Stockfish Testing Queue

Finished - 4333 tests

19-07-03 sg lmr_null diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 7609 W: 1666 L: 1794 D: 4149
sprt @ 10+0.1 th 1 No LMR for capture/promotion if null move search.
19-07-03 sg probcut_checks2 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 8953 W: 1940 L: 2060 D: 4953
sprt @ 10+0.1 th 1 Don't prune checks in ProbCut move generation.
19-07-03 sg probcut_checks diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 4200 W: 854 L: 998 D: 2348
sprt @ 10+0.1 th 1 Allow quiet checks in ProbCut.
19-07-03 sg lmr_probcut diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 30162 W: 6598 L: 6614 D: 16950
sprt @ 10+0.1 th 1 Less reduction in probCut search. (Park with prio -1)
19-07-03 sg lmr_null diff
LLR: -2.94 (-2.94,2.94) [0.50,4.50]
Total: 26401 W: 5862 L: 5895 D: 14644
sprt @ 10+0.1 th 1 Use in null move search decreased depth for reduction calculation.
19-07-03 sg reduction_formula diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 8344 W: 1804 L: 1928 D: 4612
sprt @ 10+0.1 th 1 Increase move count by 1/8 in reduction formula (cap to MAXMOVE-1). (Park with prio -1)
19-07-03 sg lmr_probcut diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 16822 W: 3755 L: 3836 D: 9231
sprt @ 10+0.1 th 1 Less reduction for captures/promotions in probCut search. (park with prio -1)
19-07-03 sg lmr_probcut diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 15712 W: 3503 L: 3590 D: 8619
sprt @ 10+0.1 th 1 Use in probcut search decreased depth for reduction calculation.
19-07-03 sg lmr_excluded diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 15544 W: 3467 L: 3554 D: 8523
sprt @ 10+0.1 th 1 Inspired by Elberto's tests use singular search decreased move count for reduction claculation.
19-07-03 sg lmr_probcut^ diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 13859 W: 2998 L: 3094 D: 7767
sprt @ 10+0.1 th 1 Use in probcut search decreased move count for reduction calculation.
19-07-03 sg lmr_null^ diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 10952 W: 2403 L: 2514 D: 6035
sprt @ 10+0.1 th 1 Use in null move search decreased move count for reduction calculation.
19-07-03 sg lmr_excluded diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 10391 W: 2274 L: 2387 D: 5730
sprt @ 10+0.1 th 1 Use in singular search decreased depth for reduction calculation.
19-07-01 sg prune_cmh diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 21048 W: 4558 L: 4619 D: 11871
sprt @ 10+0.1 th 1 Exclude moves with main history >= 9300 from cmh based pruning.
19-07-01 sg prune_cmh diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 40298 W: 8961 L: 8926 D: 22411
sprt @ 10+0.1 th 1 Exclude moves with main history >= 10000 from cmh based pruning.
19-07-01 sg prune_cmh diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 23477 W: 5218 L: 5266 D: 12993
sprt @ 10+0.1 th 1 Exclude moves with cont[3] history >= 24450 from cmh based pruning.
19-07-01 sg prune_cmh diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 21766 W: 4797 L: 4854 D: 12115
sprt @ 10+0.1 th 1 Exclude counter moves with main history >= 9300 from cmh pruning (This doubles the effect).
19-07-01 sg prune_cmh diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 11243 W: 2467 L: 2576 D: 6200
sprt @ 10+0.1 th 1 Exclude counter moves with main history >= 10000 from cmh based pruning.
19-06-28 sg prune_passer diff
LLR: -2.94 (-2.94,2.94) [0.50,4.50]
Total: 13554 W: 2952 L: 3049 D: 7553
sprt @ 10+0.1 th 1 Don't prune captures of a pawn if a new passed pawn is created. Exclude queen and rook captures.
19-06-28 sg prune_passer diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 12587 W: 2773 L: 2875 D: 6939
sprt @ 10+0.1 th 1 Don't prune captures of a pawn if a new advanced passed pawn is created..
19-06-28 sg prune_passer diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 8570 W: 1836 L: 1958 D: 4776
sprt @ 10+0.1 th 1 Don't prune captures of a pawn if a new passed pawn is created. Exclude queen captures.
19-06-28 sg prune_passer diff
LLR: -2.94 (-2.94,2.94) [0.50,4.50]
Total: 5488 W: 1147 L: 1284 D: 3057
sprt @ 10+0.1 th 1 Don't prune captures of a pawn if a new passed pawn is created..
19-06-28 sg twoMoveHistory diff
LLR: -2.94 (-2.94,2.94) [0.50,4.50]
Total: 8698 W: 1855 L: 1976 D: 4867
sprt @ 10+0.1 th 1 Use twoMoveHistory for quiet move ordering but update only at PV nodes.
19-06-28 sg twoMoveHistory3 diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 8895 W: 1937 L: 2058 D: 4900
sprt @ 10+0.1 th 1 Use two last own moves as index for twoMoveHistory. Use it for quiet move sorting.
19-06-28 sg twoMoveHistory2 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 5612 W: 1180 L: 1317 D: 3115
sprt @ 10+0.1 th 1 Add compensation to stat score which retains average reduction.
19-06-28 sg twoMoveHistory2 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 5156 W: 1085 L: 1224 D: 2847
sprt @ 10+0.1 th 1 Add compensation to stat score.
19-06-28 sg twoMoveHistory2 diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 4044 W: 830 L: 975 D: 2239
sprt @ 10+0.1 th 1 Use half twoMoveHistory in stat score with compensation.
