FuryBench Testing Framework

home toIntegralPlentyChessObsidianCaissaQuanticadeSerendipitySpaghetSchoenemannClarityIlluminaPawnocchioVineAnura
Active : 1 Machines / 30 Threads / 29.37 MNPS
Priority 0
Quanticade DarkNeutrino
scale-adjust
nnue.c: Adjust eval scale so it matches net60 on average
diff10.0+0.10
ELO    -2.06 +- 5.03 (95%)
LLR    -0.62 (-2.25, 2.89) [0.00, 3.00]
GAMES  N: 4392 W: 1012 L: 1038 D: 2342
PENTA  [4, 521, 1173, 493, 5]
Finished
Quanticade DarkNeutrino
[Integral] main
Lets see how bad it is
diff10.0+0.10
ELO    -48.77 +- 6.94 (95%)
GAMES  N: 2488/10000 W: 428 L: 775 D: 1285
PENTA  [13, 487, 581, 160, 3]
Quanticade DarkNeutrino
master
net61 (#439)
diff60.0+0.60
ELO    95.59 +- 13.95 (95%)
GAMES  N: 544/40000 W: 211 L: 65 D: 268
PENTA  [0, 16, 96, 158, 2]
Quanticade DarkNeutrino
master
net61 (#439)
diff10.0+0.10
ELO    113.60 +- 4.71 (95%)
GAMES  N: 6292/10000 W: 2716 L: 729 D: 2847
PENTA  [8, 158, 1030, 1739, 211]
Quanticade DarkNeutrino
net61
net61
diff60.0+0.60
ELO    10.33 +- 4.27 (95%)
LLR    2.89 (-2.25, 2.89) [0.00, 3.00]
GAMES  N: 5788 W: 1464 L: 1292 D: 3032
PENTA  [1, 570, 1582, 738, 3]
Quanticade DarkNeutrino
net61
TIPP chonker relabel + 3B of 25k node data
diff10.0+0.10
ELO    10.81 +- 4.53 (95%)
LLR    2.89 (-2.25, 2.89) [0.00, 3.00]
GAMES  N: 5982 W: 1525 L: 1339 D: 3118
PENTA  [15, 636, 1508, 812, 20]
Quanticade DarkNeutrino
tipp-chonker
Fully trained
diffN=20000
ELO    11.69 +- 2.16 (95%)
GAMES  N: 40002/40000 W: 12040 L: 10695 D: 17267
PENTA  [837, 4143, 8608, 5664, 749]
Quanticade DarkNeutrino
alpha-reduction
search.c: Reduce other moves if we found at least one score improvement
diff60.0+0.60
ELO    -0.10 +- 1.39 (95%)
LLR    -2.26 (-2.25, 2.89) [0.00, 3.00]
GAMES  N: 53304 W: 12282 L: 12297 D: 28725
PENTA  [8, 5860, 14934, 5839, 11]
Quanticade DarkNeutrino
tipp-chonker
diffN=50000
ELO    37.18 +- 2.39 (95%)
GAMES  N: 25310/40000 W: 7923 L: 5225 D: 12162
PENTA  [133, 2039, 5816, 4331, 336]
Quanticade DarkNeutrino
alpha-reduction
search.c: Reduce other moves if we found at least one score improvement
diff10.0+0.10
ELO    0.46 +- 1.31 (95%)
LLR    -0.94 (-2.25, 2.89) [0.00, 3.00]
GAMES  N: 65558 W: 15403 L: 15317 D: 34838
PENTA  [78, 7636, 17281, 7690, 94]
Quanticade DarkNeutrino
se-triple-noisy-margin
Source: Triple noisy margin
diff60.0+0.60
ELO    -1.22 +- 2.16 (95%)
LLR    -2.25 (-2.25, 2.89) [0.00, 3.00]
GAMES  N: 22204 W: 5157 L: 5235 D: 11812
PENTA  [3, 2508, 6158, 2430, 3]
Quanticade DarkNeutrino
alpha-reduction
search.c: Reduce other moves if we found at least one score improvement
diff10.0+0.10
ELO    0.00 +- 0.00 (95%)
LLR    0.00 (-2.25, 2.89) [0.00, 3.00]
GAMES  N: 0 W: 0 L: 0 D: 0
PENTA  [0, 0, 0, 0, 0]
Quanticade DarkNeutrino
master
search.c: Add correction margin to SE (#438)
diff10.0+0.10
ELO    71.74 +- 3.68 (95%)
GAMES  N: 10020/10000 W: 3635 L: 1595 D: 4790
PENTA  [12, 528, 2050, 2248, 172]
Quanticade DarkNeutrino
master
search.c: Add correction margin to SE (#438)
diff60.0+0.60
ELO    76.36 +- 3.33 (95%)
GAMES  N: 10000/10000 W: 3613 L: 1450 D: 4937
PENTA  [1, 391, 2126, 2408, 74]
Quanticade DarkNeutrino
alpha-raises-lmr
search.c: Reduce more based on alpha raises
diff60.0+0.60
ELO    -2.48 +- 2.77 (95%)
LLR    -2.26 (-2.25, 2.89) [0.00, 3.00]
GAMES  N: 13426 W: 3024 L: 3120 D: 7282
PENTA  [2, 1527, 3750, 1433, 1]
Quanticade DarkNeutrino
hse-4
Tree: Use locally explored values
diff60.0+0.60
ELO    -1.24 +- 2.15 (95%)
LLR    -2.25 (-2.25, 2.89) [0.00, 3.00]
GAMES  N: 21912 W: 5050 L: 5128 D: 11734
PENTA  [5, 2426, 6168, 2356, 1]
Quanticade DarkNeutrino
hse-4
Values explored over 500 iterations using chess tuning tools.
