Stockfish Testing Queue

Pending - 0 tests 0.0 hrs

None

Active - 0 tests

Finished - 1124 tests

18-09-18 pb0 newPvResearch diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 77406 W: 16643 L: 16400 D: 44363
sprt @ 10+0.1 th 1 Experiment: research subtree after a high fail with low seldepth
18-09-18 pb0 fixFHpv2 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 967 W: 160 L: 288 D: 519
sprt @ 10+0.1 th 1 Store pv-move in TT also when value == alpha but not upd-stats test against passed #1771
18-09-18 pb0 fixFHpv2 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 2189 W: 421 L: 542 D: 1226
sprt @ 10+0.1 th 1 Store pv-move in TT also when value == alpha test against passed #1771
18-09-17 pb0 legalEvasions diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 12197 W: 2628 L: 2699 D: 6870
sprt @ 10+0.1 th 1 Avoid some further illegal evasion move generation by including sliders adjacent pseudo-attacks (they are real) in the area (bitboard) the king cannot access.
18-09-17 pb0 noCrossNmp diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 10295 W: 2131 L: 2212 D: 5952
sprt @ 10+0.1 th 1 Are crossing null moves useful?
18-09-14 pb0 fixFHpv2 diff
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 202411 W: 32734 L: 32897 D: 136780
sprt @ 60+0.6 th 1 Fixes truncated pv after fail high, version 2 (simplier) Check for no regression. no functional change
18-09-14 pb0 fixFHpv diff
LLR: -0.45 (-2.94,2.94) [-3.00,1.00]
Total: 40870 W: 6672 L: 6760 D: 27438
sprt @ 60+0.6 th 1 LTC: Fixes pv after fail high. Currently after a FH the reported pv is truncated. Length of pv should correspond to reported selDepth. Check for no regression. no functional change
18-09-14 pb0 assignSSply diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 73967 W: 15949 L: 15825 D: 42193
sprt @ 10+0.1 th 1 Assign ss->ply once instead of doing it in each node. Is this faster? no functional change
18-09-14 pb0 fixFHpv2 diff
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 16300 W: 3556 L: 3424 D: 9320
sprt @ 10+0.1 th 1 Fixes truncated pv after fail high, version 2 (simplier) Check for no regression. no functional change
18-09-14 pb0 fixFHpv diff
LLR: 3.06 (-2.94,2.94) [-3.00,1.00]
Total: 45046 W: 9622 L: 9540 D: 25884
sprt @ 10+0.1 th 1 Fixes pv after fail high. Currently after a FH the reported pv is truncated. Length of pv should correspond to reported selDepth. Check for no regression. no functional change
18-09-11 pb0 conseqFailHigh diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 78221 W: 12795 L: 12611 D: 52815
sprt @ 60+0.6 th 1 Speculative LTC, as this may kick in more often at higher tc's. At very high rootDepth, divide it by 2 after 2 consecutive fail high's. Finds mate in 27 on 1k3b1q/pP2p1p1/P1K1P1Pp/7P/2B5/8/8/8 w - - 0 1 in 1 minute 15 second, whereas master seems to take forever to find mate.
18-09-13 pb0 adjustedFH diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 63862 W: 10309 L: 10184 D: 43369
sprt @ 60+0.6 th 1 LTC More prudent approach Finds M27 on FEN 1k3b1q/pP2p1p1/P1K1P1Pp/7P/2B5/8/8/8 w - - 0 1 in 1 in less than one minute.
18-09-13 pb0 adjustedFH diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 1289 W: 224 L: 349 D: 716
sprt @ 10+0.1 th 1 Another try fixing also the multiPV issue Finds M27 on FEN 1k3b1q/pP2p1p1/P1K1P1Pp/7P/2B5/8/8/8 w - - 0 1 in 30 s
18-09-13 pb0 pinnedShelter diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 19566 W: 4200 L: 4236 D: 11130
sprt @ 10+0.1 th 1 Idea by Brian: "shelter strength" can be reduced if one of the king shelter pawns is pinned
