Stockfish Testing Queue

Finished - 18520 tests

12-06-17 Sa FutilityTuning diff
9903/10000 iterations
20000/20000 games played
20000 @ 20+0.2 th 1 Seems like slope is still decreasing. Reduced throughput
13-06-17 sn undefended4 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 21503 W: 3915 L: 3948 D: 13640
sprt @ 10+0.1 th 1 Exchange less pieces when we have a king attack, take 3
13-06-17 pb pvs_aspiration diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 8886 W: 1580 L: 1667 D: 5639
sprt @ 10+0.1 th 1 Aspiration: immediately research with wider window if first rootmove fails low (see chessprogramming 'PVS and Aspiration', approach B).
13-06-17 pb bishop_knight_ex diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 16541 W: 3029 L: 3083 D: 10429
sprt @ 10+0.1 th 1 Take 3: On recapture, trade our bishop with knight (good capture), but not do it, if it's our last & opponent keeps the bishop pair.
11-06-17 Sa LMR2TuningModel diff
55157/60000 iterations
120000/120000 games played
120000 @ 5+0.05 th 1 Tune all parameters
13-06-17 pb bishop_knight_ex4 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 6465 W: 1149 L: 1247 D: 4069
sprt @ 10+0.1 th 1 Take 4: like take 3, but restrict it all to the recapture square.
12-06-17 El king_end diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 32284 W: 5842 L: 5830 D: 20612
sprt @ 10+0.1 th 1 Final try: (0,70) penalty
13-06-17 Vo qspt5 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 30260 W: 5450 L: 5447 D: 19363
sprt @ 10+0.1 th 1 last try...
12-06-17 Sa LMRKing diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 35392 W: 6428 L: 6402 D: 22562
sprt @ 10+0.1 th 1 Reduce certain king moves
12-06-17 sn futility_relax3 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 32934 W: 5898 L: 5884 D: 21152
sprt @ 10+0.1 th 1 Relax futility condition, take 5
12-06-17 Sa LMRCastling diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 31083 W: 5660 L: 5652 D: 19771
sprt @ 10+0.1 th 1 Treat castling like escaping capture
12-06-17 pb bishop_knight_ex diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 20244 W: 2616 L: 2668 D: 14960
sprt @ 60+0.6 th 1 LTC: Take 2: Handle Bishop Knight trade as good-capture in dynamic situations (on recapture)
12-06-17 sg fix_pet21 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 28850 W: 5258 L: 5260 D: 18332
sprt @ 10+0.1 th 1 Increase moveCount in QS only for non-captures. Take 4
12-06-17 Vo ape diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 20784 W: 3765 L: 3801 D: 13218
sprt @ 10+0.1 th 1 advance pawn extension
12-06-17 sg fix_pet21 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 23841 W: 4350 L: 4373 D: 15118
sprt @ 10+0.1 th 1 Fix the issue with PET#21 (see forum) by allowing evasion pruning only for material more than a bishop
12-06-17 Vo qsptB diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 13277 W: 2374 L: 2443 D: 8460
sprt @ 10+0.1 th 1 Fixed Buggy Patch
12-06-17 Sa ReduceDepthForRetreat diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 29677 W: 5450 L: 5448 D: 18779
sprt @ 10+0.1 th 1 Reduce retreats at cut nodes.
12-06-17 Sa LMRIsSingularReduction diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 10307 W: 1825 L: 1906 D: 6576
sprt @ 10+0.1 th 1 Reduce cut nodes when the TT move is singular
12-06-17 II nullmove diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 6102 W: 1066 L: 1165 D: 3871
sprt @ 10+0.1 th 1 Take 2 with new structure.
12-06-17 II nullmove diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 4246 W: 729 L: 837 D: 2680
sprt @ 10+0.1 th 1 Log null move - few more tries with new structure
12-06-17 Sa LMRCastling diff
LLR: 0.05 (-2.94,2.94) [0.00,5.00]
Total: 1140 W: 208 L: 201 D: 731
sprt @ 10+0.1 th 1 Treat castling like escaping capture.
12-06-17 Vo qsptB diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 1907 W: 290 L: 407 D: 1210
sprt @ 10+0.1 th 1 Another idea...
12-06-17 Sa LMRIsSingularReduction diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 14806 W: 2731 L: 2793 D: 9282
sprt @ 10+0.1 th 1 In non-cutNode, reduce by one ply if the TT move is singular.
12-06-17 Vo qspt4 diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 8185 W: 1385 L: 1475 D: 5325
sprt @ 10+0.1 th 1 Take 4
12-06-17 pb bishop_knight_ex diff
LLR: 2.96 (-2.94,2.94) [0.00,5.00]
Total: 23347 W: 4369 L: 4143 D: 14835
sprt @ 10+0.1 th 1 Take 2: Handle Bishop Knight trade as good-capture in dynamic situations (on recapture)
12-06-17 Vo qsfpB diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 6214 W: 1056 L: 1155 D: 4003
sprt @ 10+0.1 th 1 Tryout another idea...
