SolsticeBench

Finished
TestTCEngineAuthorResult
corr-ext SPRT diff
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
imp-lmp SPRT diff
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
root-contempt SPRT diff
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
corr-pstruct SPRT diff
search: abstract pawn-shape correction history
4.0+0.04SolsticeDragonLLR: -1.40 (-2.94, 2.94) [0.00, 5.00]
Games: 790 W: 167 L: 225 D: 398
Ptnml(0-2): 16, 128, 158, 84, 9
corr-passer SPRT diff
search: passed-pawn correction history
4.0+0.04SolsticeDragonLLR: -0.94 (-2.94, 2.94) [0.00, 5.00]
Games: 512 W: 108 L: 145 D: 259
Ptnml(0-2): 8, 81, 111, 52, 4
corr-king SPRT diff
search: king-shelter correction history
4.0+0.04SolsticeDragonLLR: -0.81 (-2.94, 2.94) [0.00, 5.00]
Games: 698 W: 149 L: 178 D: 371
Ptnml(0-2): 6, 103, 158, 78, 4
corr-threat SPRT diff
search: pawn-threat correction history
4.0+0.04SolsticeDragonLLR: -0.81 (-2.94, 2.94) [0.00, 5.00]
Games: 512 W: 107 L: 139 D: 266
Ptnml(0-2): 6, 87, 100, 59, 4
corr-mat SPRT diff
search: material-signature correction history
4.0+0.04SolsticeDragonLLR: -0.13 (-2.94, 2.94) [0.00, 5.00]
Games: 1034 W: 258 L: 259 D: 517
Ptnml(0-2): 16, 131, 223, 132, 15
killer-shift SPRT diff
search: killers and low-ply history carried to the next move, two plies on
4.0+0.04SolsticeDragonLLR: -0.65 (-2.94, 2.94) [0.00, 5.00]
Games: 876 W: 214 L: 239 D: 423
Ptnml(0-2): 24, 109, 189, 100, 16
knobtune141 SPRT diff
search: knob tune #141 values (11 knobs, 32k games at 4+0.04)
4.0+0.04SolsticeDragonLLR: -0.47 (-2.94, 2.94) [0.00, 5.00]
Games: 9692 W: 2360 L: 2338 D: 4994
Ptnml(0-2): 119, 1190, 2223, 1178, 136
mdp SPRT diff
search: mate distance pruning (non-regression)
4.0+0.04SolsticeDragonLLR: 0.16 (-2.94, 2.94) [-5.00, 0.00]
Games: 7424 W: 1728 L: 1752 D: 3944
Ptnml(0-2): 81, 881, 1814, 853, 83
eval-order SPRT diff
search: order quiets by the network's child eval at deep nodes
4.0+0.04SolsticeDragonLLR: -1.04 (-2.94, 2.94) [0.00, 5.00]
Games: 3328 W: 764 L: 792 D: 1772
Ptnml(0-2): 39, 417, 784, 381, 43
lmr-noisy SPRT diff
search: LMR for captures and promotions
4.0+0.04SolsticeDragonLLR: 2.95 (-2.94, 2.94) [0.00, 5.00]
Games: 66000 W: 15843 L: 15445 D: 34712
Ptnml(0-2): 777, 7905, 15347, 8085, 886
master SPSA diff
Knob tune: only the 11 neutral knobs of 3e7c9c2 (improving in RFP/LMP/futility/LMR, history in futility and history pruning, NMP by eval margin, quadratic quiet SEE margin, score-drop TM), 4+0.04
4.0+0.04SolsticeDragonTuning 11 Parameters
500/500 Iterations
32014/32000 Games Played
persist-swap SPRT diff
speed: per-search setup 0.95 -> 0.47 ms (same search)
4.0+0.04SolsticeDragonLLR: -0.23 (-2.94, 2.94) [-5.00, 0.00]
Games: 2816 W: 666 L: 686 D: 1464
Ptnml(0-2): 32, 319, 726, 299, 32
spsa138 SPRT diff
search: SPSA tune #138 values (68 params, 64k games)
4.0+0.04SolsticeDragonLLR: 3.07 (-2.94, 2.94) [0.00, 5.00]
Games: 6496 W: 1603 L: 1451 D: 3442
Ptnml(0-2): 70, 735, 1506, 847, 90
master SPSA diff
SPSA round 4: 68 params on master b11e0fe (after tune #134 +16.5 and low-ply history +5.4), 2+0.02
2.0+0.02SolsticeDragonTuning 68 Parameters
1000/1000 Iterations
64024/64000 Games Played
spsa134 SPRT diff
search: SPSA tune #134 values (68 params, 64k games)
4.0+0.04SolsticeDragonLLR: 2.99 (-2.94, 2.94) [0.00, 5.00]
Games: 2736 W: 702 L: 572 D: 1462
Ptnml(0-2): 20, 282, 656, 368, 42
fut-hist SPRT diff
search: futility margin += quiet history / 64
4.0+0.04SolsticeDragonLLR: -1.10 (-2.94, 2.94) [0.00, 5.00]
Games: 896 W: 176 L: 217 D: 503
Ptnml(0-2): 18, 116, 210, 97, 7
master SPSA diff
SPSA round 3: 68 params on master 1909cdc (after #121 +19.9, #129 +8.8; LMR terms now in 1/1024 ply, +3 new LMR terms), 2+0.02
2.0+0.02SolsticeDragonTuning 68 Parameters
1000/1000 Iterations
64006/64000 Games Played
lowply-hist SPRT diff
search: low-ply history in quiet ordering
4.0+0.04SolsticeDragonLLR: 2.99 (-2.94, 2.94) [0.00, 5.00]
Games: 10740 W: 2632 L: 2465 D: 5643
Ptnml(0-2): 126, 1239, 2490, 1372, 143
capthist-see SPRT diff
search: capture history in ordering and in the good / bad capture split
4.0+0.04SolsticeDragonLLR: -0.58 (-2.94, 2.94) [0.00, 5.00]
Games: 1216 W: 279 L: 297 D: 640
Ptnml(0-2): 15, 155, 279, 151, 8
tt-keep-move SPRT diff
search: keep the TT move on a fail low
4.0+0.04SolsticeDragonLLR: 2.98 (-2.94, 2.94) [0.00, 5.00]
Games: 5648 W: 1417 L: 1274 D: 2957
Ptnml(0-2): 63, 609, 1346, 734, 72
iir-all SPRT diff
search: IIR at every node without a TT move
4.0+0.04SolsticeDragonLLR: -0.74 (-2.94, 2.94) [0.00, 5.00]
Games: 1408 W: 315 L: 339 D: 754
Ptnml(0-2): 13, 189, 331, 151, 20
mdp SPRT diff
search: mate distance pruning
4.0+0.04SolsticeDragonLLR: -0.11 (-2.94, 2.94) [0.00, 5.00]
Games: 296 W: 65 L: 68 D: 163
Ptnml(0-2): 4, 31, 81, 28, 4