Stockfish Testing Queue

Finished - 40618 tests

14-02-15 rst remove_evasion_prunable diff
LLR: 2.96 (-2.94,2.94) [-4.00,0.00]
Total: 63564 W: 11727 L: 11790 D: 40047
sprt @ 15+0.05 th 1 Simplification test. Remove evasionPrunable from qsearch
14-02-15 Tha ttstore diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 21481 W: 3920 L: 3969 D: 13592
sprt @ 15+0.05 th 1 Speed up of TranspositionTable::store()
14-02-16 uri limit_ply2 diff
ELO: -3.61 +-2.8 (95%) LOS: 0.5%
Total: 22689 W: 4145 L: 4381 D: 14163
40000 @ 20 th 1 testing 26 plies
14-02-16 rst remove_evasion_prunable diff
LLR: 2.96 (-2.94,2.94) [-4.00,0.00]
Total: 166336 W: 25714 L: 26065 D: 114557
sprt @ 60+0.05 th 1 Simplification test. Remove evasionPrunable from qsearch
14-02-16 rst lmp diff
ELO: -4.38 +-3.1 (95%) LOS: 0.3%
Total: 20000 W: 4132 L: 4384 D: 11484
20000 @ 5+0.05 th 1 LMP attempt. Super short TC. Take 1
14-02-16 rst lmp diff
ELO: 0.33 +-3.1 (95%) LOS: 58.2%
Total: 20000 W: 4249 L: 4230 D: 11521
20000 @ 5+0.05 th 1 LMP attempt. Super short TC. Take 2
14-02-16 vdb probcut diff
LLR: -2.96 (-2.94,2.94) [-4.00,0.00]
Total: 14816 W: 2709 L: 2920 D: 9187
sprt @ 15+0.05 th 1 Another attempt at moving probcut into the main move loop, this time taking into account SEE values.
14-02-16 jki capGain2 diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 13973 W: 2550 L: 2619 D: 8804
sprt @ 15+0.05 th 1 CapGains, try 2
14-02-16 jos 3fold_fix diff
LLR: 2.96 (-2.94,2.94) [-4.00,0.00]
Total: 60681 W: 11028 L: 11079 D: 38574
sprt @ 15+0.05 th 1 Test as simplification/bugfix as vdb proposed.
14-02-16 Fis ext_promotions diff
LLR: -2.93 (-2.94,2.94) [-1.50,4.50]
Total: 68531 W: 12806 L: 12731 D: 42994
sprt @ 15+0.05 th 1 Extend promotions same way as checks.
14-02-16 uri faster_iterations diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 21034 W: 3842 L: 3892 D: 13300
sprt @ 15+0.05 th 1 jump by 2 every iteration and not by 1 to get deeper faster(idea is taken from Junior that even Jumps by 3) bench is constant because I cared not to do it when the depth is limited but it solves some test positions faster
14-02-17 uri faster_iterations diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 3895 W: 680 L: 775 D: 2440
sprt @ 15+0.05 th 1 Testing another idea that is simpler and and here all the benchs of odd depths are smaller from depth 9 to depth 23(maybe I need to tweak time management by reducing slowmover but I prefer first to test the simple change without modifications)
14-02-17 uri faster_iterations diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 4389 W: 732 L: 825 D: 2832
sprt @ 15+0.05 th 1 second try with reduction of slow_mover(if both tries give negative score I will stop the tries on increasing depth by more than 1)
14-02-17 jos 3fold_fix diff
LLR: -2.96 (-2.94,2.94) [-4.00,0.00]
Total: 138429 W: 20905 L: 21470 D: 96054
sprt @ 60+0.05 th 1 LTC: Test as simplification/bugfix as vdb proposed.
14-02-17 rst lmp diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 11630 W: 2116 L: 2191 D: 7323
sprt @ 15+0.05 th 1 LMP attempt. Final take 3.
