Stockfish Testing Queue

Finished - 1408 tests

19-05-16 31m simplify_cpasser diff
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 14019 W: 3162 L: 3025 D: 7832
sprt @ 10+0.1 th 1 Just consider the blockSq (s + Up), not the entire forward_file_bb. Should be a much simpler expression.
19-05-16 31m passer_kingprox diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 7753 W: 1680 L: 1807 D: 4266
sprt @ 10+0.1 th 1 A shot in the dark: make the second push in passed pawn evaluation mirror the first, merely with half effect.
19-05-15 31m LongDiagonal_passer^ diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 22582 W: 4946 L: 4999 D: 12637
sprt @ 10+0.1 th 1 No friendly passed pawns in LongDiagonalBishop blockers.
19-05-15 31m LongDiagonal_passer diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 18272 W: 4075 L: 4149 D: 10048
sprt @ 10+0.1 th 1 No friendly, unblocked passed pawns in LongDiagonalBishop blockers.
19-05-15 31m combo_KFA_KDC diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 8099 W: 1777 L: 1910 D: 4412
sprt @ 10+0.1 th 1 Another attempt to combo recent promising kingDanger tweaks.
19-05-15 31m tweak_KDC2 diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 11598 W: 2605 L: 2724 D: 6269
sprt @ 10+0.1 th 1 Opposite effect.
19-05-15 31m tweak_KDC2 diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 58389 W: 13099 L: 13031 D: 32259
sprt @ 10+0.1 th 1 An experiment building on @praveentml's work.
19-05-15 31m passer_xray diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 13386 W: 2961 L: 3059 D: 7366
sprt @ 10+0.1 th 1 Much smaller effect (4).
19-05-15 31m passer_xray diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 6012 W: 1271 L: 1406 D: 3335
sprt @ 10+0.1 th 1 Inspired by a position from TCEC 15 Super Final Game 10. Include a case in passed pawn evaluation for x-ray defense of the passer: they have a R/Q attacker on the file, but so do we, and ours is R/Q-defended.
19-05-14 31m reorder_extlmr diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 38223 W: 8573 L: 8585 D: 21065
sprt @ 10+0.1 th 1 Code reordering: only apply @Vizvezdenec's PR #2146 if !captureOrPromotion. Tested on top of PR #2146.
19-05-14 31m lmr_singular diff
LLR: -0.92 (-2.94,2.94) [0.00,3.50]
Total: 16500 W: 2799 L: 2807 D: 10894
sprt @ 60+0.6 th 1 Speculative LTC for 84K yellow. Low throughput.
19-05-13 31m lmr_singular diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 84465 W: 19052 L: 18795 D: 46618
sprt @ 10+0.1 th 1 Only if !captureOrPromotion.
19-05-13 31m ThreatByRank_AB2_2 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 8390 W: 1806 L: 1929 D: 4655
sprt @ 10+0.1 th 1 Fixing an unfortunate bug (an omitted "& s") in my previous three tests...my apologies. Halve ThreatByRank if the enemy piece is doubly defended but not doubly attacked. The key implication is that at least one of the defenders is not a pawn, and therefore the non-pawn piece will simply be replaced by another in the event of a trade.
19-05-13 31m lmr_singular diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 19873 W: 4397 L: 4463 D: 11013
sprt @ 10+0.1 th 1 Only if capture/promotion.
19-05-13 31m ThreatByRank_AB2 diff
LLR: -0.56 (-2.94,2.94) [0.50,4.50]
Total: 17407 W: 3890 L: 3835 D: 9682
sprt @ 10+0.1 th 1 Minor case only.
19-05-13 31m ThreatByRank_AB2^ diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 7915 W: 1672 L: 1797 D: 4446
sprt @ 10+0.1 th 1 Rook case only.
19-05-13 31m ThreatByRank_AB2 diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 38570 W: 8793 L: 8764 D: 21013
sprt @ 10+0.1 th 1 Idea again based on Bryan's analysis of Game 3 of TCEC 15 Cup Final. Halve ThreatByRank if the enemy piece is doubly defended but not doubly attacked. The key implication is that at least one of the defenders is not a pawn, and therefore the non-pawn piece will simply be replaced by another in the event of a trade.