18-09-13 pb0 adjustedFH diff
LLR: 2.95 (-2.94,2.94) [0.00,5.00]
Total: 58586 W: 12839 L: 12423 D: 33324
sprt @ 10+0.1 th 1 More prudent approach Finds M27 on FEN 1k3b1q/pP2p1p1/P1K1P1Pp/7P/2B5/8/8/8 w - - 0 1 in 1 in less than one minute.
18-09-12 pb0 adjustedFH diff
LLR: 2.94 (-2.94,2.94) [0.00,5.00]
Total: 29990 W: 6540 L: 6265 D: 17185
sprt @ 10+0.1 th 1 Take 2, merging both approaches
18-09-12 pb0 conseqFailHigh diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 4988 W: 1021 L: 1128 D: 2839
sprt @ 10+0.1 th 1 Take 3, similar approach to Fisherman ones
18-09-12 pb0 failHighResume diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 15662 W: 3289 L: 3344 D: 9029
sprt @ 10+0.1 th 1 Take 4
18-09-12 pb0 failHighResume diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 36583 W: 7837 L: 7791 D: 20955
sprt @ 10+0.1 th 1 Take 2 (normal against master)
18-09-12 pb0 failHighResume diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 9509 W: 1972 L: 2057 D: 5480
sprt @ 10+0.1 th 1 Take 3
18-09-12 pb0 failHighResume diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 9934 W: 2081 L: 2206 D: 5647
sprt @ 10+0.1 th 1 After fail high in aspiration, resume rootDepth from selDepth. Test against passed patch of mstembera with [0,4] bounds.
18-09-11 pb0 conseqFailHigh diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 43269 W: 9323 L: 9244 D: 24702
sprt @ 10+0.1 th 1 At very high rootDepth, divide it by 2 after 2 consecutive fail high's. Finds mate in 27 on 1k3b1q/pP2p1p1/P1K1P1Pp/7P/2B5/8/8/8 w - - 0 1 in 1 minute 15 second, whereas master seems to take forever to find mate.
18-09-11 pb0 conseqFailHigh diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 23246 W: 4952 L: 4970 D: 13324
sprt @ 10+0.1 th 1 Take 2
18-09-07 pb0 verify5 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 25538 W: 5160 L: 5170 D: 15208
sprt @ 15+0.15 th 1 Avoid some null-move verification (zugzwang unlikely), take 3 Little more tc as usual to dig in in verification more often.
18-09-07 pb0 verify5 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 26678 W: 5754 L: 5755 D: 15169
sprt @ 10+0.1 th 1 Avoid some null-move verification if we had plenty moves (zugzwang unlikely), take 2
18-09-07 pb0 verify5 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 16016 W: 3395 L: 3448 D: 9173
sprt @ 10+0.1 th 1 Avoid null-move verification if we had plenty moves (zugzwank unlikely)
18-09-06 pb0 noVerify diff
ELO: 0.26 +-1.9 (95%) LOS: 60.5%
Total: 40000 W: 6416 L: 6386 D: 27198
40000 @ 60+0.6 th 1 LTC: What's the tradeoff of nmp verification search nowaydays? LTC with low througput and prio -1
18-09-06 pb0 probcut22 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 19431 W: 4131 L: 4168 D: 11132
sprt @ 10+0.1 th 1 My try on ianfabs promising attempt (framework almost empty)
18-09-06 pb0 exp3 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 16233 W: 3410 L: 3462 D: 9361
sprt @ 10+0.1 th 1 Experiment
18-09-06 pb0 updkillers diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 15678 W: 3366 L: 3421 D: 8891
sprt @ 10+0.1 th 1 Don't update killers with bestmove when excludedMove is current killer
18-09-06 pb0 realDepth^ diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 20911 W: 4522 L: 4551 D: 11838
sprt @ 10+0.1 th 1 Use higher depth for singular ttMove bestmoves Take 2 (depth + ONE_PLY just for stat-bonus)
18-09-06 pb0 updkillers diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 11690 W: 2467 L: 2541 D: 6682
