Stockfish Testing Queue

Finished - 49271 tests

15-11-18 sni isolated_h diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 60486 W: 11670 L: 11612 D: 37204
sprt @ 10+0.1 th 1 More penalty for isolated A and H pawns
15-11-18 jos dr1 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 7460 W: 948 L: 1042 D: 5470
sprt @ 5+0.1 th 7 Since the search tree becomes narrower towards the endgame, it seems more logical to increase the rootDepth for the helpers. Test with 8moves book.
15-11-18 sni bishops2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 8443 W: 1527 L: 1616 D: 5300
sprt @ 10+0.1 th 1 Take 5
15-11-18 mbo tm_resolve_faillow diff
LLR: 2.95 (-2.94,2.94) [0.00,5.00]
Total: 4651 W: 961 L: 814 D: 2876
sprt @ 10+0.1 th 1 Time management: Use the score from previous search to resolve fail lows.
15-11-18 sni bishops2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 9600 W: 1786 L: 1870 D: 5944
sprt @ 10+0.1 th 1 Take 4
15-11-18 Voy mob diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 3039 W: 527 L: 640 D: 1872
sprt @ 10+0.1 th 1 Mob based on pawns idea... (fixed bench)
15-11-18 sg king_shelter diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 18912 W: 3603 L: 3646 D: 11663
sprt @ 10+0.1 th 1 Penalty for possible king shelter pawn break
15-11-18 sni bishops2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 16238 W: 3107 L: 3161 D: 9970
sprt @ 10+0.1 th 1 Take 3
15-11-16 Fis thread_voting diff
LLR: 2.95 (-2.94,2.94) [0.00,5.00]
Total: 24992 W: 3720 L: 3511 D: 17761
sprt @ 20+0.2 th 7 All threads vote on best move to play. LTC
15-11-18 Voy lmr diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 15454 W: 2894 L: 2952 D: 9608
sprt @ 10+0.1 th 1 Take 2
15-11-18 Roc LostInSpace diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 10911 W: 2089 L: 2167 D: 6655
sprt @ 10+0.1 th 1 Include non-blocking pawns when evaluating space
15-11-17 roh helper-limit-max-depth diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 6617 W: 1001 L: 1098 D: 4518
sprt @ 10+0.1 th 7 Formula to limit helper thread maximum depth.
15-11-17 Roc G_Outpost diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 2683 W: 426 L: 540 D: 1717
sprt @ 10+0.1 th 1 Consider outpost also when shielded by a non-pawn
15-11-17 sg outposts diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 8257 W: 1582 L: 1672 D: 5003
sprt @ 10+0.1 th 1 Recognize pawns in front which constrain the attack span
15-11-17 Voy lmr diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 16845 W: 3250 L: 3302 D: 10293
sprt @ 10+0.1 th 1 lmr tweak
15-11-16 sni asymmetric_search2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 82694 W: 15793 L: 15554 D: 51347
sprt @ 10+0.1 th 1 Asymmetric risk management, take 2
15-11-16 sni bishops2 diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 67950 W: 13148 L: 12971 D: 41831
sprt @ 10+0.1 th 1 Opposite bishops scale factor by number of pawns, take 2
15-11-17 Voy VeryQuiet diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 8891 W: 1662 L: 1749 D: 5480
sprt @ 10+0.1 th 1 Extend a capture if 10 consecutive plies were quiet.
15-11-17 luc MoveEasiness diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 10668 W: 2011 L: 2090 D: 6567
sprt @ 10+0.1 th 1 Consecutive easy moves, handled within TM. Simplified and hopefully better tuned...
15-11-17 Roc LastStopper diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 4705 W: 878 L: 985 D: 2842
sprt @ 10+0.1 th 1 S(20, 20) when only one weak enemy pawn prevents one of our pawn to become a passer
15-11-16 Voy RFt4 diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 43828 W: 6387 L: 6411 D: 31030
sprt @ 60+0.4 th 1 LTC: Take 4.
15-11-16 sni asymmetric_search2 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 26603 W: 4987 L: 4997 D: 16619
sprt @ 10+0.1 th 1 Asymmetric risk management, take 3
15-11-16 luc tuneMoveEasiness diff
14868/15000 iterations
30000/30000 games played
30000 @ 20+0.2 th 1 As the queue is empty, let's try a more sensible tuning of a simplified version...
15-11-16 sni bishops2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 7626 W: 1417 L: 1510 D: 4699
sprt @ 10+0.1 th 1 Opposite bishops scale factor by number of pawns
15-11-16 jos stable_search diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 9155 W: 1710 L: 1796 D: 5649
sprt @ 10+0.1 th 1 A last desperate try, which will very probably fail. But it would help to reduce the number of fail-highs/fail-lows per iteration, which may also be part of the problem seen in game 22.
15-11-16 roh helper-limit-max-depth diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 6409 W: 974 L: 1071 D: 4364
sprt @ 10+0.1 th 7 2nd Attempt at converting earlier success at helper thread depth into a formula.
15-11-16 tvi HashMove diff
ELO: -3.77 +-4.0 (95%) LOS: 3.1%
Total: 10319 W: 1741 L: 1853 D: 6725
20000 @ 5+0.1 th 7 Race detection. Is this (near) ELO neutral with 7 threads? Prio -1
15-11-15 jos no_drawsore_to_tt diff
ELO: 1.32 +-4.1 (95%) LOS: 73.7%
Total: 10000 W: 1809 L: 1771 D: 6420
10000 @ 5+0.1 th 3 Don't save draw scores into the transposition table. A quick measurement with 3 threads.
15-11-09 Fis thread_voting diff
LLR: 2.95 (-2.94,2.94) [0.00,5.00]
Total: 77576 W: 12614 L: 12192 D: 52770
sprt @ 10+0.1 th 7 All threads vote on best move to play.
