Finished
TestTCEngineAuthorResult
net: bullet-trained gen4 recipe (25% result), first bullet net
4.0+0.04SolsticeDragonLLR: 3.15 (-2.94, 2.94) [0.00, 5.00]
Games: 784 W: 297 L: 135 D: 352
Ptnml(0-2): 8, 43, 155, 151, 35
search: SPSA tune #166 values (round 2 after #164)
4.0+0.04SolsticeDragonLLR: 2.98 (-2.94, 2.94) [0.00, 5.00]
Games: 6962 W: 1772 L: 1617 D: 3573
Ptnml(0-2): 99, 772, 1605, 885, 120
SPSA round on master dca7749 (81 params, after #164)
2.0+0.02SolsticeDragonTuning 81 Parameters
1000/1000 Iterations
64002/64000 Games Played
search: SPSA tune #164 values (81 params, 64k games)
4.0+0.04SolsticeDragonLLR: 2.95 (-2.94, 2.94) [0.00, 5.00]
Games: 12910 W: 3197 L: 3017 D: 6696
Ptnml(0-2): 172, 1532, 2913, 1620, 218
SPSA round on master 5f7835d (81 params)
2.0+0.02SolsticeDragonTuning 81 Parameters
1000/1000 Iterations
64054/64000 Games Played
search: eval-difference history (2/raw unit, cap 1000, no base)
4.0+0.04SolsticeDragonLLR: -3.07 (-2.94, 2.94) [0.00, 5.00]
Games: 688 W: 113 L: 248 D: 327
Ptnml(0-2): 19, 144, 135, 45, 1
search: post-LMR continuation history update
4.0+0.04SolsticeDragonLLR: -2.47 (-2.94, 2.94) [0.00, 5.00]
Games: 10152 W: 2435 L: 2493 D: 5224
Ptnml(0-2): 139, 1239, 2372, 1193, 133
search: LMR +1 for quiets when the TT move is a capture
4.0+0.04SolsticeDragonLLR: -3.05 (-2.94, 2.94) [0.00, 5.00]
Games: 9486 W: 2266 L: 2352 D: 4868
Ptnml(0-2): 139, 1208, 2120, 1152, 124
eval: x (200 - halfmove)/200 (50-move damping)
4.0+0.04SolsticeDragonLLR: 0.27 (-2.94, 2.94) [0.00, 5.00]
Games: 20082 W: 4904 L: 4807 D: 10371
Ptnml(0-2): 260, 2438, 4574, 2483, 286
speed measurement only (nps per machine), not for merging: correction-history prefetch (asm volatile)
4.0+0.04SolsticeDragonLLR: 1.28 (-2.94, 2.94) [-5.00, 0.00]
Games: 4476 W: 1077 L: 1046 D: 2353
Ptnml(0-2): 40, 489, 1155, 508, 46
speed measurement only (nps per machine), not for merging: correction-history prefetch
4.0+0.04SolsticeDragonLLR: -0.50 (-2.94, 2.94) [-5.00, 0.00]
Games: 456 W: 94 L: 115 D: 247
Ptnml(0-2): 5, 64, 108, 49, 2
speed measurement only (nps per machine), not for merging: int16 corrections, early TT prefetch
4.0+0.04SolsticeDragonLLR: -0.12 (-2.94, 2.94) [-5.00, 0.00]
Games: 3932 W: 924 L: 944 D: 2064
Ptnml(0-2): 41, 438, 1023, 428, 36
speed measurement only (nps per machine), not for merging: AVX2 kernels on Skylake-SP
4.0+0.04SolsticeDragonLLR: -1.01 (-2.94, 2.94) [-5.00, 0.00]
Games: 1470 W: 349 L: 395 D: 726
Ptnml(0-2): 29, 162, 384, 146, 14
speed measurement only (nps per machine), not for merging: aligned weights, fused output layer, TT huge pages
4.0+0.04SolsticeDragonLLR: 0.29 (-2.94, 2.94) [-5.00, 0.00]
Games: 380 W: 102 L: 91 D: 187
Ptnml(0-2): 8, 35, 93, 46, 8
tt: quiescence may take only shallow (depth <= 2) entries of earlier searches
4.0+0.04SolsticeDragonLLR: -1.36 (-2.94, 2.94) [0.00, 5.00]
Games: 17474 W: 4173 L: 4155 D: 9146
Ptnml(0-2): 222, 2064, 4134, 2108, 209
tt: search generation, quiescence stores over old entries (LTC 16+0.16)
16.0+0.16SolsticeDragonLLR: -0.81 (-2.94, 2.94) [0.00, 5.00]
Games: 5752 W: 1290 L: 1298 D: 3164
Ptnml(0-2): 48, 687, 1395, 717, 29
tt: a search generation, so quiescence can store over old entries
4.0+0.04SolsticeDragonLLR: -1.24 (-2.94, 2.94) [0.00, 5.00]
Games: 6272 W: 1529 L: 1553 D: 3190
Ptnml(0-2): 74, 765, 1489, 727, 81
search: null-move threat reduces LMR
4.0+0.04SolsticeDragonLLR: -1.24 (-2.94, 2.94) [0.00, 5.00]
Games: 5348 W: 1305 L: 1334 D: 2709
Ptnml(0-2): 85, 648, 1240, 613, 88
qsearch: keep the TT move on a stand pat or fail low
4.0+0.04SolsticeDragonLLR: -1.18 (-2.94, 2.94) [0.00, 5.00]
Games: 7808 W: 1810 L: 1825 D: 4173
Ptnml(0-2): 92, 942, 1837, 955, 78
tt: a shallow bound does not overwrite a deeper entry for the same position
4.0+0.04SolsticeDragonLLR: -2.11 (-2.94, 2.94) [0.00, 5.00]
Games: 28224 W: 6715 L: 6682 D: 14827
Ptnml(0-2): 346, 3452, 6475, 3501, 338
eval: material scaling (300 + mat) / (300 + 78)
4.0+0.04SolsticeDragonLLR: 0.09 (-2.94, 2.94) [0.00, 5.00]
Games: 22848 W: 5446 L: 5346 D: 12056
Ptnml(0-2): 302, 2690, 5329, 2812, 291
search: history update for a quiet best move at PV nodes without a cutoff
4.0+0.04SolsticeDragonLLR: -2.18 (-2.94, 2.94) [0.00, 5.00]
Games: 976 W: 195 L: 278 D: 503
Ptnml(0-2): 19, 145, 228, 92, 4
search: extend PV nodes where the eval correction is large
4.0+0.04SolsticeDragonLLR: -0.59 (-2.94, 2.94) [0.00, 5.00]
Games: 1276 W: 316 L: 335 D: 625
Ptnml(0-2): 19, 161, 297, 142, 19
search: improving in late move pruning
4.0+0.04SolsticeDragonLLR: -0.71 (-2.94, 2.94) [0.00, 5.00]
Games: 1088 W: 252 L: 277 D: 559
Ptnml(0-2): 19, 144, 234, 137, 10
search: draw value from the root side's score (root contempt)
4.0+0.04SolsticeDragonLLR: 0.00 (-2.94, 2.94) [0.00, 5.00]
Games: 0 W: 0 L: 0 D: 0
Ptnml(0-2): 0, 0, 0, 0, 0