19-05-13 31m lmr_singular diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 15294 W: 3420 L: 3509 D: 8365
sprt @ 10+0.1 th 1 Variant of @locutus2's patch, reordered to not further decrease reduction 1 ply for singularExtension if we already have done so 2 plies for ttPv.
19-05-13 31m LongDiagonal_oppB diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 12043 W: 2651 L: 2756 D: 6636
sprt @ 10+0.1 th 1 This is functionally equivalent to the 54K yellow but should be substantially simpler/faster. Hopefully the 54K wasn't simply chance...
19-05-13 31m LongDiagonal_oppB diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 20690 W: 4685 L: 4746 D: 11259
sprt @ 10+0.1 th 1 Restore 54K yellow, but narrow the effect--exclude cases where the problematic enemy bishop is attacked by our pawns.
19-05-12 31m LongDiagonal_oppB diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 22261 W: 4993 L: 5047 D: 12221
sprt @ 10+0.1 th 1 Divisor 4.
19-05-12 31m LongDiagonal_oppB diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 54444 W: 12364 L: 12256 D: 29824
sprt @ 10+0.1 th 1 No LongDiagonalBishop if opposed along the long diagonal by the enemy bishop. Based on Bryan's analysis of Game 3 of TCEC 15 Cup Final.
19-05-12 31m weakBlockedCPP diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 20594 W: 4627 L: 4689 D: 11278
sprt @ 10+0.1 th 1 Less effect: bonus / 4
19-05-12 31m weakBlockedCPP diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 29407 W: 6618 L: 6636 D: 16153
sprt @ 10+0.1 th 1 No candidate passed pawn bonus if the pawn is weak (doubly attacked, but neither doubly-defended nor pawn-defended) and blocked. Based on Bryan's analysis of Game 3 of TCEC 15 Cup Final.
19-05-12 31m KD_pawnweak^ diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 30003 W: 6761 L: 6776 D: 16466
sprt @ 10+0.1 th 1 Based on Bryan's analysis of Game 3 of TCEC 15 Cup Final, moves 26-31. In particular, White has a seemingly undeserved kingDanger penalty. If the only weak, non-pawn-defended kingRing squares are merely attacked by the enemy king or pawns, don't count the weak squares at all.
19-05-12 31m KD_pawnweak diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 27698 W: 6249 L: 6275 D: 15174
sprt @ 10+0.1 th 1 Half effect.
19-05-12 31m combo_KFA_N_KDC diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 26217 W: 5883 L: 5943 D: 14391
sprt @ 10+0.1 th 1 @praveentml's cs1^ looks promising but appears to be struggling to pass LTC--perhaps these other tweaks will work well with it. Combo of myself, @praveentml, and @locutus2. (Revised description to accurately reflect patch)
19-05-12 31m LongDiagonal_oppB diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 11420 W: 2491 L: 2599 D: 6330
sprt @ 10+0.1 th 1 Half effect.
19-05-11 31m combo_KFA_CH diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 20664 W: 3435 L: 3523 D: 13706
sprt @ 60+0.6 th 1 LTC for combo.
19-05-11 31m combo_KFA_CH diff
LLR: 2.95 (-2.94,2.94) [0.00,4.00]
Total: 30645 W: 7079 L: 6790 D: 16776
sprt @ 10+0.1 th 1 Combo of the linear KFA tweak, quadratic KFA tweak, and captureHistory tweak. (fixed bench)
19-05-10 31m combo_KFA_N_CH diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 45304 W: 10292 L: 10275 D: 24737
sprt @ 10+0.1 th 1 Merge in @Vizvezdenec's captHistWTw2 (May 5, 97K STC yellow).
19-05-08 31m TBPP_king diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 24332 W: 5362 L: 5406 D: 13564
sprt @ 10+0.1 th 1 Opposite effect, canceling S(10, 10) of the ThreatByPawnPush bonus.
19-05-08 31m TBPP_king diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 10985 W: 2358 L: 2468 D: 6159
sprt @ 10+0.1 th 1 Since excluding the king is terrible, introduce a special case S(10, 10) bonus for it.
19-05-08 31m TBPP_king diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 7908 W: 1695 L: 1821 D: 4392
sprt @ 10+0.1 th 1 Don't include the enemy king in ThreatByPawnPush, since it would be better handled in king safety.
