_a_pass_before_move_60": 0, "end_two_passes": 91, "end_repetition": 5, "end_move_limit": 0}, "weak": {"games": 24, "score": 1.0, "wins": 24, "losses": 0, "score_as_black": 1.0, "score_as_white": 1.0, "mean_margin": 81.0, "mean_length": 127.25, "stones_captured_per_game": 88.75, "stones_lost_per_game": 2.1666666666666665, "passes_per_game": 1.5, "games_with_a_pass_before_move_60": 0, "end_two_passes": 24, "end_repetition": 0, "end_move_limit": 0}}, "files": "/work/pass_3/", "readme": "games.json: one entry per game (opponent, net_color, result +1/-1, margin = net's area score minus the opponent's incl. komi, end, stones captured / lost, moves: ply, by, move (row*9+col, 81 = pass), and for the network's moves its top-5 [move, probability]). boards.npy: (games, 162, 81) int8, the board AFTER each ply (1 Black, -1 White, 0 empty; index row*9+col); valid.npy: (games, 162) which plies were played."}

{"ok": true, "pass": 3, "kind": "games", "per_opponent": {"random": {"games": 8, "score": 1.0, "wins": 8, "losses": 0, "score_as_black": 1.0, "score_as_white": 1.0, "mean_margin": 79.375, "mean_length": 130.25, "stones_captured_per_game": 100.375, "stones_lost_per_game": 1.0, "passes_per_game": 1.875, "games_with_a_pass_before_move_60": 0, "end_two_passes": 8, "end_repetition": 0, "end_move_limit": 0}, "strong": {"games": 96, "score": 0.5520833134651184, "wins": 53, "losses": 43, "score_as_black": 0.4583333432674408, "score_as_white": 0.6458333134651184, "mean_margin": 35.35416793823242, "mean_length": 110.35416666666667, "stones_captured_per_game": 47.5, "stones_lost_per_game": 11.322916666666666, "passes_per_game": 1.875, "games_with_a_pass_before_move_60": 0, "end_two_passes": 91, "end_repetition": 5, "end_move_limit": 0}, "weak": {"games": 24, "score": 1.0, "wins": 24, "losses": 0, "score_as_black": 1.0, "score_as_white": 1.0, "mean_margin": 81.0, "mean_length": 127.25, "stones_captured_per_game": 88.75, "stones_lost_per_game": 2.1666666666666665, "passes_per_game": 1.5, "games_with_a_pass_before_move_60": 0, "end_two_passes": 24, "end_repetition": 0, "end_move_limit": 0}}, "files": "/work/pass_3/", "readme": "games.json: one entry per game (opponent, net_color, result +1/-1, margin = net's area score minus the opponent's incl. komi, end, stones captured / lost, moves: ply, by, move (row*9+col, 81 = pass), and for the network's moves its top-5 [move, probability]). boards.npy: (games, 162, 81) int8, the board AFTER each ply (1 Black, -1 White, 0 empty; index row*9+col); valid.npy: (games, 162) which plies were played."}

Passes left: 7.

python

import numpy as np,json,collectionsg=json.load(open('/work/pass_3/games.json')); boards=np.load('/work/pass_3/boards.npy')adj=[[j for j in (q-9,q+9,q-1,q+1) if 0<=j<81 and abs(j//9-q//9)+abs(j%9-q%9)==1] for q in range(81)]def groups(b):    info={}    for q in range(81):        if b[q]==0 or q in info: continue        stack=[q]; stones={q}; libs=set()        while stack:            t=stack.pop()            for v in adj[t]:                if b[v]==0: libs.add(v)                elif b[v]==b[q] and v not in stones: stones.add(v);stack.append(v)        for t in stones: info[t]=(stones,libs)    return infobad=[]; stats=collections.Counter()for gi,x in enumerate(g):    if x['opponent']!='strong':continue    color=1 if x['net_color']=='black' else -1    for pi,m in enumerate(x['moves']):        if m['by']!='net' or m['move']==81:continue        before=np.zeros(81,np.int8) if pi==0 else boards[gi,pi-1]        after=boards[gi,pi]; gr=groups(before);q=m['move']        caps=sum(before==-color)-sum(after==-color)        save=any(before[t]==color and len(gr[t][1])==1 for t in adj[q])        atari_after=groups(after).get(q,(set(),set()))[1]        stats['moves']+=1        if len(atari_after)==1:            stats['self_atari']+=1            if caps==0:stats['noncapture_self_atari']+=1        if m['top5'][0][0]==q and m['top5'][0][1]>.95 and caps==0 and not save:            bad.append((gi,pi,q,len(atari_after),x['result'],m['top5'][:2]))print(stats)print('Confident, no capture or atari save:',len(bad))print(bad[:30])for gi,pi,q,libs,res,top in bad[:4]:    print('\n',gi,pi,q,libs,res,top)    print('\n'.join(''.join({0:'.',1:'X',-1:'O'}[int(v)] for v in row) for row in boards[gi,pi-1].reshape(9,9)))

import numpy as np,json,collectionsg=json.load(open('/work/pass_3/games.json')); boards=np.load('/work/pass_3/boards.npy')adj=[[j for j in (q-9,q+9,q-1,q+1) if 0<=j<81 and abs(j//9-q//9)+abs(j%9-q%9)==1] for q in range(81)]def groups(b):    info={}    for q in range(81):        if b[q]==0 or q in info: continue        stack=[q]; stones={q}; libs=set()        while stack:            t=stack.pop()            for v in adj[t]:                if b[v]==0: libs.add(v)                elif b[v]==b[q] and v not in stones: stones.add(v);stack.append(v)        for t in stones: info[t]=(stones,libs)    return infobad=[]; stats=collections.Counter()for gi,x in enumerate(g):    if x['opponent']!='strong':continue    color=1 if x['net_color']=='black' else -1    for pi,m in enumerate(x['moves']):        if m['by']!='net' or m['move']==81:continue        before=np.zeros(81,np.int8) if pi==0 else boards[gi,pi-1]        after=boards[gi,pi]; gr=groups(before);q=m['move']        caps=sum(before==-color)-sum(after==-color)        save=any(before[t]==color and len(gr[t][1])==1 for t in adj[q])        atari_after=groups(after).get(q,(set(),set()))[1]        stats['moves']+=1        if len(atari_after)==1:            stats['self_atari']+=1            if caps==0:stats['noncapture_self_atari']+=1        if m['top5'][0][0]==q and m['top5'][0][1]>.95 and caps==0 and not save:            bad.append((gi,pi,q,len(atari_after),x['result'],m['top5'][:2]))print(stats)print('Confident, no capture or atari save:',len(bad))print(bad[:30])for gi,pi,q,libs,res,top in bad[:4]:    print('\n',gi,pi,q,libs,res,top)    print('\n'.join(''.join({0:'.',1:'X',-1:'O'}[int(v)] for v in row) for row in boards[gi,pi-1].reshape(9,9)))

OUTPUT of Python run 8 of at most 120:

Counter({'moves': 5123, 'self_atari': 251, 'noncapture_self_atari': 141})Confident, no capture or atari save: 605[(8, 0, 40, 4, -1, [[40, 1.0], [31, 0.0]]), (8, 46, 5, 2, -1, [[5, 0.9883], [4, 0.0113]]), (8, 62, 37, 8, -1, [[37, 0.9565], [46, 0.0257]]), (8, 64, 46, 8, -1, [[46, 0.9982], [36, 0.0009]]), (8, 66, 55, 8, -1, [[55, 0.9827], [36, 0.0159]]), (8, 72, 27, 1, -1, [[27, 1.0], [77, 0.0]]), (8, 74, 77, 2, -1, [[77, 0.9861], [66, 0.0105]]), (8, 76, 76, 3, -1, [[76, 0.9999], [67, 0.0001]]), (8, 78, 67, 2, -1, [[67, 0.9958], [75, 0.004]]), (8, 82, 2, 1, -1, [[2, 1.0], [78, 0.0]]), (8, 84, 78, 2, -1, [[78, 0.9661], [77, 0.0333]]), (8, 86, 79, 1, -1, [[79, 1.0], [77, 0.0]]), (8, 88, 76, 2, -1, [[76, 0.9575], [81, 0.0412]]), (8, 90, 67, 1, -1, [[67, 1.0], [81, 0.0]]), (9, 0, 40, 4, 1, [[40, 1.0], [31, 0.0]]), (9, 26, 67, 1, 1, [[67, 1.0], [38, 0.0]]), (9, 58, 35, 15, 1, [[35, 0.9978], [2, 0.0022]]), (9, 68, 8, 1, 1, [[8, 0.9996], [66, 0.0003]]), (9, 100, 25, 23, 1, [[25, 0.952], [23, 0.0125]]), (9, 106, 15, 22, 1, [[15, 0.9964], [69, 0.0027]]), (10, 0, 40, 4, -1, [[40, 1.0], [31, 0.0]]), (10, 38, 43, 2, -1, [[43, 1.0], [20, 0.0]]), (10, 56, 16, 6, -1, [[16, 0.9638], [28, 0.0151]]), (10, 62, 54, 1, -1, [[54, 1.0], [19, 0.0]]), (10, 66, 76, 1, -1, [[76, 1.0], [19, 0.0]]), (10, 70, 27, 6, -1, [[27, 1.0], [19, 0.0]]), (10, 74, 19, 3, -1, [[19, 0.9927], [61, 0.0042]]), (10, 76, 4, 1, -1, [[4, 0.9949], [9, 0.0051]]), (11, 0, 40, 4, -1, [[40, 1.0], [31, 0.0]]), (11, 34, 38, 9, -1, [[38, 0.9635], [67, 0.0117]])]
 8 0 40 4 -1 [[40, 1.0], [31, 0.0]].O.OOXX.XOO.OOOXX.OOOOOOX.XOOXXXXXX.OXXXXXX.XOXXXOXXX.OXOOOOXXXOOOOXOOOO.OOOX.XXO
 8 46 5 2 -1 [[5, 0.9883], [4, 0.0113]]............OXOX....OOOOXO..OXXXXXX...XXXXX.X..XXOXXX...OOOOXXX.....OOOO.........
 8 62 37 8 -1 [[37, 0.9565], [46, 0.0257]].O.OOXX.XO..OOOXX..OOOOOX.X.OXXXXXX...XXXXX.X..XXOXXX...OOOOXXX..O..OOOO.........
 8 64 46 8 -1 [[46, 0.9982], [36, 0.0009]].O.OOXX.XO..OOOXX..OOOOOX.X.OXXXXXX..XXXXXX.XO.XXOXXX...OOOOXXX..O..OOOO.........