sprt @ 10+0.1 th 1 Don't update killers with bestmove when excludedMove is set
18-09-06 pb0 exclude_update diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 10376 W: 2199 L: 2279 D: 5898
sprt @ 10+0.1 th 1 Don't update move stats when excluded move is set, take 2
18-09-06 pb0 exclude_update diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 12765 W: 2658 L: 2727 D: 7380
sprt @ 10+0.1 th 1 Don't update move stats when excluded move is set
18-09-06 pb0 realDepth diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 3516 W: 680 L: 793 D: 2043
sprt @ 10+0.1 th 1 Use higher depth for singular ttMove bestmoves (depth + ONE_PLY for TT and stat-bonus)
18-09-06 pb0 realDepth diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 429 W: 39 L: 166 D: 224
sprt @ 10+0.1 th 1 Use higher depth for singular ttMove bestmoves Take 3 (depth + ONE_PLY just for TT-save)
18-09-05 pb0 passedTTmove diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 9913 W: 2081 L: 2164 D: 5668
sprt @ 10+0.1 th 1 Tweak moveCountPruning according to position of ttmove in scoring take 2
18-09-05 pb0 passedTTmove diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 8974 W: 1894 L: 1981 D: 5099
sprt @ 10+0.1 th 1 Tweak moveCountPruning according to position of ttmove in scoring fix uninitialized variable
18-09-05 pb0 singular7 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 4058 W: 803 L: 914 D: 2341
sprt @ 10+0.1 th 1 Mark singular extended nodes in TT and then re-search more shallow
18-09-05 pb0 singular7 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 3384 W: 669 L: 783 D: 1932
sprt @ 10+0.1 th 1 Mark singlular extend. nodes in TT to avoid some re-search (take 3)
18-09-05 pb0 singular7 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 4349 W: 869 L: 978 D: 2502
sprt @ 10+0.1 th 1 Mark singlular extend. nodes in TT to avoid some re-search (fixed logic)
18-09-05 pb0 singular7 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 13666 W: 2914 L: 2978 D: 7774
sprt @ 10+0.1 th 1 Mark singlular extended nodes in TT to avoid some singular re-search
18-09-04 pb0 futilityMCtweak diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 21889 W: 4675 L: 4700 D: 12514
sprt @ 10+0.1 th 1 Futility movecount formula tweak
18-09-04 pb0 futilityMCtweak diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 27642 W: 5857 L: 5913 D: 15872
sprt @ 10+0.1 th 1 Futility movecount formula tweak
18-09-04 pb0 retireCheckExtension diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 8333 W: 1722 L: 1812 D: 4799
sprt @ 10+0.1 th 1 Take 3
18-09-04 pb0 resetTTMove2 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 5409 W: 1115 L: 1220 D: 3074
sprt @ 10+0.1 th 1 Reset obsolete TT-Moves (tt-moves with stat-score < -14000)
18-09-04 pb0 resetTTMove2 diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 2835 W: 540 L: 656 D: 1639
sprt @ 10+0.1 th 1 Reset obsolete TT-Moves (tt-moves with negative stat-scores)
18-09-04 pb0 retireCheckExtension diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 16466 W: 3415 L: 3466 D: 9585
sprt @ 10+0.1 th 1 Retire check extension al low depths when there are 6 or more evasions
18-09-04 pb0 retireCheckExtension diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 10340 W: 2199 L: 2280 D: 5861
sprt @ 10+0.1 th 1 Retire check extension at low depths when there are 4 or more evasions
18-09-03 pb0 aligned diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 55596 W: 11681 L: 11631 D: 32284
sprt @ 10+0.1 th 1 Could this be faster? (non-functional change)