FuryBench Testing Framework

home toIntegralPlentyChessObsidianCaissaQuanticadeSerendipitySpaghetSchoenemannClarityIlluminaPawnocchioVineAnura
No logs for event exist
Active : 25 Machines / 740 Threads / 577.3 MNPS
Priority 0
Quanticade DarkNeutrino
tunable
SMP VLTC tune preparing for release time
smp diff20.0+0.20
TUNE   134 parameters
ITER   1,666 / 5,000
GAMES  6,666 / 20,000
Caissa Witek
lmp-off-in-check
Search: Skip LMP when in check
diff10.0+0.10
ELO    -0.35 +- 1.23 (95%)
LLR    -2.23 (-2.94, 2.94) [0.00, 2.00]
GAMES  N: 80402 W: 19248 L: 19328 D: 41826
PENTA  [218, 9721, 20403, 9641, 218]
Priority -1
Caissa Witek
nmp-margin-ramp
Search: Depth-ramped null move margin on corrected eval
diff10.0+0.10
ELO    0.00 +- 0.00 (95%)
LLR    0.00 (-2.94, 2.94) [0.00, 2.00]
GAMES  N: 0 W: 0 L: 0 D: 0
PENTA  [0, 0, 0, 0, 0]
Caissa Witek
pawn-hist-statscore
diff10.0+0.10
ELO    -2.01 +- 3.94 (95%)
LLR    -0.65 (-2.94, 2.94) [0.00, 2.00]
GAMES  N: 8116 W: 1923 L: 1970 D: 4223
PENTA  [20, 1032, 2012, 963, 31]
Priority -10
Caissa Witek
see-quiet-threshold
SEE: Fix early exit for quiet moves with a positive threshold
diff8.0+0.08
ELO    0.78 +- 2.73 (95%)
LLR    0.69 (-2.94, 2.94) [-2.00, 0.00]
GAMES  N: 16474 W: 3954 L: 3917 D: 8603
PENTA  [65, 1919, 4218, 1984, 51]
Finished
Caissa Witek
master
NN: Output buckets split by queens on the board 16 output subnets instead of 8: piece count bucket * 2 + "any queen on the board". Net version 17. ELO 1.40 +- 1.08 (95%) SPRT 10.0+0.10s Threads=1 Hash=16MB LLR 3.01 (-2.94, 2.94) [0.00, 2.00] GAMES N: 106708 W: 25536 L: 25106 D: 56066 PENTA [401, 12544, 26993, 13056, 360] https://furybench.com/test/8043/ ELO 1.35 +- 1.28 (95%) SPRT 60.0+0.60s Threads=1 Hash=128MB LLR 2.95 (-2.94, 2.94) [-0.50, 2.50] GAMES N: 67256 W: 15894 L: 15633 D: 35729 PENTA [33, 7757, 17779, 8034, 25] https://furybench.com/test/8045/ Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
diff10.0+0.10
ELO    18.07 +- 4.55 (95%)
GAMES  N: 6274/40000 W: 1663 L: 1337 D: 3274
PENTA  [16, 638, 1527, 916, 40]
Caissa Witek
legal-evasions
diff60.0+0.60
ELO    5.35 +- 3.69 (95%)
LLR    2.00 (-2.94, 2.94) [-0.50, 2.50]
GAMES  N: 8060 W: 1913 L: 1789 D: 4358
PENTA  [2, 879, 2148, 995, 6]
Caissa Witek
legal-evasions
diffN=10000
ELO    6.26 +- 3.18 (95%)
LLR    3.01 (-2.94, 2.94) [0.00, 2.00]
GAMES  N: 23490 W: 7860 L: 7437 D: 8193
PENTA  [829, 2435, 4846, 2754, 881]
Caissa Witek
legal-evasions
diff10.0+0.10
ELO    7.42 +- 3.02 (95%)
LLR    2.95 (-2.94, 2.94) [0.00, 2.00]
GAMES  N: 13020 W: 3213 L: 2935 D: 6872
PENTA  [22, 1421, 3364, 1663, 40]