Counter({'moves': 5123, 'self_atari': 251, 'noncapture_self_atari': 141})Confident, no capture or atari save: 605[(8, 0, 40, 4, -1, [[40, 1.0], [31, 0.0]]), (8, 46, 5, 2, -1, [[5, 0.9883], [4, 0.0113]]), (8, 62, 37, 8, -1, [[37, 0.9565], [46, 0.0257]]), (8, 64, 46, 8, -1, [[46, 0.9982], [36, 0.0009]]), (8, 66, 55, 8, -1, [[55, 0.9827], [36, 0.0159]]), (8, 72, 27, 1, -1, [[27, 1.0], [77, 0.0]]), (8, 74, 77, 2, -1, [[77, 0.9861], [66, 0.0105]]), (8, 76, 76, 3, -1, [[76, 0.9999], [67, 0.0001]]), (8, 78, 67, 2, -1, [[67, 0.9958], [75, 0.004]]), (8, 82, 2, 1, -1, [[2, 1.0], [78, 0.0]]), (8, 84, 78, 2, -1, [[78, 0.9661], [77, 0.0333]]), (8, 86, 79, 1, -1, [[79, 1.0], [77, 0.0]]), (8, 88, 76, 2, -1, [[76, 0.9575], [81, 0.0412]]), (8, 90, 67, 1, -1, [[67, 1.0], [81, 0.0]]), (9, 0, 40, 4, 1, [[40, 1.0], [31, 0.0]]), (9, 26, 67, 1, 1, [[67, 1.0], [38, 0.0]]), (9, 58, 35, 15, 1, [[35, 0.9978], [2, 0.0022]]), (9, 68, 8, 1, 1, [[8, 0.9996], [66, 0.0003]]), (9, 100, 25, 23, 1, [[25, 0.952], [23, 0.0125]]), (9, 106, 15, 22, 1, [[15, 0.9964], [69, 0.0027]]), (10, 0, 40, 4, -1, [[40, 1.0], [31, 0.0]]), (10, 38, 43, 2, -1, [[43, 1.0], [20, 0.0]]), (10, 56, 16, 6, -1, [[16, 0.9638], [28, 0.0151]]), (10, 62, 54, 1, -1, [[54, 1.0], [19, 0.0]]), (10, 66, 76, 1, -1, [[76, 1.0], [19, 0.0]]), (10, 70, 27, 6, -1, [[27, 1.0], [19, 0.0]]), (10, 74, 19, 3, -1, [[19, 0.9927], [61, 0.0042]]), (10, 76, 4, 1, -1, [[4, 0.9949], [9, 0.0051]]), (11, 0, 40, 4, -1, [[40, 1.0], [31, 0.0]]), (11, 34, 38, 9, -1, [[38, 0.9635], [67, 0.0117]])]
 8 0 40 4 -1 [[40, 1.0], [31, 0.0]].O.OOXX.XOO.OOOXX.OOOOOOX.XOOXXXXXX.OXXXXXX.XOXXXOXXX.OXOOOOXXXOOOOXOOOO.OOOX.XXO
 8 46 5 2 -1 [[5, 0.9883], [4, 0.0113]]............OXOX....OOOOXO..OXXXXXX...XXXXX.X..XXOXXX...OOOOXXX.....OOOO.........
 8 62 37 8 -1 [[37, 0.9565], [46, 0.0257]].O.OOXX.XO..OOOXX..OOOOOX.X.OXXXXXX...XXXXX.X..XXOXXX...OOOOXXX..O..OOOO.........
 8 64 46 8 -1 [[46, 0.9982], [36, 0.0009]].O.OOXX.XO..OOOXX..OOOOOX.X.OXXXXXX..XXXXXX.XO.XXOXXX...OOOOXXX..O..OOOO.........

python

import jsonsrc=open('/work/design_v3.py').read()src=src.replace("first_center=0.)","first_center=0., triangle=0.)")# White was unused. Reuse its global slot for the empty-board indicator,# leaving 24 final-layer-4 units for symmetric central shape features.src=src.replace("w(1,2,487,1)\n    b(1,488,1)","b(1,487,1)")src=src.replace("w(1,3*q,488,-1)","w(1,3*q,487,-1)")src=src.replace("w(1,3*q+1,488,-1)","w(1,3*q+1,487,-1)")src=src.replace("; w(k,488,488,1)","")src=src.replace("w(5,488,P,p['first_center'])","w(5,487,P,p['first_center'])")src=src.replace("    line=[-1.2",'''    central_points=[q for q,(r,c) in enumerate(xy) if 2<=r<=6 and 2<=c<=6 and q!=40]    shape_units={q:488+i for i,q in enumerate(central_points)}    for q,j in shape_units.items():        r,c=xy[q]        b(4,j,-2)        for t in adj[q]: w(4,u(0,t),j,1)        for dr,dc in [(-1,-1),(-1,1),(1,-1),(1,1)]:            w(4,u(0,(r+dr)*9+c+dc),j,1)    line=[-1.2''')src=src.replace("central=max(0.",'''if q in shape_units: w(5,shape_units[q],P,-p['triangle'])        central=max(0.''')open('/work/design_v4.py','w').write(src)exec(src)config=install(dict(own_adj=2.,opp_adj=1.5,crowd=3.,temperature_scale=2.5,                    opening=3.,defend=1.,first_center=20.,triangle=1.5))json.dump(config,open('/work/config_v4.json','w'),indent=2)print(config)

import jsonsrc=open('/work/design_v3.py').read()src=src.replace("first_center=0.)","first_center=0., triangle=0.)")# White was unused. Reuse its global slot for the empty-board indicator,# leaving 24 final-layer-4 units for symmetric central shape features.src=src.replace("w(1,2,487,1)\n    b(1,488,1)","b(1,487,1)")src=src.replace("w(1,3*q,488,-1)","w(1,3*q,487,-1)")src=src.replace("w(1,3*q+1,488,-1)","w(1,3*q+1,487,-1)")src=src.replace("; w(k,488,488,1)","")src=src.replace("w(5,488,P,p['first_center'])","w(5,487,P,p['first_center'])")src=src.replace("    line=[-1.2",'''    central_points=[q for q,(r,c) in enumerate(xy) if 2<=r<=6 and 2<=c<=6 and q!=40]    shape_units={q:488+i for i,q in enumerate(central_points)}    for q,j in shape_units.items():        r,c=xy[q]        b(4,j,-2)        for t in adj[q]: w(4,u(0,t),j,1)        for dr,dc in [(-1,-1),(-1,1),(1,-1),(1,1)]:            w(4,u(0,(r+dr)*9+c+dc),j,1)    line=[-1.2''')src=src.replace("central=max(0.",'''if q in shape_units: w(5,shape_units[q],P,-p['triangle'])        central=max(0.''')open('/work/design_v4.py','w').write(src)exec(src)config=install(dict(own_adj=2.,opp_adj=1.5,crowd=3.,temperature_scale=2.5,                    opening=3.,defend=1.,first_center=20.,triangle=1.5))json.dump(config,open('/work/config_v4.json','w'),indent=2)print(config)

OUTPUT of Python run 9 of at most 120:

{'capture': 11.0, 'capture_extra': 3.0, 'save': 7.0, 'danger': 10.0, 'eye': 22.0, 'isolation': 3.0, 'isolation_threshold': 2.5, 'base': 1.3, 'pass_bias': -1.0, 'attack': 1.3, 'own_adj': 2.0, 'opp_adj': 1.5, 'opening': 3.0, 'save_danger': 2.0, 'safe2': 1.0, 'safe3': 1.0, 'cap_safety': 1.0, 'crowd': 3.0, 'temperature_scale': 2.5, 'defend': 1.0, 'first_center': 20.0, 'triangle': 1.5}

{'capture': 11.0, 'capture_extra': 3.0, 'save': 7.0, 'danger': 10.0, 'eye': 22.0, 'isolation': 3.0, 'isolation_threshold': 2.5, 'base': 1.3, 'pass_bias': -1.0, 'attack': 1.3, 'own_adj': 2.0, 'opp_adj': 1.5, 'opening': 3.0, 'save_danger': 2.0, 'safe2': 1.0, 'safe3': 1.0, 'cap_safety': 1.0, 'crowd': 3.0, 'temperature_scale': 2.5, 'defend': 1.0, 'first_center': 20.0, 'triangle': 1.5}

Games pass 4 of 10 done.

{"ok": true, "pass": 4, "kind": "games", "per_opponent": {"random": {"games": 8, "score": 1.0, "wins": 8, "losses": 0, "score_as_black": 1.0, "score_as_white": 1.0, "mean_margin": 81.0, "mean_length": 121.0, "stones_captured_per_game": 94.75, "stones_lost_per_game": 0.125, "passes_per_game": 1.0, "games_with_a_pass_before_move_60": 0, "end_two_passes": 8, "end_repetition": 0, "end_move_limit": 0}, "strong": {"games": 96, "score": 0.7916666865348816, "wins": 76, "losses": 20, "score_as_black": 0.7291666865348816, "score_as_white": 0.8541666865348816, "mean_margin": 60.09375, "mean_length": 118.08333333333333, "stones_captured_per_game": 67.73958333333333, "stones_lost_per_game": 8.5625, "passes_per_game": 1.5208333333333333, "games_with_a_pass_before_move_60": 0, "end_two_passes": 89, "end_repetition": 7, "end_move_limit": 0}, "weak": {"games": 24, "score": 0.9583333134651184, "wins": 23, "losses": 1, "score_as_black": 0.9166666865348816, "score_as_white": 1.0, "mean_margin": 74.625, "mean_length": 124.95833333333333, "stones_captured_per_game": 84.25, "stones_lost_per_game": 3.1666666666666665, "passes_per_game": 2.4166666666666665, "games_with_a_pass_before_move_60": 0, "end_two_passes": 23, "end_repetition": 1, "end_move_limit": 0}}, "files": "/work/pass_4/", "readme": "games.json: one entry per game (opponent, net_color, result +1/-1, margin = net's area score minus the opponent's incl. komi, end, stones captured / lost, moves: ply, by, move (row*9+col, 81 = pass), and for the network's moves its top-5 [move, probability]). boards.npy: (games, 162, 81) int8, the board AFTER each ply (1 Black, -1 White, 0 empty; index row*9+col); valid.npy: (games, 162) which plies were played."}

{"ok": true, "pass": 4, "kind": "games", "per_opponent": {"random": {"games": 8, "score": 1.0, "wins": 8, "losses": 0, "score_as_black": 1.0, "score_as_white": 1.0, "mean_margin": 81.0, "mean_length": 121.0, "stones_captured_per_game": 94.75, "stones_lost_per_game": 0.125, "passes_per_game": 1.0, "games_with_a_pass_before_move_60": 0, "end_two_passes": 8, "end_repetition": 0, "end_move_limit": 0}, "strong": {"games": 96, "score": 0.7916666865348816, "wins": 76, "losses": 20, "score_as_black": 0.7291666865348816, "score_as_white": 0.8541666865348816, "mean_margin": 60.09375, "mean_length": 118.08333333333333, "stones_captured_per_game": 67.73958333333333, "stones_lost_per_game": 8.5625, "passes_per_game": 1.5208333333333333, "games_with_a_pass_before_move_60": 0, "end_two_passes": 89, "end_repetition": 7, "end_move_limit": 0}, "weak": {"games": 24, "score": 0.9583333134651184, "wins": 23, "losses": 1, "score_as_black": 0.9166666865348816, "score_as_white": 1.0, "mean_margin": 74.625, "mean_length": 124.95833333333333, "stones_captured_per_game": 84.25, "stones_lost_per_game": 3.1666666666666665, "passes_per_game": 2.4166666666666665, "games_with_a_pass_before_move_60": 0, "end_two_passes": 23, "end_repetition": 1, "end_move_limit": 0}}, "files": "/work/pass_4/", "readme": "games.json: one entry per game (opponent, net_color, result +1/-1, margin = net's area score minus the opponent's incl. komi, end, stones captured / lost, moves: ply, by, move (row*9+col, 81 = pass), and for the network's moves its top-5 [move, probability]). boards.npy: (games, 162, 81) int8, the board AFTER each ply (1 Black, -1 White, 0 empty; index row*9+col); valid.npy: (games, 162) which plies were played."}

Passes left: 6.

python

import json,numpy as np,collectionsg=json.load(open('/work/pass_4/games.json')); boards=np.load('/work/pass_4/boards.npy')def show(b): return '\n'.join(''.join({0:'.',1:'X',-1:'O'}[int(v)] for v in row) for row in np.asarray(b).reshape(9,9))for i,x in enumerate(g):    if x['result']<0:        print(i,x['opponent'],x['net_color'],x['margin'],x['end'],              'caps/lost',x['stones_captured_by_net'],x['stones_lost_by_net'])print('\nREPETITION DETAILS')for gi,x in enumerate(g):    if x['end']!='repetition':continue    n=len(x['moves'])    print('\nGAME',gi,x['result'],x['net_color'],x['margin'])    for pi in range(max(0,n-5),n):        print(x['moves'][pi])        print(show(boards[gi,pi]))    if gi>30:break

import json,numpy as np,collectionsg=json.load(open('/work/pass_4/games.json')); boards=np.load('/work/pass_4/boards.npy')def show(b): return '\n'.join(''.join({0:'.',1:'X',-1:'O'}[int(v)] for v in row) for row in np.asarray(b).reshape(9,9))for i,x in enumerate(g):    if x['result']<0:        print(i,x['opponent'],x['net_color'],x['margin'],x['end'],              'caps/lost',x['stones_captured_by_net'],x['stones_lost_by_net'])print('\nREPETITION DETAILS')for gi,x in enumerate(g):    if x['end']!='repetition':continue    n=len(x['moves'])    print('\nGAME',gi,x['result'],x['net_color'],x['margin'])    for pi in range(max(0,n-5),n):        print(x['moves'][pi])        print(show(boards[gi,pi]))    if gi>30:break