19-05-08 31m pawnPush_kingRing diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 21269 W: 4677 L: 4736 D: 11856
sprt @ 10+0.1 th 1 +25%
19-05-08 31m pawnPush_kingRing^ diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 19805 W: 4342 L: 4409 D: 11054
sprt @ 10+0.1 th 1 -25%
19-05-08 31m pawnPush_kingRing diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 4740 W: 958 L: 1099 D: 2683
sprt @ 10+0.1 th 1 Since that 7K red was unexpectedly terrible, try opposite effect: +50% rather than -50%.
19-05-08 31m pawnPush_kingRing diff
LLR: -2.94 (-2.94,2.94) [0.50,4.50]
Total: 7584 W: 1648 L: 1775 D: 4161
sprt @ 10+0.1 th 1 Halve ThreatByPawnPush when the pawn being pushed is currently in our kingRing. Often this push is not a real threat, since it is dangerous to our king. Inspired by Bryan's analysis of Pohl 60.061.
19-05-06 31m combo_KFA_N^ diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 52319 W: 8813 L: 8799 D: 34707
sprt @ 60+0.6 th 1 LTC: linear KFA and knight material tweaks.
19-05-06 31m combo_KFA_N diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 29360 W: 6480 L: 6528 D: 16352
sprt @ 10+0.1 th 1 Revert linear-term KFA tweak, restore KFA quadratic-term tweak. Keep knight material tweak.
19-05-06 31m combo_KFA_N diff
LLR: 2.95 (-2.94,2.94) [0.00,4.00]
Total: 29088 W: 6543 L: 6264 D: 16281
sprt @ 10+0.1 th 1 Merge new master and revert quadratic-term tweak, leaving the linear KFA and knight material tweaks.
19-05-06 31m pawnPush_hanging diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 7297 W: 1542 L: 1670 D: 4085
sprt @ 10+0.1 th 1 No ThreatByPawnPush when the pawn is the only defender of an otherwise-hanging piece, and the threat is against the enemy that will capture it. (Note that this enemy must be a rook or queen on the same file.)
19-05-06 31m pawnPush_hanging diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 11648 W: 2534 L: 2641 D: 6473
sprt @ 10+0.1 th 1 Based on Bryan's analysis of Pohl 60.061 (SF 0-1 Houdini). No ThreatByPawnPush if the pawn is the only defender of an otherwise-hanging piece.
19-05-05 31m combo_KFA_N diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 32191 W: 7198 L: 7234 D: 17759
sprt @ 10+0.1 th 1 Combo of @sg's knight tweaks with the recent promising KFA tweaks by myself/@SFisGOD/@Vizvezdenec (LTC 126K yellow). Submitted a bit more quickly than usual because of the upcoming TCEC Superfinal--please let me know if I should wait instead.
19-05-05 31m advance2 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 8677 W: 1826 L: 1947 D: 4904
sprt @ 10+0.1 th 1 Exclude edge files.
19-05-05 31m advance diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 12931 W: 2806 L: 2907 D: 7218
sprt @ 10+0.1 th 1 Half effect for the pawns not on the most advanced rank.
19-05-05 31m tweak_KFA2 diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 37565 W: 8246 L: 8262 D: 21057
sprt @ 10+0.1 th 1 Experiment with even larger decreases in the constant term. -25.
19-05-05 31m advance diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 8803 W: 1879 L: 2000 D: 4924
sprt @ 10+0.1 th 1 My interpretation of @snicolet's 47K STC green / 91K LTC yellow. One potential concern with giving a bonus for the number of pawns on our most-advanced-pawn's rank is the implicit penalization of pawn pushes. Suppose we have three pawns on RANK_6: their bonus is 3 * r^2 / 2 = 37.5 cp. But if one is pushed to RANK_7, the bonus is cut in half: r^2 / 2 = 18. Therefore, lessen this drop: give the larger bonus of the most-advanced pawn's rank and the rank behind it.
19-05-04 31m tweak_KFA2 diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 34031 W: 5770 L: 5814 D: 22447
sprt @ 60+0.6 th 1 LTC: Constant -20.
19-05-04 31m tweak_KFA2 diff
LLR: 2.95 (-2.94,2.94) [0.00,4.00]
Total: 29789 W: 6809 L: 6525 D: 16455
sprt @ 10+0.1 th 1 Constant -20.