15-11-16 Voy RFt4 diff
LLR: 2.95 (-2.94,2.94) [0.00,4.00]
Total: 21483 W: 4100 L: 3868 D: 13515
sprt @ 10+0.1 th 1 Take 4.
15-11-15 lbr test diff
LLR: -2.96 (-2.94,2.94) [-3.00,1.00]
Total: 8071 W: 1037 L: 1197 D: 5837
sprt @ 20+0.2 th 7 LTC: fix bestThread
15-11-15 Voy RFt2 diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 33359 W: 4925 L: 4979 D: 23455
sprt @ 60+0.4 th 1 LTC: Reduction formula tweak...
15-11-15 sni asymmetric_search diff
ELO: 2.87 +-2.9 (95%) LOS: 97.2%
Total: 20000 W: 3815 L: 3650 D: 12535
20000 @ 10+0.1 th 1 Estimate elo value of square (0,0) of risk matrix
15-11-15 sni asymmetric_search diff
ELO: -2.14 +-2.9 (95%) LOS: 7.8%
Total: 20000 W: 3687 L: 3810 D: 12503
20000 @ 10+0.1 th 1 Estimate elo value of square (0,1) of risk matrix
15-11-16 luc OpponentTime diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 15720 W: 2964 L: 3021 D: 9735
sprt @ 10+0.1 th 1 Basic attempt to adjust our time usage depending on opponent's time (take 1)
15-11-15 sni asymmetric_search diff
ELO: -0.23 +-3.0 (95%) LOS: 44.1%
Total: 20000 W: 3815 L: 3828 D: 12357
20000 @ 10+0.1 th 1 Estimate elo value of square (1,1) of risk matrix
15-11-10 mco pull_clean_deepest diff
LLR: -2.96 (-2.94,2.94) [-5.00,0.00]
Total: 46936 W: 7320 L: 7623 D: 31993
sprt @ 10+0.1 th 7 Simple regression test to verify patch is correct: it is a no functional change, but has an impact only in SMP case that cannot be verified with bench, so better to test. TAKE 2 (added a speed optimization).
15-11-15 sni asymmetric_search diff
ELO: -2.31 +-3.0 (95%) LOS: 6.3%
Total: 20000 W: 3713 L: 3846 D: 12441
20000 @ 10+0.1 th 1 Estimate elo value of square (1,0) of risk matrix
15-11-13 jos stable_search diff
ELO: -0.80 +-2.6 (95%) LOS: 27.3%
Total: 20000 W: 2886 L: 2932 D: 14182
20000 @ 60+0.4 th 1 Measure the impact of a somewhat more stable search. Now at LTC. (The real benefit of this patch will most likely show up at even longer tc, and when analyzing.)
15-11-15 lbr tt64 diff
ELO: -4.02 +-2.9 (95%) LOS: 0.3%
Total: 16856 W: 2475 L: 2670 D: 11711
20000 @ 40+0.4 th 1 LTC (rescaled throughput): 64-bit keys, 4 entries per cluster. realistic hash pressure. how much elo do we lose ?
15-11-15 Voy RFt3 diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 17090 W: 3244 L: 3342 D: 10504
sprt @ 10+0.1 th 1 Take 3.
15-11-15 SC varianceRazor diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 10331 W: 1970 L: 2051 D: 6310
sprt @ 10+0.1 th 1 Variance in razoring. To make razor pruning more flat over variance, increase razor margin when variance is low.
15-11-15 lbr ttcopy diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 27593 W: 4851 L: 4860 D: 17882
sprt @ 5+0.05 th 7 load TTEntry one shot, to limit TT races. currently, we work with a pointer, and reload tte->widget() every time we need a widget(), all while the TT changes under our feet. a bit dangerous for my taste.
15-11-15 tvi HashMove diff
ELO: 0.24 +-4.2 (95%) LOS: 54.5%
Total: 10000 W: 1916 L: 1909 D: 6175
10000 @ 5+0.1 th 3 A variant on Hyatt's lockless hash, reordered fields, 3 thr 5+0.1/Hash=4
15-11-09 lbr test diff
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 114131 W: 18120 L: 18143 D: 77868
sprt @ 10+0.1 th 7 fix bestThread
15-11-15 lbr tt64 diff
ELO: -4.19 +-3.0 (95%) LOS: 0.3%
Total: 20000 W: 3696 L: 3937 D: 12367
20000 @ 10+0.1 th 1 64-bit keys, 4 entries per cluster. realistic hash pressure. how much elo do we lose ?
15-11-15 tvi HashMove diff
LLR: -0.32 (-2.94,2.94) [0.00,5.00]
Total: 607 W: 108 L: 119 D: 380
sprt @ 5+0.1 th 3 A variant on Hyatt's lockless hash, 3 thr 5+0.1/Hash=4
15-11-15 lbr tt64^ diff
ELO: -4.12 +-3.0 (95%) LOS: 0.3%
Total: 20000 W: 3711 L: 3948 D: 12341
20000 @ 10+0.1 th 1 64-bit keys, 2 entries per cluster. realistic hash pressure. how much elo do we lose ?
15-11-13 tvi HashMove diff
LLR: -2.96 (-2.94,2.94) [-3.00,1.00]
Total: 131306 W: 25201 L: 25596 D: 80509
sprt @ 5+0.05 th 3 Maintain old behaviour (any move is better than no move), but do not replace saved high depth moves with new low depth moves. Test with 3 threads, 5+0.05
15-11-15 Voy RFt2 diff
LLR: 2.96 (-2.94,2.94) [0.00,4.00]
Total: 43045 W: 8439 L: 8126 D: 26480
sprt @ 10+0.1 th 1 Reduction formula tweak...