Caissa Witek
legal-movegen
diff10.0+0.10
ELO    -0.72 +- 1.57 (95%)
LLR    -2.95 (-2.94, 2.94) [-0.50, 2.50]
GAMES  N: 49168 W: 11711 L: 11813 D: 25644
PENTA  [159, 5859, 12627, 5803, 136]
Caissa Witek
legal-movegen-2
diff8.0+0.08
ELO    -2.15 +- 2.33 (95%)
LLR    -2.95 (-2.94, 2.94) [-0.50, 2.50]
GAMES  N: 23160 W: 5482 L: 5625 D: 12053
PENTA  [94, 2835, 5864, 2694, 93]
Caissa Witek
legal-movegen
diffN=10000
ELO    2.18 +- 2.47 (95%)
LLR    1.32 (-2.94, 2.94) [0.00, 2.00]
GAMES  N: 39010 W: 12706 L: 12461 D: 13843
PENTA  [1396, 4229, 8064, 4366, 1450]
Caissa Witek
see-pinned-recapture
diffN=10000
ELO    10.32 +- 4.11 (95%)
LLR    3.13 (-2.94, 2.94) [0.00, 2.00]
GAMES  N: 14038 W: 4816 L: 4399 D: 4823
PENTA  [437, 1477, 2931, 1580, 594]
Caissa Witek
see-pinned-recapture
diff10.0+0.10
ELO    1.91 +- 1.37 (95%)
LLR    2.95 (-2.94, 2.94) [0.00, 2.00]
GAMES  N: 66182 W: 16070 L: 15707 D: 34405
PENTA  [205, 7811, 16716, 8134, 225]
Caissa Witek
history-piece-to-factorizer
diff10.0+0.10
ELO    -4.21 +- 3.09 (95%)
LLR    -1.95 (-2.94, 2.94) [0.00, 2.00]
GAMES  N: 12882 W: 3002 L: 3158 D: 6722
PENTA  [41, 1620, 3283, 1448, 49]
Quanticade DarkNeutrino
non-root-iir
search.c: Disable IIR in root node
diff10.0+0.10
ELO    -0.03 +- 1.10 (95%)
LLR    2.99 (-2.94, 2.94) [-3.00, 1.00]
GAMES  N: 96452 W: 22716 L: 22725 D: 51011
PENTA  [165, 11376, 25152, 11369, 164]
Caissa Witek
queen-output-buckets
NN: Output buckets split by queens on the board 16 output subnets instead of 8: piece count bucket * 2 + "any queen on the board". Net version 17. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
diff60.0+0.60
ELO    1.35 +- 1.28 (95%)
LLR    2.95 (-2.94, 2.94) [-0.50, 2.50]
GAMES  N: 67256 W: 15894 L: 15633 D: 35729
PENTA  [33, 7757, 17779, 8034, 25]
Clarity Vast
cm_2_new
new data mini net test before i kick off a full datagen run
diff8.0+0.08
ELO    -21.64 +- 8.47 (95%)
LLR    -2.27 (-2.25, 2.89) [0.00, 3.00]
GAMES  N: 2508 W: 594 L: 750 D: 1164
PENTA  [67, 344, 544, 276, 23]
Clarity Vast
cm_2_new
fixed nodes
diffN=20000
ELO    -31.78 +- 9.55 (95%)
LLR    -2.63 (-2.25, 2.89) [0.00, 3.00]
GAMES  N: 2138 W: 423 L: 618 D: 1097
PENTA  [62, 336, 438, 201, 32]
Caissa Witek
queen-output-buckets
NN: Output buckets split by queens on the board 16 output subnets instead of 8: piece count bucket * 2 + "any queen on the board". Net version 17. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
diff10.0+0.10
ELO    1.40 +- 1.08 (95%)
LLR    3.01 (-2.94, 2.94) [0.00, 2.00]
GAMES  N: 106708 W: 25536 L: 25106 D: 56066
PENTA  [401, 12544, 26993, 13056, 360]