OUTPUT of Python run 10 of at most 120:

12 strong black -16.5 two passes caps/lost 5 1715 strong black 6.5 repetition caps/lost 13 817 strong black -25.5 two passes caps/lost 21 5321 strong black -3.5 two passes caps/lost 8 2122 strong black -4.5 two passes caps/lost 15 1735 strong black -6.5 two passes caps/lost 24 3637 strong black -14.5 two passes caps/lost 6 1539 strong black -16.5 two passes caps/lost 32 7241 strong black 1.5 repetition caps/lost 7 443 strong black -16.5 two passes caps/lost 17 3044 strong black -0.5 two passes caps/lost 20 4645 strong black 2.5 repetition caps/lost 18 1350 strong black -13.5 two passes caps/lost 3 957 strong white 49.5 repetition caps/lost 50 468 strong white -7.5 two passes caps/lost 8 3979 strong white -11.5 two passes caps/lost 9 2582 strong white 7.5 repetition caps/lost 3 488 strong white 37.5 repetition caps/lost 42 1090 strong white -13.5 repetition caps/lost 14 3493 strong white -2.5 two passes caps/lost 16 49106 weak black 16.5 repetition caps/lost 39 14
REPETITION DETAILS
GAME 15 -1 black 6.5{'ply': 84, 'by': 'net', 'move': 18, 'top5': [[18, 1.0], [72, 0.0], [8, 0.0], [81, 0.0], [80, 0.0]]}.X.X.XXO.XXX.XXXXXXXXXXXOOXOXX.XOOOOOOXXXXOOOOOOOOOO..OOOXOXXOOOOXXX.XOO.X.X.X.X.{'ply': 85, 'by': 'opponent', 'move': 52}.X.X.XXO.XXX.XXXXXXXXXXXOOXOXX.XOOOOOOXXXXOOOOOOOOOOO.OOOXOXXOOOOXXX.XOO.X.X.X.X.{'ply': 86, 'by': 'net', 'move': 72, 'top5': [[72, 1.0], [8, 0.0], [81, 0.0], [80, 0.0], [68, 0.0]]}.X.X.XXO.XXX.XXXXXXXXXXXOOXOXX.XOOOOOOXXXXOOOOOOOOOOO.OOOXOXXOOOOXXX.XOOXX.X.X.X.{'ply': 87, 'by': 'opponent', 'move': 74}.X.X.XXO.XXX.XXXXXXXXXXXOOXOXX.XOOOOOOXXXXOOOOOOOOOOO.OOOXOXXOOOOXXX.XOO..OX.X.X.{'ply': 88, 'by': 'net', 'move': 73, 'top5': [[73, 1.0], [8, 0.0], [81, 0.0], [80, 0.0], [72, 0.0]]}.X.X.XXO.XXX.XXXXXXXXXXXOOXOXX.XOOOOOOXXXXOOOOOOOOOOO.OOOXOXXOOOOXXX.XOO.X.X.X.X.
GAME 41 -1 black 1.5{'ply': 48, 'by': 'net', 'move': 13, 'top5': [[13, 0.9998], [44, 0.0001], [62, 0.0001], [68, 0.0], [67, 0.0]]}..........XOOXXOX..XOOOO.O.X.XXOOXO..X.XXX.X..XXXOOXXOXOOOOXXX...................{'ply': 49, 'by': 'opponent', 'move': 4}....O.....XOOXXOX..XOOOO.O.X.XXOOXO..X.XXX.X..XXXOOXXOXOOOOXXX...................{'ply': 50, 'by': 'net', 'move': 24, 'top5': [[24, 1.0], [44, 0.0], [62, 0.0], [68, 0.0], [67, 0.0]]}....O.....XOOXXOX..XOOOOXO.X.XXOOXO..X.XXX.X..XXXOOXXOXOOOOXXX...................{'ply': 51, 'by': 'opponent', 'move': 42}....O.....XOOXXOX..XOOOO.O.X.XXOO.O..X.XXXOX..XXXOOXXOXOOOOXXX...................{'ply': 52, 'by': 'net', 'move': 33, 'top5': [[33, 1.0], [44, 0.0], [62, 0.0], [68, 0.0], [69, 0.0]]}....O.....XOOXXOX..XOOOO.O.X.XXOOXO..X.XXX.X..XXXOOXXOXOOOOXXX...................

12 strong black -16.5 two passes caps/lost 5 1715 strong black 6.5 repetition caps/lost 13 817 strong black -25.5 two passes caps/lost 21 5321 strong black -3.5 two passes caps/lost 8 2122 strong black -4.5 two passes caps/lost 15 1735 strong black -6.5 two passes caps/lost 24 3637 strong black -14.5 two passes caps/lost 6 1539 strong black -16.5 two passes caps/lost 32 7241 strong black 1.5 repetition caps/lost 7 443 strong black -16.5 two passes caps/lost 17 3044 strong black -0.5 two passes caps/lost 20 4645 strong black 2.5 repetition caps/lost 18 1350 strong black -13.5 two passes caps/lost 3 957 strong white 49.5 repetition caps/lost 50 468 strong white -7.5 two passes caps/lost 8 3979 strong white -11.5 two passes caps/lost 9 2582 strong white 7.5 repetition caps/lost 3 488 strong white 37.5 repetition caps/lost 42 1090 strong white -13.5 repetition caps/lost 14 3493 strong white -2.5 two passes caps/lost 16 49106 weak black 16.5 repetition caps/lost 39 14
REPETITION DETAILS
GAME 15 -1 black 6.5{'ply': 84, 'by': 'net', 'move': 18, 'top5': [[18, 1.0], [72, 0.0], [8, 0.0], [81, 0.0], [80, 0.0]]}.X.X.XXO.XXX.XXXXXXXXXXXOOXOXX.XOOOOOOXXXXOOOOOOOOOO..OOOXOXXOOOOXXX.XOO.X.X.X.X.{'ply': 85, 'by': 'opponent', 'move': 52}.X.X.XXO.XXX.XXXXXXXXXXXOOXOXX.XOOOOOOXXXXOOOOOOOOOOO.OOOXOXXOOOOXXX.XOO.X.X.X.X.{'ply': 86, 'by': 'net', 'move': 72, 'top5': [[72, 1.0], [8, 0.0], [81, 0.0], [80, 0.0], [68, 0.0]]}.X.X.XXO.XXX.XXXXXXXXXXXOOXOXX.XOOOOOOXXXXOOOOOOOOOOO.OOOXOXXOOOOXXX.XOOXX.X.X.X.{'ply': 87, 'by': 'opponent', 'move': 74}.X.X.XXO.XXX.XXXXXXXXXXXOOXOXX.XOOOOOOXXXXOOOOOOOOOOO.OOOXOXXOOOOXXX.XOO..OX.X.X.{'ply': 88, 'by': 'net', 'move': 73, 'top5': [[73, 1.0], [8, 0.0], [81, 0.0], [80, 0.0], [72, 0.0]]}.X.X.XXO.XXX.XXXXXXXXXXXOOXOXX.XOOOOOOXXXXOOOOOOOOOOO.OOOXOXXOOOOXXX.XOO.X.X.X.X.
GAME 41 -1 black 1.5{'ply': 48, 'by': 'net', 'move': 13, 'top5': [[13, 0.9998], [44, 0.0001], [62, 0.0001], [68, 0.0], [67, 0.0]]}..........XOOXXOX..XOOOO.O.X.XXOOXO..X.XXX.X..XXXOOXXOXOOOOXXX...................{'ply': 49, 'by': 'opponent', 'move': 4}....O.....XOOXXOX..XOOOO.O.X.XXOOXO..X.XXX.X..XXXOOXXOXOOOOXXX...................{'ply': 50, 'by': 'net', 'move': 24, 'top5': [[24, 1.0], [44, 0.0], [62, 0.0], [68, 0.0], [67, 0.0]]}....O.....XOOXXOX..XOOOOXO.X.XXOOXO..X.XXX.X..XXXOOXXOXOOOOXXX...................{'ply': 51, 'by': 'opponent', 'move': 42}....O.....XOOXXOX..XOOOO.O.X.XXOO.O..X.XXXOX..XXXOOXXOXOOOOXXX...................{'ply': 52, 'by': 'net', 'move': 33, 'top5': [[33, 1.0], [44, 0.0], [62, 0.0], [68, 0.0], [69, 0.0]]}....O.....XOOXXOX..XOOOO.O.X.XXOOXO..X.XXX.X..XXXOOXXOXOOOOXXX...................

python

import jsonsrc=open('/work/design_v4.py').read()src=src.replace("triangle=0.)","triangle=0., capture_guard=0.)")src=src.replace("w(5,u(3,t),D,-p['cap_safety'])","# Captures are accounted for after their reliability gate, in layer 6.")src=src.replace("if wm: w(5,u(0,t),P,wm)","""guard_weight={1:1.,2:1.,3:.5}.get(d,0.)*p['capture_guard']            if guard_weight:                w(5,u(1,t),C,-guard_weight)                w(5,u(3,t),C,guard_weight)            if wm: w(5,u(0,t),P,wm)""")src=src.replace("w(6,D,u(3,q),1)",                "w(6,D,u(3,q),1); w(6,C,u(3,q),-p['cap_safety'])")open('/work/design_v5.py','w').write(src)exec(src)cfg=json.load(open('/work/config_v4.json'))cfg['capture_guard']=.2config=install(cfg)json.dump(config,open('/work/config_v5.json','w'),indent=2)print(config)

import jsonsrc=open('/work/design_v4.py').read()src=src.replace("triangle=0.)","triangle=0., capture_guard=0.)")src=src.replace("w(5,u(3,t),D,-p['cap_safety'])","# Captures are accounted for after their reliability gate, in layer 6.")src=src.replace("if wm: w(5,u(0,t),P,wm)","""guard_weight={1:1.,2:1.,3:.5}.get(d,0.)*p['capture_guard']            if guard_weight:                w(5,u(1,t),C,-guard_weight)                w(5,u(3,t),C,guard_weight)            if wm: w(5,u(0,t),P,wm)""")src=src.replace("w(6,D,u(3,q),1)",                "w(6,D,u(3,q),1); w(6,C,u(3,q),-p['cap_safety'])")open('/work/design_v5.py','w').write(src)exec(src)cfg=json.load(open('/work/config_v4.json'))cfg['capture_guard']=.2config=install(cfg)json.dump(config,open('/work/config_v5.json','w'),indent=2)print(config)

OUTPUT of Python run 11 of at most 120:

{'capture': 11.0, 'capture_extra': 3.0, 'save': 7.0, 'danger': 10.0, 'eye': 22.0, 'isolation': 3.0, 'isolation_threshold': 2.5, 'base': 1.3, 'pass_bias': -1.0, 'attack': 1.3, 'own_adj': 2.0, 'opp_adj': 1.5, 'opening': 3.0, 'save_danger': 2.0, 'safe2': 1.0, 'safe3': 1.0, 'cap_safety': 1.0, 'crowd': 3.0, 'temperature_scale': 2.5, 'defend': 1.0, 'first_center': 20.0, 'triangle': 1.5, 'capture_guard': 0.2}

{'capture': 11.0, 'capture_extra': 3.0, 'save': 7.0, 'danger': 10.0, 'eye': 22.0, 'isolation': 3.0, 'isolation_threshold': 2.5, 'base': 1.3, 'pass_bias': -1.0, 'attack': 1.3, 'own_adj': 2.0, 'opp_adj': 1.5, 'opening': 3.0, 'save_danger': 2.0, 'safe2': 1.0, 'safe3': 1.0, 'cap_safety': 1.0, 'crowd': 3.0, 'temperature_scale': 2.5, 'defend': 1.0, 'first_center': 20.0, 'triangle': 1.5, 'capture_guard': 0.2}

Games pass 5 of 10 done.

