FuryBench Testing Framework

home toIntegralPlentyChessObsidianCaissaQuanticadeSerendipitySpaghetSchoenemannClarityIlluminaPawnocchioVineAnura
Finished
Caissa Witek
hmc-input
Eval: Half-move counter inputs (net version 16) Net format version 16 adds 12 feature transformer rows after the piece inputs: bucket b = min(hmc / 8, 12) activates row b - 1 in both perspectives, no row below hmc 8. The incremental update compares the bucket at the two ends of the update path; the accumulator cache stays piece-only and the row is added when a node is built from it. `utils upgradeNet` converts a version 15 net with zero rows, which is an exact no-op (bench 5907427, same as master). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
diff10.0+0.10
ELO    -4.21 +- 2.01 (95%)
LLR    -2.95 (-2.94, 2.94) [0.00, 2.00]
GAMES  N: 13378 W: 4395 L: 4557 D: 4426
PENTA  [2, 855, 5137, 693, 2]
Caissa Witek
hmc-input
Eval: Half-move counter inputs (net version 16) Net format version 16 adds 12 feature transformer rows after the piece inputs: bucket b = min(hmc / 8, 12) activates row b - 1 in both perspectives, no row below hmc 8. The incremental update compares the bucket at the two ends of the update path; the accumulator cache stays piece-only and the row is added when a node is built from it. `utils upgradeNet` converts a version 15 net with zero rows, which is an exact no-op (bench 5907427, same as master). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
diff10.0+0.10
ELO    -8.63 +- 3.33 (95%)
LLR    -2.95 (-2.94, 2.94) [0.00, 2.00]
GAMES  N: 9180 W: 2127 L: 2355 D: 4698
PENTA  [0, 1122, 2577, 888, 3]
Integral Fury
factorised-ob
diff10.0+0.10
ELO    -1.66 +- 2.49 (95%)
LLR    -2.27 (-2.25, 2.89) [0.00, 3.00]
GAMES  N: 19102 W: 4619 L: 4710 D: 9773
PENTA  [57, 2269, 4977, 2204, 44]
Integral Fury
factorised-ob
diffN=20000
ELO    1.81 +- 1.35 (95%)
LLR    3.44 (-2.25, 2.89) [0.00, 3.00]
GAMES  N: 95164 W: 27140 L: 26645 D: 41379
PENTA  [1523, 11399, 21319, 11742, 1599]
Integral Fury
non_capthist_malus_decay
diff40.0+0.40
ELO    1.95 +- 1.53 (95%)
LLR    2.89 (-2.25, 2.89) [0.00, 3.00]
GAMES  N: 44314 W: 10913 L: 10664 D: 22737
PENTA  [11, 4783, 12323, 5026, 14]
PlentyChess Yoshie2000
refuted-by-probcut
diff8.0+0.08
ELO    1.01 +- 0.80 (95%)
LLR    2.91 (-2.25, 2.89) [0.00, 2.50]
GAMES  N: 180496 W: 44044 L: 43519 D: 92933
PENTA  [280, 21019, 47108, 21578, 263]
Integral Fury
limit_history_static_eval_updates
diff10.0+0.10
ELO    -0.74 +- 2.37 (95%)
LLR    -1.53 (-2.25, 2.89) [0.00, 3.00]
GAMES  N: 20620 W: 4992 L: 5036 D: 10592
PENTA  [42, 2426, 5420, 2378, 44]
PlentyChess Yoshie2000
0193
diff8.0+0.08
ELO    0.18 +- 2.12 (95%)
LLR    -0.48 (-2.25, 2.89) [0.00, 2.50]
GAMES  N: 26306 W: 6500 L: 6486 D: 13320
PENTA  [49, 3137, 6773, 3139, 55]
PlentyChess Yoshie2000
0193
diff40.0+0.40
ELO    -2.74 +- 2.93 (95%)
LLR    -1.81 (-2.25, 2.89) [0.00, 2.50]
GAMES  N: 12692 W: 3060 L: 3160 D: 6472
PENTA  [1, 1530, 3390, 1418, 7]
Integral Fury
non_capthist_malus_decay
diff10.0+0.10
ELO    4.39 +- 2.69 (95%)
LLR    2.90 (-2.25, 2.89) [0.00, 3.00]
GAMES  N: 15980 W: 4076 L: 3874 D: 8030
PENTA  [20, 1788, 4189, 1956, 37]
PlentyChess Yoshie2000
0193
diffN=20000
ELO    1.89 +- 1.50 (95%)
LLR    2.91 (-2.25, 2.89) [0.00, 2.50]
GAMES  N: 73532 W: 20653 L: 20254 D: 32625
PENTA  [972, 8876, 16790, 9037, 1091]
Integral Fury
pc_imp_depth
diff10.0+0.10
ELO    -2.12 +- 2.69 (95%)
LLR    -2.26 (-2.25, 2.89) [0.00, 3.00]
GAMES  N: 15754 W: 3899 L: 3995 D: 7860
PENTA  [32, 1862, 4183, 1770, 30]
Integral Fury
seval_overestimate_lmr
diff10.0+0.10
ELO    -2.31 +- 4.64 (95%)
LLR    -0.81 (-2.25, 2.89) [0.00, 3.00]
GAMES  N: 5274 W: 1317 L: 1352 D: 2605
PENTA  [6, 636, 1391, 595, 9]
Integral Fury
d12_fp
diff10.0+0.10
ELO    0.93 +- 1.64 (95%)
LLR    0.40 (-2.25, 2.89) [0.00, 3.00]
GAMES  N: 42762 W: 10629 L: 10514 D: 21619
PENTA  [71, 4935, 11266, 5026, 83]
PlentyChess Yoshie2000
0193
diff8.0+0.08
ELO    6.58 +- 3.17 (95%)
LLR    2.89 (-2.25, 2.89) [0.00, 2.50]
GAMES  N: 11990 W: 3139 L: 2912 D: 5939
PENTA  [24, 1333, 3070, 1528, 40]