14-02-17 rst invading_move diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 3836 W: 649 L: 744 D: 2443
sprt @ 15+0.05 th 1 Replace pos.advanced_pawn_push(move) by pos.invading_move(move). Proposal from Stephane Nicolet
14-02-17 rst multicut diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 15227 W: 2787 L: 2852 D: 9588
sprt @ 15+0.05 th 1 Simple MultiCut. Take 1
14-02-17 uri fix_extend_promotions diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 11655 W: 2114 L: 2189 D: 7352
sprt @ 15+0.05 th 1 the previous patch that failed with more than 50% extended also under promotion and I fix it to extend only queen promotions.
14-02-17 vdb probcutafternullmove diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 7522 W: 1359 L: 1445 D: 4718
sprt @ 15+0.05 th 1 Remove one of the preconditions for probcut.
14-02-17 joa fp_margin diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 13299 W: 2969 L: 3039 D: 7291
sprt @ 15+0.05 th 1 futility_margin 90
14-02-17 joa fp_margin diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 4864 W: 1014 L: 1109 D: 2741
sprt @ 15+0.05 th 1 futility margin 80
14-02-17 joa fp_margin diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 20680 W: 4597 L: 4645 D: 11438
sprt @ 15+0.05 th 1 futility margin 110
14-02-17 joa fp_margin diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 14992 W: 3290 L: 3355 D: 8347
sprt @ 15+0.05 th 1 futility margin 120
14-02-17 rst ttmove_reduce diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 10041 W: 1799 L: 1878 D: 6364
sprt @ 15+0.05 th 1 reduce ttMove if SE search indicates that ttMove is not singular
14-02-18 jhe bishop_pawns diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 18852 W: 3540 L: 3596 D: 11716
sprt @ 15+0.05 th 1 Increase midgame penalty to 12 per pawn.
14-02-18 pec still_first diff
LLR: 2.97 (-2.94,2.94) [-1.50,4.50]
Total: 18907 W: 3475 L: 3322 D: 12110
sprt @ 15+0.05 th 1 Distribute part of still first move time to other moves
14-02-18 jos options diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 9836 W: 2123 L: 2204 D: 5509
sprt @ 15+0.05 th 1 Try SPSA values after 20k iterations, although they look a bit suspicious.
14-02-19 gli still_first diff
LLR: 2.96 (-2.94,2.94) [0.00,6.00]
Total: 19044 W: 2997 L: 2811 D: 13236
sprt @ 60+0.05 th 1 LTC for PE: Distribute part of still first move time to other moves
14-02-19 pec still_first diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 8253 W: 1438 L: 1521 D: 5294
sprt @ 15+0.05 th 1 Redistribute more of still first move time to other moves. Also remove constant and incidentally redistribute time throughout the game. Note that this is simplification and there is no intention to test it at sprt [0, 6] as well as previous patch. Tested at STC with sprt[-1.5 4.5] to pre-select positive changes
14-02-19 Fis ext_queenattacks diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 4466 W: 779 L: 872 D: 2815
sprt @ 15+0.05 th 1 Extend attacks on the queen same way as checks since they are also mostly forcing moves.
14-02-19 pec still_first diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 5436 W: 956 L: 1047 D: 3433
sprt @ 15+0.05 th 1 Redistribute more of still first move time to other moves. No redistributing time throughout the game.