12-06-17 sn futility_relax3 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 18145 W: 2255 L: 2315 D: 13575
sprt @ 60+0.6 th 1 LTC: Relax futility condition, take 4
12-06-17 sg fix_pet21 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 5449 W: 930 L: 1032 D: 3487
sprt @ 10+0.1 th 1 Now check material condition only for moveCount = 2. Take 3. Set other test to prio -1.
12-06-17 sg fix_pet21 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 15707 W: 2821 L: 2879 D: 10007
sprt @ 10+0.1 th 1 Now check material condition only for moveCount > 2. Take 2
12-06-17 Vo qspt3 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 29250 W: 5323 L: 5323 D: 18604
sprt @ 10+0.1 th 1 take 3
12-06-17 pb bishop_knight_ex diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 18785 W: 3409 L: 3454 D: 11922
sprt @ 10+0.1 th 1 Sort Bishop-Knight exchange among quiet moves (with zero value)
12-06-17 Sa ReduceDepthForRetreat diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 5155 W: 858 L: 961 D: 3336
sprt @ 10+0.1 th 1 Reduce retreats only in AllNode when it is not check, not in check, not escape capture.
12-06-17 Vo qspt2 diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 68505 W: 12515 L: 12348 D: 43642
sprt @ 10+0.1 th 1 take 2
11-06-17 jh unsafe_threats diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 6977 W: 1205 L: 1301 D: 4471
sprt @ 10+0.1 th 1 More accurate unsafe pawn threats
11-06-17 Vo qspt diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 63145 W: 11596 L: 11451 D: 40098
sprt @ 10+0.1 th 1 stc
11-06-17 fa kpsqt diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 54437 W: 10042 L: 10011 D: 34384
sprt @ 10+0.1 th 1 king psqt local tune.
11-06-17 Vo sct2 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 17169 W: 3124 L: 3176 D: 10869
sprt @ 10+0.1 th 1 stc
11-06-17 sn futility_relax3 diff
LLR: 2.95 (-2.94,2.94) [0.00,5.00]
Total: 22812 W: 4219 L: 3997 D: 14596
sprt @ 10+0.1 th 1 Relax futility condition, take 4
11-06-17 Vo statAvg diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 26199 W: 4835 L: 4847 D: 16517
sprt @ 10+0.1 th 1 stc
11-06-17 sg passed_pawn_state diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 19128 W: 3531 L: 3574 D: 12023
sprt @ 10+0.1 th 1 I think more tuning is needed but first test the tuned parameters
10-06-17 Sa SimplifyDepthInProbCutC diff
LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 59350 W: 10793 L: 10738 D: 37819
sprt @ 10+0.1 th 1 Depth calculation inside loop is executed less often than outside.
11-06-17 sn EvalInfo diff
LLR: 2.94 (-2.94,2.94) [-3.00,1.00]
Total: 25286 W: 4689 L: 4576 D: 16021
sprt @ 10+0.1 th 1 Vérification for PR #1135
11-06-17 sn safe_pawn_threats2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 11939 W: 2133 L: 2207 D: 7599
sprt @ 10+0.1 th 1 More precise ThreatBySafePawn, take 4
11-06-17 El king_end diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 28565 W: 5267 L: 5269 D: 18029
sprt @ 10+0.1 th 1 Use penalty (0,100)
10-06-17 Sa FutilityTuning diff
9920/10000 iterations
20000/20000 games played
20000 @ 20+0.2 th 1 Continuing tuning because the original tuning was at 5+0.5 and because it did not seem to reach stability.
10-06-17 Vo sct diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 35006 W: 6251 L: 6228 D: 22527
sprt @ 10+0.1 th 1 stat comparison tweak
10-06-17 Sa FutilityMargin diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 36147 W: 4773 L: 4772 D: 26602
sprt @ 60+0.6 th 1 LTC for tuned futility margins
10-06-17 Ro XRayDefense diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 24281 W: 4416 L: 4437 D: 15428
sprt @ 10+0.1 th 1 Negate some queen attacks
10-06-17 Vo lmrRoot diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 12846 W: 2269 L: 2340 D: 8237
sprt @ 10+0.1 th 1 stc
09-06-17 sg tune_passed_pawn_state diff
29241/30000 iterations
60000/68000 games played
68000 @ 10+0.1 th 1 Generalize the different bonus by defended and unsafe squares to an bonus array and tune it. Additionally we can use now different values for middle and endgame. The starting values are calculated to natch functional the master.