{"ok": true, "pass": 5, "kind": "games", "per_opponent": {"random": {"games": 8, "score": 1.0, "wins": 8, "losses": 0, "score_as_black": 1.0, "score_as_white": 1.0, "mean_margin": 66.25, "mean_length": 115.0, "stones_captured_per_game": 82.25, "stones_lost_per_game": 0.0, "passes_per_game": 1.5, "games_with_a_pass_before_move_60": 0, "end_two_passes": 8, "end_repetition": 0, "end_move_limit": 0}, "strong": {"games": 96, "score": 0.8333333134651184, "wins": 80, "losses": 16, "score_as_black": 0.7916666865348816, "score_as_white": 0.875, "mean_margin": 60.47916793823242, "mean_length": 120.0625, "stones_captured_per_game": 75.5, "stones_lost_per_game": 7.052083333333333, "passes_per_game": 2.1041666666666665, "games_with_a_pass_before_move_60": 0, "end_two_passes": 95, "end_repetition": 1, "end_move_limit": 0}, "weak": {"games": 24, "score": 1.0, "wins": 24, "losses": 0, "score_as_black": 1.0, "score_as_white": 1.0, "mean_margin": 75.16666412353516, "mean_length": 123.66666666666667, "stones_captured_per_game": 81.16666666666667, "stones_lost_per_game": 2.7083333333333335, "passes_per_game": 2.0833333333333335, "games_with_a_pass_before_move_60": 0, "end_two_passes": 24, "end_repetition": 0, "end_move_limit": 0}}, "files": "/work/pass_5/", "readme": "games.json: one entry per game (opponent, net_color, result +1/-1, margin = net's area score minus the opponent's incl. komi, end, stones captured / lost, moves: ply, by, move (row*9+col, 81 = pass), and for the network's moves its top-5 [move, probability]). boards.npy: (games, 162, 81) int8, the board AFTER each ply (1 Black, -1 White, 0 empty; index row*9+col); valid.npy: (games, 162) which plies were played."}

{"ok": true, "pass": 5, "kind": "games", "per_opponent": {"random": {"games": 8, "score": 1.0, "wins": 8, "losses": 0, "score_as_black": 1.0, "score_as_white": 1.0, "mean_margin": 66.25, "mean_length": 115.0, "stones_captured_per_game": 82.25, "stones_lost_per_game": 0.0, "passes_per_game": 1.5, "games_with_a_pass_before_move_60": 0, "end_two_passes": 8, "end_repetition": 0, "end_move_limit": 0}, "strong": {"games": 96, "score": 0.8333333134651184, "wins": 80, "losses": 16, "score_as_black": 0.7916666865348816, "score_as_white": 0.875, "mean_margin": 60.47916793823242, "mean_length": 120.0625, "stones_captured_per_game": 75.5, "stones_lost_per_game": 7.052083333333333, "passes_per_game": 2.1041666666666665, "games_with_a_pass_before_move_60": 0, "end_two_passes": 95, "end_repetition": 1, "end_move_limit": 0}, "weak": {"games": 24, "score": 1.0, "wins": 24, "losses": 0, "score_as_black": 1.0, "score_as_white": 1.0, "mean_margin": 75.16666412353516, "mean_length": 123.66666666666667, "stones_captured_per_game": 81.16666666666667, "stones_lost_per_game": 2.7083333333333335, "passes_per_game": 2.0833333333333335, "games_with_a_pass_before_move_60": 0, "end_two_passes": 24, "end_repetition": 0, "end_move_limit": 0}}, "files": "/work/pass_5/", "readme": "games.json: one entry per game (opponent, net_color, result +1/-1, margin = net's area score minus the opponent's incl. komi, end, stones captured / lost, moves: ply, by, move (row*9+col, 81 = pass), and for the network's moves its top-5 [move, probability]). boards.npy: (games, 162, 81) int8, the board AFTER each ply (1 Black, -1 White, 0 empty; index row*9+col); valid.npy: (games, 162) which plies were played."}

Passes left: 5.

python

import json,numpy as np,collectionsg=json.load(open('/work/pass_5/games.json')); boards=np.load('/work/pass_5/boards.npy')adj=[[j for j in (q-9,q+9,q-1,q+1) if 0<=j<81 and abs(j//9-q//9)+abs(j%9-q%9)==1] for q in range(81)]def group_info(b):    info={}    for q in range(81):        if b[q]==0 or q in info:continue        stack=[q];ss={q};ll=set()        while stack:            t=stack.pop()            for v in adj[t]:                if b[v]==0:ll.add(v)                elif b[v]==b[q] and v not in ss:ss.add(v);stack.append(v)        for t in ss:info[t]=(ss,ll)    return infodef show(b): return '\n'.join(''.join({0:'.',1:'X',-1:'O'}[int(v)] for v in row) for row in np.asarray(b).reshape(9,9))stats=collections.Counter();bad=[]for gi,x in enumerate(g):    if x['opponent']!='strong':continue    if x['result']<0:        print('LOSS',gi,x['net_color'],x['margin'],x['end'],x['stones_lost_by_net'])    color=1 if x['net_color']=='black' else -1    for pi,m in enumerate(x['moves']):        if m['by']!='net' or m['move']==81:continue        before=np.zeros(81,np.int8) if pi==0 else boards[gi,pi-1]        after=boards[gi,pi];gr=group_info(before);q=m['move']        caps=sum(before==-color)-sum(after==-color)        grp,libs=group_info(after)[q]        stats['moves']+=1        if len(libs)==1:            stats['self_atari']+=1            if caps==0:                stats['noncapture_self_atari']+=1                if len(grp)>=3:bad.append((gi,pi,q,len(grp),x['result'],m['top5']))print(stats)print('Large group self-ataris:',bad[:12])for gi,pi,q,n,res,top in bad[:3]:    print('\n',gi,pi,q,n,res,top)    print(show(boards[gi,pi-1]))loss=[i for i,x in enumerate(g) if x['opponent']=='strong' and x['result']<0]for gi in loss[:2]:    print('\nGAME',gi)    x=g[gi]    for pi,m in enumerate(x['moves']):        if pi<34 and m['by']=='net':print(pi,m['move'],m['top5'][:3])        if pi in [15,31,47,len(x['moves'])-1]:print('after',pi,'\n'+show(boards[gi,pi]))

import json,numpy as np,collectionsg=json.load(open('/work/pass_5/games.json')); boards=np.load('/work/pass_5/boards.npy')adj=[[j for j in (q-9,q+9,q-1,q+1) if 0<=j<81 and abs(j//9-q//9)+abs(j%9-q%9)==1] for q in range(81)]def group_info(b):    info={}    for q in range(81):        if b[q]==0 or q in info:continue        stack=[q];ss={q};ll=set()        while stack:            t=stack.pop()            for v in adj[t]:                if b[v]==0:ll.add(v)                elif b[v]==b[q] and v not in ss:ss.add(v);stack.append(v)        for t in ss:info[t]=(ss,ll)    return infodef show(b): return '\n'.join(''.join({0:'.',1:'X',-1:'O'}[int(v)] for v in row) for row in np.asarray(b).reshape(9,9))stats=collections.Counter();bad=[]for gi,x in enumerate(g):    if x['opponent']!='strong':continue    if x['result']<0:        print('LOSS',gi,x['net_color'],x['margin'],x['end'],x['stones_lost_by_net'])    color=1 if x['net_color']=='black' else -1    for pi,m in enumerate(x['moves']):        if m['by']!='net' or m['move']==81:continue        before=np.zeros(81,np.int8) if pi==0 else boards[gi,pi-1]        after=boards[gi,pi];gr=group_info(before);q=m['move']        caps=sum(before==-color)-sum(after==-color)        grp,libs=group_info(after)[q]        stats['moves']+=1        if len(libs)==1:            stats['self_atari']+=1            if caps==0:                stats['noncapture_self_atari']+=1                if len(grp)>=3:bad.append((gi,pi,q,len(grp),x['result'],m['top5']))print(stats)print('Large group self-ataris:',bad[:12])for gi,pi,q,n,res,top in bad[:3]:    print('\n',gi,pi,q,n,res,top)    print(show(boards[gi,pi-1]))loss=[i for i,x in enumerate(g) if x['opponent']=='strong' and x['result']<0]for gi in loss[:2]:    print('\nGAME',gi)    x=g[gi]    for pi,m in enumerate(x['moves']):        if pi<34 and m['by']=='net':print(pi,m['move'],m['top5'][:3])        if pi in [15,31,47,len(x['moves'])-1]:print('after',pi,'\n'+show(boards[gi,pi]))

OUTPUT of Python run 12 of at most 120:

LOSS 19 black -33.5 two passes 63LOSS 24 black -18.5 two passes 30LOSS 29 black -10.5 two passes 15LOSS 32 black -14.5 two passes 9LOSS 35 black -18.5 two passes 23LOSS 38 black -17.5 two passes 11LOSS 43 black -4.5 two passes 18LOSS 44 black -26.5 two passes 33LOSS 47 black -21.5 two passes 18LOSS 55 black -6.5 two passes 12LOSS 60 white -3.5 two passes 13LOSS 67 white 19.5 repetition 12LOSS 72 white -7.5 two passes 22LOSS 74 white -3.5 two passes 70LOSS 91 white -5.5 two passes 19LOSS 96 white -0.5 two passes 19Counter({'moves': 5576, 'self_atari': 92, 'noncapture_self_atari': 62})Large group self-ataris: [(15, 62, 4, 10, 1, [[4, 1.0], [5, 0.0], [16, 0.0], [35, 0.0], [34, 0.0]]), (16, 66, 5, 3, 1, [[5, 0.8449], [11, 0.0597], [66, 0.0304], [3, 0.0225], [21, 0.0189]]), (17, 66, 4, 8, 1, [[4, 0.997], [70, 0.0017], [5, 0.001], [6, 0.0001], [76, 0.0]]), (24, 68, 6, 14, -1, [[6, 0.8687], [7, 0.1267], [9, 0.0034], [44, 0.0009], [35, 0.0001]]), (24, 76, 38, 4, -1, [[38, 0.9551], [32, 0.0449], [23, 0.0], [9, 0.0], [5, 0.0]]), (24, 80, 41, 6, -1, [[41, 0.964], [9, 0.0353], [72, 0.0006], [0, 0.0], [81, 0.0]]), (24, 88, 1, 3, -1, [[1, 0.7305], [81, 0.2687], [18, 0.0007], [73, 0.0], [68, 0.0]]), (29, 92, 60, 3, -1, [[60, 1.0], [67, 0.0], [81, 0.0], [66, 0.0], [32, 0.0]]), (43, 96, 11, 3, -1, [[11, 0.9995], [20, 0.0005], [29, 0.0], [81, 0.0], [1, 0.0]]), (44, 86, 18, 3, -1, [[1, 0.5685], [81, 0.2686], [18, 0.1629], [77, 0.0], [12, 0.0]]), (47, 74, 70, 3, -1, [[70, 1.0], [52, 0.0], [21, 0.0], [20, 0.0], [81, 0.0]]), (55, 62, 77, 4, -1, [[77, 1.0], [10, 0.0], [22, 0.0], [78, 0.0], [71, 0.0]])]
 15 62 4 10 1 [[4, 1.0], [5, 0.0], [16, 0.0], [35, 0.0], [34, 0.0]]..OX......O.OXXO...OOXXOO..OXOOXO...XXOXXXOOO.XOOOOXOXX.XXOXXXX.XX.XOOO.X.XXXX.O.
 16 66 5 3 1 [[5, 0.8449], [11, 0.0597], [66, 0.0304], [3, 0.0225], [21, 0.0189]]....O.......XOXXOOOOX.XOOOOXX...XXOX.XXXXX.XXX.X.X.X.X.XXX.X.X.X...XOX.X.....O.X.
 17 66 4 8 1 [[4, 0.997], [70, 0.0017], [5, 0.001], [6, 0.0001], [76, 0.0]]..OO.....OOXXXXOO.XXXOOOOO.OOOOOXXXOOOXXXX.XXOXXX..XX.XX..XX.X..XXXXXX..X........
GAME 190 40 [[40, 1.0], [31, 0.0], [39, 0.0]]2 31 [[31, 0.4207], [39, 0.2552], [41, 0.2552]]4 41 [[39, 0.2916], [41, 0.2916], [30, 0.1191]]6 59 [[59, 0.4843], [32, 0.4274], [42, 0.0486]]8 39 [[39, 0.4056], [57, 0.3433], [32, 0.2006]]10 32 [[57, 0.9449], [51, 0.0222], [32, 0.0149]]12 33 [[60, 0.3643], [33, 0.3422], [57, 0.2352]]14 60 [[60, 0.7843], [30, 0.1397], [38, 0.0387]]after 15 ...............................XXX.....XXXO....OOOOOO.....OXX....................16 38 [[30, 0.6543], [38, 0.189], [57, 0.1338]]18 61 [[57, 0.5929], [61, 0.2237], [34, 0.1786]]20 57 [[57, 0.99], [30, 0.005], [29, 0.0015]]22 66 [[66, 0.9972], [67, 0.0028], [30, 0.0]]24 67 [[67, 0.9997], [65, 0.0002], [69, 0.0001]]26 69 [[69, 0.9768], [30, 0.0085], [29, 0.0053]]28 77 [[77, 1.0], [70, 0.0], [25, 0.0]]30 30 [[30, 0.8754], [25, 0.0419], [70, 0.0418]]after 31 ........................O....OXXXXO..OXXXXOO...OOOOOO...OXOXXX...OXX.X.......X...32 70 [[46, 0.288], [70, 0.2329], [25, 0.2215]]after 47 ..OX.......OXO.....OXXO.O...OOXXXXO..OXXXXOO..OOOOOOO..XOXOXXX.OXOXX.XX...X..X...after 109 .OOOOOOOOOO.OOOOOOOOOOOOOOO.OO.OX.OOXOO.O.OOOXOOOOOOO.XXOXOXXXO.XOXX.XXXX.XX.X.X.
GAME 240 40 [[40, 1.0], [31, 0.0], [39, 0.0]]2 31 [[31, 0.4207], [39, 0.2552], [41, 0.2552]]4 41 [[39, 0.2916], [41, 0.2916], [30, 0.1191]]6 59 [[59, 0.4843], [32, 0.4274], [42, 0.0486]]8 39 [[39, 0.4056], [57, 0.3433], [32, 0.2006]]10 57 [[57, 0.9449], [51, 0.0222], [32, 0.0149]]12 38 [[38, 0.4378], [56, 0.178], [51, 0.1513]]14 66 [[66, 0.4999], [51, 0.4676], [67, 0.0325]]after 15 ...............................X......XXXXO....OOOOO....OXOX......X..............16 60 [[60, 0.9734], [67, 0.0103], [32, 0.0103]]18 67 [[67, 0.9997], [32, 0.0002], [65, 0.0001]]20 65 [[69, 0.5262], [65, 0.4098], [32, 0.0288]]22 69 [[69, 0.8477], [32, 0.0451], [37, 0.0376]]24 77 [[77, 1.0], [37, 0.0], [32, 0.0]]26 32 [[37, 0.3987], [32, 0.2645], [30, 0.1959]]28 30 [[30, 0.8601], [37, 0.089], [43, 0.018]]30 55 [[55, 0.7926], [21, 0.1196], [28, 0.0497]]after 31 ....................O.O......OXXXO...OXXXXO...OOOOOOO..XOXOXX....XXX.X.......X...32 23 [[23, 1.0], [64, 0.0], [28, 0.0]]after 47 ....X.......XOXX...OOOOXOO.XXOXXXOO..OXXXXO...OOOOOOO.OXOXOXX...XXXX.X...O...X...after 90 XXOOO.OO.XO.OOO..O.OOOO.OO.XXOXXXOOOOOXXOXOO.OOOOOOOOOOXOXOXXXOXXXXX.XXXX.XX.X.X.

LOSS 19 black -33.5 two passes 63LOSS 24 black -18.5 two passes 30LOSS 29 black -10.5 two passes 15LOSS 32 black -14.5 two passes 9LOSS 35 black -18.5 two passes 23LOSS 38 black -17.5 two passes 11LOSS 43 black -4.5 two passes 18LOSS 44 black -26.5 two passes 33LOSS 47 black -21.5 two passes 18LOSS 55 black -6.5 two passes 12LOSS 60 white -3.5 two passes 13LOSS 67 white 19.5 repetition 12LOSS 72 white -7.5 two passes 22LOSS 74 white -3.5 two passes 70LOSS 91 white -5.5 two passes 19LOSS 96 white -0.5 two passes 19Counter({'moves': 5576, 'self_atari': 92, 'noncapture_self_atari': 62})Large group self-ataris: [(15, 62, 4, 10, 1, [[4, 1.0], [5, 0.0], [16, 0.0], [35, 0.0], [34, 0.0]]), (16, 66, 5, 3, 1, [[5, 0.8449], [11, 0.0597], [66, 0.0304], [3, 0.0225], [21, 0.0189]]), (17, 66, 4, 8, 1, [[4, 0.997], [70, 0.0017], [5, 0.001], [6, 0.0001], [76, 0.0]]), (24, 68, 6, 14, -1, [[6, 0.8687], [7, 0.1267], [9, 0.0034], [44, 0.0009], [35, 0.0001]]), (24, 76, 38, 4, -1, [[38, 0.9551], [32, 0.0449], [23, 0.0], [9, 0.0], [5, 0.0]]), (24, 80, 41, 6, -1, [[41, 0.964], [9, 0.0353], [72, 0.0006], [0, 0.0], [81, 0.0]]), (24, 88, 1, 3, -1, [[1, 0.7305], [81, 0.2687], [18, 0.0007], [73, 0.0], [68, 0.0]]), (29, 92, 60, 3, -1, [[60, 1.0], [67, 0.0], [81, 0.0], [66, 0.0], [32, 0.0]]), (43, 96, 11, 3, -1, [[11, 0.9995], [20, 0.0005], [29, 0.0], [81, 0.0], [1, 0.0]]), (44, 86, 18, 3, -1, [[1, 0.5685], [81, 0.2686], [18, 0.1629], [77, 0.0], [12, 0.0]]), (47, 74, 70, 3, -1, [[70, 1.0], [52, 0.0], [21, 0.0], [20, 0.0], [81, 0.0]]), (55, 62, 77, 4, -1, [[77, 1.0], [10, 0.0], [22, 0.0], [78, 0.0], [71, 0.0]])]
 15 62 4 10 1 [[4, 1.0], [5, 0.0], [16, 0.0], [35, 0.0], [34, 0.0]]..OX......O.OXXO...OOXXOO..OXOOXO...XXOXXXOOO.XOOOOXOXX.XXOXXXX.XX.XOOO.X.XXXX.O.
 16 66 5 3 1 [[5, 0.8449], [11, 0.0597], [66, 0.0304], [3, 0.0225], [21, 0.0189]]....O.......XOXXOOOOX.XOOOOXX...XXOX.XXXXX.XXX.X.X.X.X.XXX.X.X.X...XOX.X.....O.X.
 17 66 4 8 1 [[4, 0.997], [70, 0.0017], [5, 0.001], [6, 0.0001], [76, 0.0]]..OO.....OOXXXXOO.XXXOOOOO.OOOOOXXXOOOXXXX.XXOXXX..XX.XX..XX.X..XXXXXX..X........
GAME 190 40 [[40, 1.0], [31, 0.0], [39, 0.0]]2 31 [[31, 0.4207], [39, 0.2552], [41, 0.2552]]4 41 [[39, 0.2916], [41, 0.2916], [30, 0.1191]]6 59 [[59, 0.4843], [32, 0.4274], [42, 0.0486]]8 39 [[39, 0.4056], [57, 0.3433], [32, 0.2006]]10 32 [[57, 0.9449], [51, 0.0222], [32, 0.0149]]12 33 [[60, 0.3643], [33, 0.3422], [57, 0.2352]]14 60 [[60, 0.7843], [30, 0.1397], [38, 0.0387]]after 15 ...............................XXX.....XXXO....OOOOOO.....OXX....................16 38 [[30, 0.6543], [38, 0.189], [57, 0.1338]]18 61 [[57, 0.5929], [61, 0.2237], [34, 0.1786]]20 57 [[57, 0.99], [30, 0.005], [29, 0.0015]]22 66 [[66, 0.9972], [67, 0.0028], [30, 0.0]]24 67 [[67, 0.9997], [65, 0.0002], [69, 0.0001]]26 69 [[69, 0.9768], [30, 0.0085], [29, 0.0053]]28 77 [[77, 1.0], [70, 0.0], [25, 0.0]]30 30 [[30, 0.8754], [25, 0.0419], [70, 0.0418]]after 31 ........................O....OXXXXO..OXXXXOO...OOOOOO...OXOXXX...OXX.X.......X...32 70 [[46, 0.288], [70, 0.2329], [25, 0.2215]]after 47 ..OX.......OXO.....OXXO.O...OOXXXXO..OXXXXOO..OOOOOOO..XOXOXXX.OXOXX.XX...X..X...after 109 .OOOOOOOOOO.OOOOOOOOOOOOOOO.OO.OX.OOXOO.O.OOOXOOOOOOO.XXOXOXXXO.XOXX.XXXX.XX.X.X.
GAME 240 40 [[40, 1.0], [31, 0.0], [39, 0.0]]2 31 [[31, 0.4207], [39, 0.2552], [41, 0.2552]]4 41 [[39, 0.2916], [41, 0.2916], [30, 0.1191]]6 59 [[59, 0.4843], [32, 0.4274], [42, 0.0486]]8 39 [[39, 0.4056], [57, 0.3433], [32, 0.2006]]10 57 [[57, 0.9449], [51, 0.0222], [32, 0.0149]]12 38 [[38, 0.4378], [56, 0.178], [51, 0.1513]]14 66 [[66, 0.4999], [51, 0.4676], [67, 0.0325]]after 15 ...............................X......XXXXO....OOOOO....OXOX......X..............16 60 [[60, 0.9734], [67, 0.0103], [32, 0.0103]]18 67 [[67, 0.9997], [32, 0.0002], [65, 0.0001]]20 65 [[69, 0.5262], [65, 0.4098], [32, 0.0288]]22 69 [[69, 0.8477], [32, 0.0451], [37, 0.0376]]24 77 [[77, 1.0], [37, 0.0], [32, 0.0]]26 32 [[37, 0.3987], [32, 0.2645], [30, 0.1959]]28 30 [[30, 0.8601], [37, 0.089], [43, 0.018]]30 55 [[55, 0.7926], [21, 0.1196], [28, 0.0497]]after 31 ....................O.O......OXXXO...OXXXXO...OOOOOOO..XOXOXX....XXX.X.......X...32 23 [[23, 1.0], [64, 0.0], [28, 0.0]]after 47 ....X.......XOXX...OOOOXOO.XXOXXXOO..OXXXXO...OOOOOOO.OXOXOXX...XXXX.X...O...X...after 90 XXOOO.OO.XO.OOO..O.OOOO.OO.XXOXXXOOOOOXXOXOO.OOOOOOOOOOXOXOXXXOXXXXX.XXXX.XX.X.X.

