Finished
WorldRecklesslinear-cutoff-count-lmrdiff8.0+0.08
LLR: 3.05 (-2.25, 2.89) [0.00, 4.00]
Games: 19856 W: 4816 L: 4615 D: 10425
Ptnml(0-2): 101, 2292, 4932, 2511, 92
WorldRecklesslmr-history-tablediff8.0+0.08
LLR: -2.42 (-2.25, 2.89) [0.00, 4.00]
Games: 36118 W: 8625 L: 8626 D: 18867
Ptnml(0-2): 178, 4311, 9070, 4334, 166
history table for LMR results (stm-from-to)
WorldRecklessstyx_v1diff8.0+0.08
LLR: -2.27 (-2.25, 2.89) [0.00, 4.00]
Games: 12604 W: 3115 L: 3183 D: 6306
Ptnml(0-2): 93, 1559, 3050, 1523, 77
Styx's first network (retraining on latest bullet + rounding in quantization)
WorldRecklessstyx_v1diffN=25000
LLR: -2.30 (-2.25, 2.89) [0.00, 4.00]
Games: 73364 W: 23453 L: 23295 D: 26616
Ptnml(0-2): 2489, 8357, 14897, 8385, 2554
Styx's first network (retraining on latest bullet + rounding in quantization)
WorldRecklesslikely-fail-highdiff8.0+0.08
LLR: -2.28 (-2.25, 2.89) [0.00, 4.00]
Games: 32714 W: 7774 L: 7779 D: 17161
Ptnml(0-2): 162, 3857, 8311, 3878, 149
apply cut node reduction for a potential fail high
WorldRecklesspawn-hist2diff8.0+0.08
LLR: -1.79 (-2.25, 2.89) [0.00, 4.00]
Games: 11518 W: 2710 L: 2757 D: 6051
Ptnml(0-2): 50, 1438, 2825, 1401, 45
static eval pawn history updates
WorldRecklessnmp-history-2diff8.0+0.08
LLR: -2.63 (-2.25, 2.89) [0.00, 4.00]
Games: 24740 W: 5978 L: 6024 D: 12738
Ptnml(0-2): 124, 3042, 6078, 3008, 118
increase history range with offset as compensation
WorldRecklesslikely-fail-high-2diff8.0+0.08
LLR: -2.10 (-2.25, 2.89) [0.00, 4.00]
Games: 9526 W: 2279 L: 2347 D: 4900
Ptnml(0-2): 52, 1200, 2313, 1160, 38
more reduction for a potential fail high
WorldRecklesslikely-fail-lowdiff8.0+0.08
LLR: -2.38 (-2.25, 2.89) [0.00, 4.00]
Games: 4420 W: 1029 L: 1125 D: 2266
Ptnml(0-2): 29, 565, 1104, 497, 15
no ttpv reduction for a potential fail low
WorldRecklessgranular-ultimate-patch-2diff8.0+0.08
LLR: -2.04 (-2.25, 2.89) [0.00, 4.00]
Games: 8754 W: 2079 L: 2147 D: 4528
Ptnml(0-2): 43, 1127, 2097, 1075, 35
scale the ultimate patch linearly - take 2
WorldRecklesspawn-hist3diff8.0+0.08
LLR: -2.42 (-2.25, 2.89) [0.00, 4.00]
Games: 9610 W: 2289 L: 2370 D: 4951
Ptnml(0-2): 42, 1201, 2393, 1134, 35
factorize pawn history on piece-to
WorldRecklessgranular-ultimate-patchdiff8.0+0.08
LLR: -1.56 (-2.25, 2.89) [0.00, 4.00]
Games: 6616 W: 1565 L: 1616 D: 3435
Ptnml(0-2): 18, 840, 1655, 765, 30
scale the ultimate patch linearly
WorldRecklessnmp-butterfly-and-pawn-corrhistdiff8.0+0.08
LLR: -1.65 (-2.25, 2.89) [0.00, 4.00]