14-02-19 rst multicut diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 1991 W: 393 L: 497 D: 1101
sprt @ 15+0.05 th 1 Simple MultiCut. Take 2
14-02-19 hxi scaling diff
ELO: -0.47 +-3.1 (95%) LOS: 38.3%
Total: 20000 W: 4133 L: 4160 D: 11707
20000 @ 5+0.05 th 1 scaling measure variant A (p0=40,p2=60)
14-02-19 hxi scaling diff
ELO: -1.30 +-3.1 (95%) LOS: 20.6%
Total: 20000 W: 4160 L: 4235 D: 11605
20000 @ 5+0.05 th 1 scaling measure variant B (p0=48,p2=60)
14-02-19 hxi scaling diff
ELO: 0.66 +-3.1 (95%) LOS: 66.2%
Total: 20000 W: 4177 L: 4139 D: 11684
20000 @ 5+0.05 th 1 scaling measure variant C (p0=44,p2=56)
14-02-19 hxi scaling diff
ELO: 1.18 +-3.1 (95%) LOS: 77.4%
Total: 20000 W: 4115 L: 4047 D: 11838
20000 @ 5+0.05 th 1 scaling measure variant D (p0=44,p2=64)
14-02-19 Fis ext_pawn_kicks diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 4016 W: 658 L: 752 D: 2606
sprt @ 15+0.05 th 1 Extend moves where a pawn kicks another piece.
14-02-20 Fis ext_queenattacks diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 4567 W: 813 L: 906 D: 2848
sprt @ 15+0.05 th 1 Extending attacks on the queen but skipping if the queen is already attacked per Ralphs suggestion.
14-02-20 uri simplify_futility diff
LLR: 2.96 (-2.94,2.94) [-4.00,0.00]
Total: 12868 W: 2408 L: 2295 D: 8165
sprt @ 15+0.05 th 1 deleting condition from futility that I see no reason for it(it is a simplification so I use -4,0)
14-02-20 hxi piece_values diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 4592 W: 802 L: 895 D: 2895
sprt @ 15+0.05 th 1 try to move mobility bonus to piece values so avg mobility per piece and phase is near zero
14-02-20 pec slow_mover diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 35993 W: 6628 L: 6640 D: 22725
sprt @ 15+0.05 th 1 Redistribute time along game
14-02-21 pec slow_mover diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 6262 W: 1104 L: 1193 D: 3965
sprt @ 15+0.05 th 1 Redistribute more time throughout game
14-02-21 uri simplify_futility diff
LLR: 0.36 (-2.94,2.94) [-3.00,1.00]
Total: 25715 W: 3907 L: 3927 D: 17881
sprt @ 60+0.05 th 1 deleting condition from futility that I see no reason for it(it is a simplification so I use -4,0) decided to use only (-3,1) for long time control. I believe that it is an improvement but prefer not to test it in SPRT (0,6) because I am not sure if the improvement is big enough to pass and as a simplification it is better to commit it even if it is only a small improvement.
14-02-21 sg multicut diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 44420 W: 8192 L: 8182 D: 28046
sprt @ 15+0.05 th 1 Multicut take 1
14-02-21 rst multicut diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 1789 W: 288 L: 389 D: 1112
sprt @ 15+0.05 th 1 Real MultiCut. Take 1
14-02-21 rst multicut diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 12922 W: 2294 L: 2366 D: 8262
sprt @ 15+0.05 th 1 Real MultiCut. Take 2
14-02-21 jki unstop diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 10327 W: 1839 L: 1917 D: 6571
sprt @ 15+0.05 th 1 Legacy unstoppable_pawns_eval
14-02-21 uri simplify_futility diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 28199 W: 5146 L: 5178 D: 17875
sprt @ 15+0.05 th 1 another try and this time less aggressive reductions at high depths but still prune immediately when the evaluation is high because I think that it is a waste of time to search moves with very high score(usually they are pruned by null move pruning but it takes time). This time I will use normal SPRT
14-02-21 hwi 3fold_fix_hw diff
LLR: 1.57 (-2.94,2.94) [-4.00,0.00]
Total: 35155 W: 6431 L: 6469 D: 22255
sprt @ 15+0.05 th 1 Attempt at fixing 3 fold repetition problem.
14-02-21 rst multicut diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 5134 W: 903 L: 995 D: 3236
sprt @ 15+0.05 th 1 Mixture of ProbCut and MultiCut. Take 3