python

import jsonsrc=open('/work/design_v5.py').read()src=src.replace("capture_guard=0.)",                "capture_guard=0., support_adj=1.25, support_dist2=.75, support_dist3=.3125)")src=src.replace("w(5,u(0,t),I,-1.25*iw)","""support={1:p['support_adj'],2:p['support_dist2'],3:p['support_dist3']}.get(d,0.)                w(5,u(0,t),I,-support)""")open('/work/design_v6.py','w').write(src)exec(src)cfg=json.load(open('/work/config_v5.json'))cfg.update(support_adj=2.5,support_dist2=.35,support_dist3=.1,isolation_threshold=1.8)config=install(cfg)json.dump(config,open('/work/config_v6.json','w'),indent=2)print(config)

import jsonsrc=open('/work/design_v5.py').read()src=src.replace("capture_guard=0.)",                "capture_guard=0., support_adj=1.25, support_dist2=.75, support_dist3=.3125)")src=src.replace("w(5,u(0,t),I,-1.25*iw)","""support={1:p['support_adj'],2:p['support_dist2'],3:p['support_dist3']}.get(d,0.)                w(5,u(0,t),I,-support)""")open('/work/design_v6.py','w').write(src)exec(src)cfg=json.load(open('/work/config_v5.json'))cfg.update(support_adj=2.5,support_dist2=.35,support_dist3=.1,isolation_threshold=1.8)config=install(cfg)json.dump(config,open('/work/config_v6.json','w'),indent=2)print(config)

OUTPUT of Python run 13 of at most 120:

{'capture': 11.0, 'capture_extra': 3.0, 'save': 7.0, 'danger': 10.0, 'eye': 22.0, 'isolation': 3.0, 'isolation_threshold': 1.8, 'base': 1.3, 'pass_bias': -1.0, 'attack': 1.3, 'own_adj': 2.0, 'opp_adj': 1.5, 'opening': 3.0, 'save_danger': 2.0, 'safe2': 1.0, 'safe3': 1.0, 'cap_safety': 1.0, 'crowd': 3.0, 'temperature_scale': 2.5, 'defend': 1.0, 'first_center': 20.0, 'triangle': 1.5, 'capture_guard': 0.2, 'support_adj': 2.5, 'support_dist2': 0.35, 'support_dist3': 0.1}

{'capture': 11.0, 'capture_extra': 3.0, 'save': 7.0, 'danger': 10.0, 'eye': 22.0, 'isolation': 3.0, 'isolation_threshold': 1.8, 'base': 1.3, 'pass_bias': -1.0, 'attack': 1.3, 'own_adj': 2.0, 'opp_adj': 1.5, 'opening': 3.0, 'save_danger': 2.0, 'safe2': 1.0, 'safe3': 1.0, 'cap_safety': 1.0, 'crowd': 3.0, 'temperature_scale': 2.5, 'defend': 1.0, 'first_center': 20.0, 'triangle': 1.5, 'capture_guard': 0.2, 'support_adj': 2.5, 'support_dist2': 0.35, 'support_dist3': 0.1}

Games pass 6 of 10 done.

{"ok": true, "pass": 6, "kind": "games", "per_opponent": {"random": {"games": 8, "score": 1.0, "wins": 8, "losses": 0, "score_as_black": 1.0, "score_as_white": 1.0, "mean_margin": 50.25, "mean_length": 106.25, "stones_captured_per_game": 56.25, "stones_lost_per_game": 0.0, "passes_per_game": 1.75, "games_with_a_pass_before_move_60": 0, "end_two_passes": 8, "end_repetition": 0, "end_move_limit": 0}, "strong": {"games": 96, "score": 0.65625, "wins": 63, "losses": 33, "score_as_black": 0.6041666865348816, "score_as_white": 0.7083333134651184, "mean_margin": 43.78125, "mean_length": 113.80208333333333, "stones_captured_per_game": 63.208333333333336, "stones_lost_per_game": 8.260416666666666, "passes_per_game": 3.3958333333333335, "games_with_a_pass_before_move_60": 0, "end_two_passes": 94, "end_repetition": 0, "end_move_limit": 2}, "weak": {"games": 24, "score": 0.9166666865348816, "wins": 22, "losses": 2, "score_as_black": 0.9166666865348816, "score_as_white": 0.9166666865348816, "mean_margin": 71.625, "mean_length": 123.25, "stones_captured_per_game": 85.625, "stones_lost_per_game": 2.9583333333333335, "passes_per_game": 3.2916666666666665, "games_with_a_pass_before_move_60": 0, "end_two_passes": 23, "end_repetition": 1, "end_move_limit": 0}}, "files": "/work/pass_6/", "readme": "games.json: one entry per game (opponent, net_color, result +1/-1, margin = net's area score minus the opponent's incl. komi, end, stones captured / lost, moves: ply, by, move (row*9+col, 81 = pass), and for the network's moves its top-5 [move, probability]). boards.npy: (games, 162, 81) int8, the board AFTER each ply (1 Black, -1 White, 0 empty; index row*9+col); valid.npy: (games, 162) which plies were played."}

{"ok": true, "pass": 6, "kind": "games", "per_opponent": {"random": {"games": 8, "score": 1.0, "wins": 8, "losses": 0, "score_as_black": 1.0, "score_as_white": 1.0, "mean_margin": 50.25, "mean_length": 106.25, "stones_captured_per_game": 56.25, "stones_lost_per_game": 0.0, "passes_per_game": 1.75, "games_with_a_pass_before_move_60": 0, "end_two_passes": 8, "end_repetition": 0, "end_move_limit": 0}, "strong": {"games": 96, "score": 0.65625, "wins": 63, "losses": 33, "score_as_black": 0.6041666865348816, "score_as_white": 0.7083333134651184, "mean_margin": 43.78125, "mean_length": 113.80208333333333, "stones_captured_per_game": 63.208333333333336, "stones_lost_per_game": 8.260416666666666, "passes_per_game": 3.3958333333333335, "games_with_a_pass_before_move_60": 0, "end_two_passes": 94, "end_repetition": 0, "end_move_limit": 2}, "weak": {"games": 24, "score": 0.9166666865348816, "wins": 22, "losses": 2, "score_as_black": 0.9166666865348816, "score_as_white": 0.9166666865348816, "mean_margin": 71.625, "mean_length": 123.25, "stones_captured_per_game": 85.625, "stones_lost_per_game": 2.9583333333333335, "passes_per_game": 3.2916666666666665, "games_with_a_pass_before_move_60": 0, "end_two_passes": 23, "end_repetition": 1, "end_move_limit": 0}}, "files": "/work/pass_6/", "readme": "games.json: one entry per game (opponent, net_color, result +1/-1, margin = net's area score minus the opponent's incl. komi, end, stones captured / lost, moves: ply, by, move (row*9+col, 81 = pass), and for the network's moves its top-5 [move, probability]). boards.npy: (games, 162, 81) int8, the board AFTER each ply (1 Black, -1 White, 0 empty; index row*9+col); valid.npy: (games, 162) which plies were played."}

Passes left: 4.

python

import jsonsrc=open('/work/design_v6.py').read()src=src.replace("support_dist3=.3125)","support_dist3=.3125, safety_threshold=2.)")src=src.replace("b(5,D,2-len(adj[q]))","b(5,D,p['safety_threshold']-len(adj[q]))")open('/work/design_v7.py','w').write(src)exec(src)cfg=json.load(open('/work/config_v5.json'))cfg.update(temperature_scale=3.5,safety_threshold=1.7,safe2=.7,safe3=1.3,           cap_safety=.85,danger=14.)config=install(cfg)json.dump(config,open('/work/config_v7.json','w'),indent=2)print(config)

import jsonsrc=open('/work/design_v6.py').read()src=src.replace("support_dist3=.3125)","support_dist3=.3125, safety_threshold=2.)")src=src.replace("b(5,D,2-len(adj[q]))","b(5,D,p['safety_threshold']-len(adj[q]))")open('/work/design_v7.py','w').write(src)exec(src)cfg=json.load(open('/work/config_v5.json'))cfg.update(temperature_scale=3.5,safety_threshold=1.7,safe2=.7,safe3=1.3,           cap_safety=.85,danger=14.)config=install(cfg)json.dump(config,open('/work/config_v7.json','w'),indent=2)print(config)

OUTPUT of Python run 14 of at most 120:

{'capture': 11.0, 'capture_extra': 3.0, 'save': 7.0, 'danger': 14.0, 'eye': 22.0, 'isolation': 3.0, 'isolation_threshold': 2.5, 'base': 1.3, 'pass_bias': -1.0, 'attack': 1.3, 'own_adj': 2.0, 'opp_adj': 1.5, 'opening': 3.0, 'save_danger': 2.0, 'safe2': 0.7, 'safe3': 1.3, 'cap_safety': 0.85, 'crowd': 3.0, 'temperature_scale': 3.5, 'defend': 1.0, 'first_center': 20.0, 'triangle': 1.5, 'capture_guard': 0.2, 'support_adj': 1.25, 'support_dist2': 0.75, 'support_dist3': 0.3125, 'safety_threshold': 1.7}

{'capture': 11.0, 'capture_extra': 3.0, 'save': 7.0, 'danger': 14.0, 'eye': 22.0, 'isolation': 3.0, 'isolation_threshold': 2.5, 'base': 1.3, 'pass_bias': -1.0, 'attack': 1.3, 'own_adj': 2.0, 'opp_adj': 1.5, 'opening': 3.0, 'save_danger': 2.0, 'safe2': 0.7, 'safe3': 1.3, 'cap_safety': 0.85, 'crowd': 3.0, 'temperature_scale': 3.5, 'defend': 1.0, 'first_center': 20.0, 'triangle': 1.5, 'capture_guard': 0.2, 'support_adj': 1.25, 'support_dist2': 0.75, 'support_dist3': 0.3125, 'safety_threshold': 1.7}

Games pass 7 of 10 done.

{"ok": true, "pass": 7, "kind": "games", "per_opponent": {"random": {"games": 8, "score": 1.0, "wins": 8, "losses": 0, "score_as_black": 1.0, "score_as_white": 1.0, "mean_margin": 68.5, "mean_length": 114.125, "stones_captured_per_game": 65.125, "stones_lost_per_game": 0.625, "passes_per_game": 1.125, "games_with_a_pass_before_move_60": 0, "end_two_passes": 8, "end_repetition": 0, "end_move_limit": 0}, "strong": {"games": 96, "score": 0.78125, "wins": 75, "losses": 21, "score_as_black": 0.7291666865348816, "score_as_white": 0.8333333134651184, "mean_margin": 55.02083206176758, "mean_length": 115.77083333333333, "stones_captured_per_game": 64.42708333333333, "stones_lost_per_game": 7.177083333333333, "passes_per_game": 1.9895833333333333, "games_with_a_pass_before_move_60": 0, "end_two_passes": 92, "end_repetition": 2, "end_move_limit": 2}, "weak": {"games": 24, "score": 0.9583333134651184, "wins": 23, "losses": 1, "score_as_black": 0.9166666865348816, "score_as_white": 1.0, "mean_margin": 72.83333587646484, "mean_length": 123.41666666666667, "stones_captured_per_game": 87.95833333333333, "stones_lost_per_game": 2.3333333333333335, "passes_per_game": 2.9166666666666665, "games_with_a_pass_before_move_60": 0, "end_two_passes": 24, "end_repetition": 0, "end_move_limit": 0}}, "files": "/work/pass_7/", "readme": "games.json: one entry per game (opponent, net_color, result +1/-1, margin = net's area score minus the opponent's incl. komi, end, stones captured / lost, moves: ply, by, move (row*9+col, 81 = pass), and for the network's moves its top-5 [move, probability]). boards.npy: (games, 162, 81) int8, the board AFTER each ply (1 Black, -1 White, 0 empty; index row*9+col); valid.npy: (games, 162) which plies were played."}