diff10.0+0.10
ELO    0.43 +- 1.56 (95%)
LLR    -0.73 (-2.25, 2.89) [0.00, 3.00]
GAMES  N: 46640 W: 10990 L: 10932 D: 24718
PENTA  [73, 5409, 12302, 5459, 77]
Quanticade DarkNeutrino
alpha-raises-lmr
search.c: Reduce more based on alpha raises
diff10.0+0.10
ELO    0.46 +- 1.17 (95%)
LLR    -1.18 (-2.25, 2.89) [0.00, 3.00]
GAMES  N: 83348 W: 19740 L: 19629 D: 43979
PENTA  [134, 9719, 21850, 9844, 127]
Quanticade DarkNeutrino
mo-checks-good
search.c: Give move ordering bonus to quiet moves that give check
diff10.0+0.10
ELO    -2.67 +- 2.94 (95%)
LLR    -2.28 (-2.25, 2.89) [0.00, 3.00]
GAMES  N: 13384 W: 3088 L: 3191 D: 7105
PENTA  [23, 1628, 3491, 1529, 21]
Quanticade DarkNeutrino
conthist-6
history.c: Add conthist 6 ply
diff10.0+0.10
ELO    -2.02 +- 2.66 (95%)
LLR    -2.27 (-2.25, 2.89) [0.00, 3.00]
GAMES  N: 16522 W: 3862 L: 3958 D: 8702
PENTA  [33, 2011, 4261, 1931, 25]
Quanticade DarkNeutrino
master
25K node datagen.
diffN=25000
ELO    -0.08 +- 0.10 (95%)
GAMES  N: 30000010/30000000 W: 10528304 L: 10535074 D: 8936632
Quanticade DarkNeutrino
[Obsidian] main
search.c: Add correction margin to SE (#438)
diff10.0+0.10
ELO    -9.83 +- 5.15 (95%)
GAMES  N: 5020/5000 W: 1191 L: 1333 D: 2496
PENTA  [24, 699, 1203, 563, 21]
Quanticade DarkNeutrino
master
same as before but LTC
diff60.0+0.60
ELO    30.93 +- 7.81 (95%)
GAMES  N: 1802/3000 W: 503 L: 343 D: 956
PENTA  [0, 140, 464, 294, 3]
Quanticade DarkNeutrino
master
Master against last dev prerelease to see whats up
diff10.0+0.10
ELO    32.77 +- 9.60 (95%)
GAMES  N: 1244/10000 W: 357 L: 240 D: 647
PENTA  [2, 91, 322, 202, 5]
Quanticade DarkNeutrino
se-correction
praying is becoming mandatory part of engine dev
diff60.0+0.60
ELO    1.13 +- 1.90 (95%)
LLR    0.64 (-2.25, 2.89) [0.00, 3.00]
GAMES  N: 28214 W: 6568 L: 6476 D: 15170
PENTA  [8, 3024, 7946, 3126, 3]
Quanticade DarkNeutrino
hse-tuned
crash smp diff20.0+0.20
ELO    -0.00 +- 3.18 (95%)
LLR    -0.36 (-2.25, 2.89) [0.00, 3.00]
GAMES  N: 9688 W: 2217 L: 2217 D: 5254
PENTA  [3, 1010, 2816, 1014, 1]