Games: 27170 W: 6546 L: 6539 D: 14085
Ptnml(0-2): 122, 3264, 6799, 3285, 115
augment butterfly history with correction history based on the pawn structure
WorldRecklesspv-history-bucketdiff8.0+0.08
LLR: -1.74 (-2.25, 2.89) [0.00, 4.00]
Games: 5836 W: 1362 L: 1424 D: 3050
Ptnml(0-2): 31, 729, 1453, 681, 24
add pv bucket to quiet history
WorldRecklesspawn-histdiff8.0+0.08
LLR: -1.01 (-2.25, 2.89) [0.00, 4.00]
Games: 6568 W: 1571 L: 1598 D: 3399
Ptnml(0-2): 31, 824, 1611, 777, 41
pawn history with Pere's fix
WorldRecklessnmp-butterfly-and-pawn-corrhist-2diff8.0+0.08
LLR: -2.10 (-2.25, 2.89) [0.00, 4.00]
Games: 14340 W: 3449 L: 3502 D: 7389
Ptnml(0-2): 72, 1747, 3588, 1688, 75
same as nmp-butterfly-and-pawn-corrhist, but sums histories instead of averaging
WorldRecklessmore-nmp-historydiff8.0+0.08
LLR: -1.44 (-2.25, 2.89) [0.00, 4.00]
Games: 22010 W: 5273 L: 5272 D: 11465
Ptnml(0-2): 105, 2639, 5511, 2650, 100
increase maximum correction
WorldRecklessrfp-static-eval-guarddiff8.0+0.08
LLR: -1.72 (-2.25, 2.89) [0.00, 4.00]
Games: 7824 W: 1856 L: 1911 D: 4057
Ptnml(0-2): 45, 952, 1961, 921, 33
skip rfp if static_eval >= beta
WorldRecklessrazoring-verificationdiff8.0+0.08
LLR: -2.26 (-2.25, 2.89) [0.00, 4.00]
Games: 8212 W: 1938 L: 2018 D: 4256
Ptnml(0-2): 37, 1071, 1970, 991, 37
WorldRecklessnull-move-historydiff8.0+0.08
LLR: 2.94 (-2.25, 2.89) [0.00, 4.00]
Games: 52018 W: 12534 L: 12241 D: 27243
Ptnml(0-2): 244, 6121, 12991, 6404, 249
adjust nmp margin by null move history (indexed by stm & previous move)
WorldRecklesslmr-cn-killersdiff8.0+0.08
LLR: -2.32 (-2.25, 2.89) [0.00, 4.00]
Games: 19612 W: 4710 L: 4757 D: 10145
Ptnml(0-2): 100, 2372, 4907, 2329, 98
don't apply cut node reduction for killers
WorldRecklessnmp-tt-fail-highdiff8.0+0.08
LLR: -2.30 (-2.25, 2.89) [0.00, 4.00]
Games: 22310 W: 5326 L: 5364 D: 11620
Ptnml(0-2): 115, 2701, 5548, 2689, 102
more nmp when tt entry suggests a fail high
WorldRecklessnmp-return-verified-scorediff8.0+0.08
LLR: -2.25 (-2.25, 2.89) [0.00, 4.00]
Games: 12562 W: 2924 L: 2987 D: 6651
Ptnml(0-2): 59, 1441, 3335, 1396, 50
WorldRecklesscorrhist-more-momentumdiff8.0+0.08
LLR: -2.26 (-2.25, 2.89) [0.00, 4.00]
Games: 14856 W: 3487 L: 3545 D: 7824
Ptnml(0-2): 66, 1815, 3709, 1787, 51
larger bonus for corrhist updates (2 * diff * depth)
WorldRecklessrfp-null-movediff8.0+0.08
LLR: -2.37 (-2.25, 2.89) [0.00, 4.00]
Games: 7968 W: 1858 L: 1942 D: 4168
Ptnml(0-2): 37, 997, 1998, 917, 35
more rfp when no tt move (likely to fail high)