{"ok": true, "pass": 7, "kind": "games", "per_opponent": {"random": {"games": 8, "score": 1.0, "wins": 8, "losses": 0, "score_as_black": 1.0, "score_as_white": 1.0, "mean_margin": 68.5, "mean_length": 114.125, "stones_captured_per_game": 65.125, "stones_lost_per_game": 0.625, "passes_per_game": 1.125, "games_with_a_pass_before_move_60": 0, "end_two_passes": 8, "end_repetition": 0, "end_move_limit": 0}, "strong": {"games": 96, "score": 0.78125, "wins": 75, "losses": 21, "score_as_black": 0.7291666865348816, "score_as_white": 0.8333333134651184, "mean_margin": 55.02083206176758, "mean_length": 115.77083333333333, "stones_captured_per_game": 64.42708333333333, "stones_lost_per_game": 7.177083333333333, "passes_per_game": 1.9895833333333333, "games_with_a_pass_before_move_60": 0, "end_two_passes": 92, "end_repetition": 2, "end_move_limit": 2}, "weak": {"games": 24, "score": 0.9583333134651184, "wins": 23, "losses": 1, "score_as_black": 0.9166666865348816, "score_as_white": 1.0, "mean_margin": 72.83333587646484, "mean_length": 123.41666666666667, "stones_captured_per_game": 87.95833333333333, "stones_lost_per_game": 2.3333333333333335, "passes_per_game": 2.9166666666666665, "games_with_a_pass_before_move_60": 0, "end_two_passes": 24, "end_repetition": 0, "end_move_limit": 0}}, "files": "/work/pass_7/", "readme": "games.json: one entry per game (opponent, net_color, result +1/-1, margin = net's area score minus the opponent's incl. komi, end, stones captured / lost, moves: ply, by, move (row*9+col, 81 = pass), and for the network's moves its top-5 [move, probability]). boards.npy: (games, 162, 81) int8, the board AFTER each ply (1 Black, -1 White, 0 empty; index row*9+col); valid.npy: (games, 162) which plies were played."}

Passes left: 3.

python

import jsonsrc=open('/work/design_v7.py').read()src=src.replace("safety_threshold=2.)",                "safety_threshold=2., pass_deficit=0., deficit_cap=16.)")src=src.replace("""b(1,487,1)    for q in range(81):        w(1,3*q,487,-1)        w(1,3*q+1,487,-1)""","w(1,2,487,1)")src=src.replace("if q==40: w(5,487,P,p['first_center'])","# Empty-board opening bonus is applied in layer 6.")src=src.replace("    line=[-1.2",'''    # Reconstruct the empty-board flag after the positional phase signal.    w(5,486,486,30); b(5,486,-29)    # A conservative late-game score proxy, used only to discourage    # resignation-by-passing while apparently behind.    b(5,487,7.5)    w(5,487,487,-15)    for t in range(81):        w(5,u(1,t),487,1)        w(5,u(0,t),487,-1)    w(6,487,487,1)    w(6,487,488,1); b(6,488,-p['deficit_cap'])    line=[-1.2''')src=src.replace("w(6,P,u(5,q),1)","""w(6,P,u(5,q),1)        if q==40: w(6,486,u(5,q),p['first_center'])""")src=src.replace("a['out.b'][81]=p['pass_bias']","""a['out.b'][81]=p['pass_bias']    a['out.W'][487,81]=-p['pass_deficit']    a['out.W'][488,81]=p['pass_deficit']""")open('/work/design_v8.py','w').write(src)exec(src)cfg=json.load(open('/work/config_v5.json'))cfg.update(pass_deficit=.5,deficit_cap=16.)config=install(cfg)json.dump(config,open('/work/config_v8.json','w'),indent=2)print(config)

import jsonsrc=open('/work/design_v7.py').read()src=src.replace("safety_threshold=2.)",                "safety_threshold=2., pass_deficit=0., deficit_cap=16.)")src=src.replace("""b(1,487,1)    for q in range(81):        w(1,3*q,487,-1)        w(1,3*q+1,487,-1)""","w(1,2,487,1)")src=src.replace("if q==40: w(5,487,P,p['first_center'])","# Empty-board opening bonus is applied in layer 6.")src=src.replace("    line=[-1.2",'''    # Reconstruct the empty-board flag after the positional phase signal.    w(5,486,486,30); b(5,486,-29)    # A conservative late-game score proxy, used only to discourage    # resignation-by-passing while apparently behind.    b(5,487,7.5)    w(5,487,487,-15)    for t in range(81):        w(5,u(1,t),487,1)        w(5,u(0,t),487,-1)    w(6,487,487,1)    w(6,487,488,1); b(6,488,-p['deficit_cap'])    line=[-1.2''')src=src.replace("w(6,P,u(5,q),1)","""w(6,P,u(5,q),1)        if q==40: w(6,486,u(5,q),p['first_center'])""")src=src.replace("a['out.b'][81]=p['pass_bias']","""a['out.b'][81]=p['pass_bias']    a['out.W'][487,81]=-p['pass_deficit']    a['out.W'][488,81]=p['pass_deficit']""")open('/work/design_v8.py','w').write(src)exec(src)cfg=json.load(open('/work/config_v5.json'))cfg.update(pass_deficit=.5,deficit_cap=16.)config=install(cfg)json.dump(config,open('/work/config_v8.json','w'),indent=2)print(config)

OUTPUT of Python run 15 of at most 120:

{'capture': 11.0, 'capture_extra': 3.0, 'save': 7.0, 'danger': 10.0, 'eye': 22.0, 'isolation': 3.0, 'isolation_threshold': 2.5, 'base': 1.3, 'pass_bias': -1.0, 'attack': 1.3, 'own_adj': 2.0, 'opp_adj': 1.5, 'opening': 3.0, 'save_danger': 2.0, 'safe2': 1.0, 'safe3': 1.0, 'cap_safety': 1.0, 'crowd': 3.0, 'temperature_scale': 2.5, 'defend': 1.0, 'first_center': 20.0, 'triangle': 1.5, 'capture_guard': 0.2, 'support_adj': 1.25, 'support_dist2': 0.75, 'support_dist3': 0.3125, 'safety_threshold': 2.0, 'pass_deficit': 0.5, 'deficit_cap': 16.0}

{'capture': 11.0, 'capture_extra': 3.0, 'save': 7.0, 'danger': 10.0, 'eye': 22.0, 'isolation': 3.0, 'isolation_threshold': 2.5, 'base': 1.3, 'pass_bias': -1.0, 'attack': 1.3, 'own_adj': 2.0, 'opp_adj': 1.5, 'opening': 3.0, 'save_danger': 2.0, 'safe2': 1.0, 'safe3': 1.0, 'cap_safety': 1.0, 'crowd': 3.0, 'temperature_scale': 2.5, 'defend': 1.0, 'first_center': 20.0, 'triangle': 1.5, 'capture_guard': 0.2, 'support_adj': 1.25, 'support_dist2': 0.75, 'support_dist3': 0.3125, 'safety_threshold': 2.0, 'pass_deficit': 0.5, 'deficit_cap': 16.0}

Games pass 8 of 10 done.

{"ok": true, "pass": 8, "kind": "games", "per_opponent": {"random": {"games": 8, "score": 1.0, "wins": 8, "losses": 0, "score_as_black": 1.0, "score_as_white": 1.0, "mean_margin": 75.5, "mean_length": 117.75, "stones_captured_per_game": 74.5, "stones_lost_per_game": 0.375, "passes_per_game": 1.25, "games_with_a_pass_before_move_60": 0, "end_two_passes": 8, "end_repetition": 0, "end_move_limit": 0}, "strong": {"games": 96, "score": 0.78125, "wins": 75, "losses": 21, "score_as_black": 0.7291666865348816, "score_as_white": 0.8333333134651184, "mean_margin": 55.89583206176758, "mean_length": 121.42708333333333, "stones_captured_per_game": 76.9375, "stones_lost_per_game": 9.270833333333334, "passes_per_game": 2.2083333333333335, "games_with_a_pass_before_move_60": 0, "end_two_passes": 94, "end_repetition": 0, "end_move_limit": 2}, "weak": {"games": 24, "score": 1.0, "wins": 24, "losses": 0, "score_as_black": 1.0, "score_as_white": 1.0, "mean_margin": 78.33333587646484, "mean_length": 129.54166666666666, "stones_captured_per_game": 91.08333333333333, "stones_lost_per_game": 3.1666666666666665, "passes_per_game": 3.125, "games_with_a_pass_before_move_60": 0, "end_two_passes": 24, "end_repetition": 0, "end_move_limit": 0}}, "files": "/work/pass_8/", "readme": "games.json: one entry per game (opponent, net_color, result +1/-1, margin = net's area score minus the opponent's incl. komi, end, stones captured / lost, moves: ply, by, move (row*9+col, 81 = pass), and for the network's moves its top-5 [move, probability]). boards.npy: (games, 162, 81) int8, the board AFTER each ply (1 Black, -1 White, 0 empty; index row*9+col); valid.npy: (games, 162) which plies were played."}

{"ok": true, "pass": 8, "kind": "games", "per_opponent": {"random": {"games": 8, "score": 1.0, "wins": 8, "losses": 0, "score_as_black": 1.0, "score_as_white": 1.0, "mean_margin": 75.5, "mean_length": 117.75, "stones_captured_per_game": 74.5, "stones_lost_per_game": 0.375, "passes_per_game": 1.25, "games_with_a_pass_before_move_60": 0, "end_two_passes": 8, "end_repetition": 0, "end_move_limit": 0}, "strong": {"games": 96, "score": 0.78125, "wins": 75, "losses": 21, "score_as_black": 0.7291666865348816, "score_as_white": 0.8333333134651184, "mean_margin": 55.89583206176758, "mean_length": 121.42708333333333, "stones_captured_per_game": 76.9375, "stones_lost_per_game": 9.270833333333334, "passes_per_game": 2.2083333333333335, "games_with_a_pass_before_move_60": 0, "end_two_passes": 94, "end_repetition": 0, "end_move_limit": 2}, "weak": {"games": 24, "score": 1.0, "wins": 24, "losses": 0, "score_as_black": 1.0, "score_as_white": 1.0, "mean_margin": 78.33333587646484, "mean_length": 129.54166666666666, "stones_captured_per_game": 91.08333333333333, "stones_lost_per_game": 3.1666666666666665, "passes_per_game": 3.125, "games_with_a_pass_before_move_60": 0, "end_two_passes": 24, "end_repetition": 0, "end_move_limit": 0}}, "files": "/work/pass_8/", "readme": "games.json: one entry per game (opponent, net_color, result +1/-1, margin = net's area score minus the opponent's incl. komi, end, stones captured / lost, moves: ply, by, move (row*9+col, 81 = pass), and for the network's moves its top-5 [move, probability]). boards.npy: (games, 162, 81) int8, the board AFTER each ply (1 Black, -1 White, 0 empty; index row*9+col); valid.npy: (games, 162) which plies were played."}

Passes left: 2.

python

import jsonexec(open('/work/design_v8.py').read())cfg=json.load(open('/work/config_v5.json'))cfg['temperature_scale']=4.config=install(cfg)json.dump(config,open('/work/config_v9.json','w'),indent=2)print('Installed the established tactical/shape design with sharper move selection.')print(config)

import jsonexec(open('/work/design_v8.py').read())cfg=json.load(open('/work/config_v5.json'))cfg['temperature_scale']=4.config=install(cfg)json.dump(config,open('/work/config_v9.json','w'),indent=2)print('Installed the established tactical/shape design with sharper move selection.')print(config)

OUTPUT of Python run 16 of at most 120:

Installed the established tactical/shape design with sharper move selection.{'capture': 11.0, 'capture_extra': 3.0, 'save': 7.0, 'danger': 10.0, 'eye': 22.0, 'isolation': 3.0, 'isolation_threshold': 2.5, 'base': 1.3, 'pass_bias': -1.0, 'attack': 1.3, 'own_adj': 2.0, 'opp_adj': 1.5, 'opening': 3.0, 'save_danger': 2.0, 'safe2': 1.0, 'safe3': 1.0, 'cap_safety': 1.0, 'crowd': 3.0, 'temperature_scale': 4.0, 'defend': 1.0, 'first_center': 20.0, 'triangle': 1.5, 'capture_guard': 0.2, 'support_adj': 1.25, 'support_dist2': 0.75, 'support_dist3': 0.3125, 'safety_threshold': 2.0, 'pass_deficit': 0.0, 'deficit_cap': 16.0}

Installed the established tactical/shape design with sharper move selection.{'capture': 11.0, 'capture_extra': 3.0, 'save': 7.0, 'danger': 10.0, 'eye': 22.0, 'isolation': 3.0, 'isolation_threshold': 2.5, 'base': 1.3, 'pass_bias': -1.0, 'attack': 1.3, 'own_adj': 2.0, 'opp_adj': 1.5, 'opening': 3.0, 'save_danger': 2.0, 'safe2': 1.0, 'safe3': 1.0, 'cap_safety': 1.0, 'crowd': 3.0, 'temperature_scale': 4.0, 'defend': 1.0, 'first_center': 20.0, 'triangle': 1.5, 'capture_guard': 0.2, 'support_adj': 1.25, 'support_dist2': 0.75, 'support_dist3': 0.3125, 'safety_threshold': 2.0, 'pass_deficit': 0.0, 'deficit_cap': 16.0}

Games pass 9 of 10 done.