19-06-28 sg twoMoveHistory2 diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 7544 W: 1640 L: 1768 D: 4136
sprt @ 10+0.1 th 1 Use twoMoveHistory for quiet stat scores in LMR.
19-06-28 sg twoMoveHistory diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 9500 W: 2073 L: 2191 D: 5236
sprt @ 10+0.1 th 1 Use quarter twoMoveHistory for quiet ordering.
19-06-28 sg twoMoveHistory diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 7305 W: 1572 L: 1701 D: 4032
sprt @ 10+0.1 th 1 Use half twoMoveHistory for quiet ordering.
19-06-28 sg twoMoveHistory diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 4886 W: 1039 L: 1180 D: 2667
sprt @ 10+0.1 th 1 Double size of twoMoveHistory table.
19-06-28 sg twoMoveHistory diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 7714 W: 1685 L: 1812 D: 4217
sprt @ 10+0.1 th 1 Use twoMoveHistory for quiet move ordering (table has same size as contHistory).
19-06-27 sg spaceScale2 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 48257 W: 10750 L: 10675 D: 26832
sprt @ 10+0.1 th 1 Like my struggeling attempt (low + rook = high) but center thresholds around the old value.
19-06-27 sg spaceScale diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 38601 W: 8648 L: 8620 D: 21333
sprt @ 10+0.1 th 1 Try again the first tested thresholds but use faster scaling.
19-06-27 sg spaceScale2 diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 16691 W: 3676 L: 3758 D: 9257
sprt @ 10+0.1 th 1 Decrease low: low + 2 rooks = high threshold.
19-06-27 sg spaceScale2 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 18373 W: 4137 L: 4210 D: 10026
sprt @ 10+0.1 th 1 low + bishop = high threshold.
19-06-27 sg spaceScale diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 10739 W: 2423 L: 2535 D: 5781
sprt @ 10+0.1 th 1 Scale faster down.
19-06-27 sg spaceScale diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 14170 W: 3151 L: 3245 D: 7774
sprt @ 10+0.1 th 1 Try it in the other direction and use as low threshold master value and add one rook to high threshold.
19-06-27 sg spaceWeighting diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 11044 W: 2408 L: 2518 D: 6118
sprt @ 10+0.1 th 1 More space bonus for 4th rank but compensate (divisor=19).
19-06-26 sg passed_pawn_formation2 diff
LLR: -2.60 (-2.94,2.94) [0.00,3.50]
Total: 78139 W: 13184 L: 13119 D: 51836
sprt @ 60+0.6 th 1 I have tried several variants of the idea to detect pawn formations which could probably create a passed pawn later. Because this is a long-term strategix conecpt i think this should do even better at higher depths. So try a speculative LTC of my best yellow attempt (76K) (Low TP). Consider only pawn formation with a majority from at least two pawns.
19-06-27 sg spaceScale diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 11261 W: 2397 L: 2506 D: 6358
sprt @ 10+0.1 th 1 Customize by using a low and high space theshold. material < low threshold use zero, material < high theshold scale down. Here highh threshold is the same as in master and low theshold is one rook lower.
19-06-27 sg spaceScale diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 20368 W: 4518 L: 4581 D: 11269
sprt @ 10+0.1 th 1 For material < SpaceThreshold/2 use zero, for < SpaceThreshold scale down.
19-06-27 sg spaceScale diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 14538 W: 3116 L: 3209 D: 8213
sprt @ 10+0.1 th 1 If material below SpaceThreshold scale score down instead of using zero.
19-06-26 sg passer2v1 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 47146 W: 10616 L: 10545 D: 25985
sprt @ 10+0.1 th 1 Detect two vs one pawn situations which can creates a candidate passer which has only one stopper on an adjacent file and a undisturbed supporter on the same file.
19-06-26 sg spaceTweak diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 12539 W: 2779 L: 2894 D: 6866
sprt @ 10+0.1 th 1 Extend space area to 5th rank and compensate by decreasing divisor to 19. Test against Vizs passed spaceAttPR patch.
19-06-26 sg spaceTweak diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 11072 W: 2460 L: 2581 D: 6031
sprt @ 10+0.1 th 1 Extend space area to 5th rank. Test against Vizs passed spaceAttPR patch.
19-06-26 sg spaceAdvance diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 12473 W: 2737 L: 2840 D: 6896
sprt @ 10+0.1 th 1 Consider for space also safe pawn pushes. Test against Vizs passed spaceAttPR patch.
19-06-26 sg spaceAttThreat diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 9208 W: 2001 L: 2120 D: 5087
sprt @ 10+0.1 th 1 Give small penalty if space square could be attacked in one move. Test against Vizs passed spaceAttPR patch.
19-06-26 sg spaceAttThreat diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 6941 W: 1489 L: 1620 D: 3832
sprt @ 10+0.1 th 1 Lower penalty (1,0). Test against Vizs passed spaceAttrPR patch.
19-06-26 sg passed_pawn_formation2 diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 37149 W: 8218 L: 8199 D: 20732
sprt @ 10+0.1 th 1 Variant of my 76K yellow test: Consider only pawn formation with a majority from at least two pawns but also exclude doubled pawns.
19-06-26 sg passed_pawn_formation3 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 29363 W: 6473 L: 6492 D: 16398
sprt @ 10+0.1 th 1 Consider only pawn formation with a majority of at least two pawns but only if no passed pawns exists.