{"ok": true, "pass": 9, "kind": "games", "per_opponent": {"random": {"games": 8, "score": 1.0, "wins": 8, "losses": 0, "score_as_black": 1.0, "score_as_white": 1.0, "mean_margin": 65.75, "mean_length": 113.625, "stones_captured_per_game": 69.875, "stones_lost_per_game": 0.375, "passes_per_game": 1.0, "games_with_a_pass_before_move_60": 0, "end_two_passes": 8, "end_repetition": 0, "end_move_limit": 0}, "strong": {"games": 96, "score": 0.7916666865348816, "wins": 76, "losses": 20, "score_as_black": 0.7083333134651184, "score_as_white": 0.875, "mean_margin": 55.47916793823242, "mean_length": 118.34375, "stones_captured_per_game": 71.58333333333333, "stones_lost_per_game": 8.447916666666666, "passes_per_game": 2.3020833333333335, "games_with_a_pass_before_move_60": 0, "end_two_passes": 94, "end_repetition": 0, "end_move_limit": 2}, "weak": {"games": 24, "score": 0.8333333134651184, "wins": 20, "losses": 4, "score_as_black": 0.75, "score_as_white": 0.9166666865348816, "mean_margin": 67.95833587646484, "mean_length": 122.45833333333333, "stones_captured_per_game": 76.5, "stones_lost_per_game": 4.416666666666667, "passes_per_game": 2.625, "games_with_a_pass_before_move_60": 0, "end_two_passes": 23, "end_repetition": 1, "end_move_limit": 0}}, "files": "/work/pass_9/", "readme": "games.json: one entry per game (opponent, net_color, result +1/-1, margin = net's area score minus the opponent's incl. komi, end, stones captured / lost, moves: ply, by, move (row*9+col, 81 = pass), and for the network's moves its top-5 [move, probability]). boards.npy: (games, 162, 81) int8, the board AFTER each ply (1 Black, -1 White, 0 empty; index row*9+col); valid.npy: (games, 162) which plies were played."}

{"ok": true, "pass": 9, "kind": "games", "per_opponent": {"random": {"games": 8, "score": 1.0, "wins": 8, "losses": 0, "score_as_black": 1.0, "score_as_white": 1.0, "mean_margin": 65.75, "mean_length": 113.625, "stones_captured_per_game": 69.875, "stones_lost_per_game": 0.375, "passes_per_game": 1.0, "games_with_a_pass_before_move_60": 0, "end_two_passes": 8, "end_repetition": 0, "end_move_limit": 0}, "strong": {"games": 96, "score": 0.7916666865348816, "wins": 76, "losses": 20, "score_as_black": 0.7083333134651184, "score_as_white": 0.875, "mean_margin": 55.47916793823242, "mean_length": 118.34375, "stones_captured_per_game": 71.58333333333333, "stones_lost_per_game": 8.447916666666666, "passes_per_game": 2.3020833333333335, "games_with_a_pass_before_move_60": 0, "end_two_passes": 94, "end_repetition": 0, "end_move_limit": 2}, "weak": {"games": 24, "score": 0.8333333134651184, "wins": 20, "losses": 4, "score_as_black": 0.75, "score_as_white": 0.9166666865348816, "mean_margin": 67.95833587646484, "mean_length": 122.45833333333333, "stones_captured_per_game": 76.5, "stones_lost_per_game": 4.416666666666667, "passes_per_game": 2.625, "games_with_a_pass_before_move_60": 0, "end_two_passes": 23, "end_repetition": 1, "end_move_limit": 0}}, "files": "/work/pass_9/", "readme": "games.json: one entry per game (opponent, net_color, result +1/-1, margin = net's area score minus the opponent's incl. komi, end, stones captured / lost, moves: ply, by, move (row*9+col, 81 = pass), and for the network's moves its top-5 [move, probability]). boards.npy: (games, 162, 81) int8, the board AFTER each ply (1 Black, -1 White, 0 empty; index row*9+col); valid.npy: (games, 162) which plies were played."}

Passes left: 1.

python

import json, numpy as npexec(open('/work/design_v5.py').read())final_config=json.load(open('/work/config_v5.json'))install(final_config)for name in NAMES:    value=load(name)    assert value.shape==tuple(SHAPES[name])    assert value.dtype==np.float32    assert np.isfinite(value).all()assert set(saved())==set(NAMES)json.dump(final_config,open('/work/final_config.json','w'),indent=2)print('Final policy saved and all 14 tensors checked.')print('Design: liberty-based group signals, guarded captures, atari escapes,')print('self-atari avoidance, eye preservation, connection/shape preferences,')print('local influence, center opening, and conservative passing.')print('No training, search, or sandbox forward evaluations were used.')

import json, numpy as npexec(open('/work/design_v5.py').read())final_config=json.load(open('/work/config_v5.json'))install(final_config)for name in NAMES:    value=load(name)    assert value.shape==tuple(SHAPES[name])    assert value.dtype==np.float32    assert np.isfinite(value).all()assert set(saved())==set(NAMES)json.dump(final_config,open('/work/final_config.json','w'),indent=2)print('Final policy saved and all 14 tensors checked.')print('Design: liberty-based group signals, guarded captures, atari escapes,')print('self-atari avoidance, eye preservation, connection/shape preferences,')print('local influence, center opening, and conservative passing.')print('No training, search, or sandbox forward evaluations were used.')

OUTPUT of Python run 17 of at most 120:

Final policy saved and all 14 tensors checked.Design: liberty-based group signals, guarded captures, atari escapes,self-atari avoidance, eye preservation, connection/shape preferences,local influence, center opening, and conservative passing.No training, search, or sandbox forward evaluations were used.

Final policy saved and all 14 tensors checked.Design: liberty-based group signals, guarded captures, atari escapes,self-atari avoidance, eye preservation, connection/shape preferences,local influence, center opening, and conservative passing.No training, search, or sandbox forward evaluations were used.

Games pass 10 of 10 done.

{"ok": true, "pass": 10, "kind": "games", "per_opponent": {"random": {"games": 8, "score": 1.0, "wins": 8, "losses": 0, "score_as_black": 1.0, "score_as_white": 1.0, "mean_margin": 73.75, "mean_length": 122.25, "stones_captured_per_game": 93.75, "stones_lost_per_game": 0.375, "passes_per_game": 1.25, "games_with_a_pass_before_move_60": 0, "end_two_passes": 8, "end_repetition": 0, "end_move_limit": 0}, "strong": {"games": 96, "score": 0.7604166865348816, "wins": 73, "losses": 23, "score_as_black": 0.7083333134651184, "score_as_white": 0.8125, "mean_margin": 50.72916793823242, "mean_length": 118.03125, "stones_captured_per_game": 68.20833333333333, "stones_lost_per_game": 9.875, "passes_per_game": 2.375, "games_with_a_pass_before_move_60": 0, "end_two_passes": 92, "end_repetition": 1, "end_move_limit": 3}, "weak": {"games": 24, "score": 0.9166666865348816, "wins": 22, "losses": 2, "score_as_black": 0.9166666865348816, "score_as_white": 0.9166666865348816, "mean_margin": 73.45833587646484, "mean_length": 123.58333333333333, "stones_captured_per_game": 84.33333333333333, "stones_lost_per_game": 3.4166666666666665, "passes_per_game": 2.2916666666666665, "games_with_a_pass_before_move_60": 0, "end_two_passes": 24, "end_repetition": 0, "end_move_limit": 0}}, "files": "/work/pass_10/", "readme": "games.json: one entry per game (opponent, net_color, result +1/-1, margin = net's area score minus the opponent's incl. komi, end, stones captured / lost, moves: ply, by, move (row*9+col, 81 = pass), and for the network's moves its top-5 [move, probability]). boards.npy: (games, 162, 81) int8, the board AFTER each ply (1 Black, -1 White, 0 empty; index row*9+col); valid.npy: (games, 162) which plies were played."}

{"ok": true, "pass": 10, "kind": "games", "per_opponent": {"random": {"games": 8, "score": 1.0, "wins": 8, "losses": 0, "score_as_black": 1.0, "score_as_white": 1.0, "mean_margin": 73.75, "mean_length": 122.25, "stones_captured_per_game": 93.75, "stones_lost_per_game": 0.375, "passes_per_game": 1.25, "games_with_a_pass_before_move_60": 0, "end_two_passes": 8, "end_repetition": 0, "end_move_limit": 0}, "strong": {"games": 96, "score": 0.7604166865348816, "wins": 73, "losses": 23, "score_as_black": 0.7083333134651184, "score_as_white": 0.8125, "mean_margin": 50.72916793823242, "mean_length": 118.03125, "stones_captured_per_game": 68.20833333333333, "stones_lost_per_game": 9.875, "passes_per_game": 2.375, "games_with_a_pass_before_move_60": 0, "end_two_passes": 92, "end_repetition": 1, "end_move_limit": 3}, "weak": {"games": 24, "score": 0.9166666865348816, "wins": 22, "losses": 2, "score_as_black": 0.9166666865348816, "score_as_white": 0.9166666865348816, "mean_margin": 73.45833587646484, "mean_length": 123.58333333333333, "stones_captured_per_game": 84.33333333333333, "stones_lost_per_game": 3.4166666666666665, "passes_per_game": 2.2916666666666665, "games_with_a_pass_before_move_60": 0, "end_two_passes": 24, "end_repetition": 0, "end_move_limit": 0}}, "files": "/work/pass_10/", "readme": "games.json: one entry per game (opponent, net_color, result +1/-1, margin = net's area score minus the opponent's incl. komi, end, stones captured / lost, moves: ply, by, move (row*9+col, 81 = pass), and for the network's moves its top-5 [move, probability]). boards.npy: (games, 162, 81) int8, the board AFTER each ply (1 Black, -1 White, 0 empty; index row*9+col); valid.npy: (games, 162) which plies were played."}

Passes left: 0.

Saved all weights. The network implements handcrafted liberty propagation, guarded captures, atari escapes, self-atari avoidance, eye preservation, and connection/shape preferences.

Across both tests of the final weights, it won:

79.7% against strong (153/192)
95.8% against weak (46/48)
100% against random (16/16)

Its tactical discipline and connected shapes should generalize, though hidden-opponent performance remains uncertain. No training, search, or sandbox forward evaluations were used.