Stockfish Testing Queue

Finished - 31762 tests

15-10-24 tvi NearDraw diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 37576 W: 5580 L: 5564 D: 26432
sprt @ 60+0.4 th 1 Penalty/Bonus for approaching draw, triple bonus. LTC
15-10-24 sg passed_pawns diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 151878 W: 29505 L: 29119 D: 93254
sprt @ 10+0.1 th 1 Consider also own pieces blocking queening path
15-10-25 bin NearDraw diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 28579 W: 5751 L: 5747 D: 17081
sprt @ 10+0.1 th 1 No idea of what im doing, but it seems to work. Moved function into qsearch. Now eval gradually decreases with depth in fortress. Side effect bench much higher + Eliminated depth explosion in drawn endgames
15-10-24 Voy PlyToHistory2 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 29233 W: 5588 L: 5585 D: 18060
sprt @ 10+0.1 th 1 Take 2.
15-10-25 IIv lazy_log_formula diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 20764 W: 3608 L: 3694 D: 13462
sprt @ 10+0.1 th 3 We have never checked whether we can use more additionalDepth on 3 threads. So, checking (0,3,5) combination.
15-10-24 jki timeman_bugfix1 diff
LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 17200 W: 3367 L: 3239 D: 10594
sprt @ 10+0.1 th 1 Fix logical bugs in MinThinkingTime handling
15-10-24 IIv lazy_smp diff
ELO: -1.39 +-16.1 (95%) LOS: 43.3%
Total: 500 W: 69 L: 71 D: 360
500 @ 10+0.1 th 23 A quick check can we do better!
15-10-24 luc NearDraw diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 17820 W: 3514 L: 3561 D: 10745
sprt @ 10+0.1 th 1 Scale eval linearly with 50-move counter (never over 40%), very basic formula. Based on Tom Vijlbrief's patch.
15-10-24 Fis NearDrawInterpolate diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 11292 W: 2129 L: 2206 D: 6957
sprt @ 10+0.1 th 1 Interpolate eval towards a draw as we approach it. Original by Tom Vijlbrief.
15-10-24 jhe log2_depth diff
ELO: 5.73 +-8.5 (95%) LOS: 90.7%
Total: 2000 W: 328 L: 295 D: 1377
2000 @ 10+0.1 th 23 Correct Depth Rounding.
15-10-24 jhe max_time diff
ELO: -8.60 +-4.6 (95%) LOS: 0.0%
Total: 8485 W: 1509 L: 1719 D: 5257
10000 @ 10+0.1 th 1 Scale MaxRatio to compensate for lack of increment.
15-10-24 jhe max_time diff
ELO: -6.74 +-4.2 (95%) LOS: 0.1%
Total: 10000 W: 1846 L: 2040 D: 6114
10000 @ 10+0.1 th 1 Don't include increment in Max Time calculation.
15-10-24 Voy PlyToHistory2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 29435 W: 4240 L: 4255 D: 20940
sprt @ 60+0.4 th 1 LTC. Version 2
15-10-21 sni pawn_span2 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 28249 W: 5370 L: 5372 D: 17507
sprt @ 10+0.1 th 1 Take 2
15-10-23 Voy HPt4 diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 79304 W: 15304 L: 15179 D: 48821
sprt @ 10+0.1 th 1 History Pruning Tweak 4
15-10-24 sni bishop5 diff
LLR: -2.95 (-2.94,2.94) [-3.00,1.00]
Total: 7632 W: 1376 L: 1548 D: 4708
sprt @ 10+0.1 th 1 Simplify opposite bishop evaluation
15-10-24 sni initiative_tweak2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 9532 W: 1886 L: 1970 D: 5676
sprt @ 10+0.1 th 1 Increase (mg,eg) gradient
15-10-23 tvi NearDraw diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 47424 W: 9111 L: 9027 D: 29286
sprt @ 10+0.1 th 1 Penalty/Bonus for approaching draw, double bonus
15-10-24 sg dynamicHP diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 7789 W: 1493 L: 1585 D: 4711
sprt @ 10+0.1 th 1 Use log2 of plies (via msb) as depth bound for history pruning. Take 2
15-10-24 tvi NearDraw diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 7427 W: 1378 L: 1472 D: 4577
sprt @ 10+0.1 th 1 Penalty/Bonus for approaching draw, try continuous scaling depending on std::abs(eval)
15-10-23 IIv lazy_smp diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 21016 W: 3510 L: 3549 D: 13957
sprt @ 10+0.1 th 7 Do we have improvement on 7 threads? The formula coincides with master's on 3 threads and have good distribution on higher number of threads - explained on the forum.
15-10-24 tvi NearDraw diff
LLR: -0.16 (-2.94,2.94) [0.00,5.00]
Total: 38 W: 6 L: 13 D: 19
sprt @ 10+0.1 th 1 Penalty/Bonus for approaching draw, only when winning/losing
15-10-23 jhe log2_depth diff
ELO: -2.92 +-5.5 (95%) LOS: 15.0%
Total: 5000 W: 802 L: 844 D: 3354
5000 @ 10+0.1 th 7 Correct Depth Rounding.
15-10-23 Voy PlyPieceToHistory diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 20208 W: 3848 L: 3885 D: 12475
sprt @ 10+0.1 th 1 Try ply-piece-toSq history stat table.
15-10-24 Voy PPTH2 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 26211 W: 5124 L: 5133 D: 15954
sprt @ 10+0.1 th 1 Version 2
15-10-23 Voy PlyToHistory2 diff
LLR: 2.96 (-2.94,2.94) [0.00,5.00]
Total: 35701 W: 7008 L: 6721 D: 21972
sprt @ 10+0.1 th 1 Version 2
15-10-23 sni pawn_span2 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 29074 W: 5617 L: 5614 D: 17843
sprt @ 10+0.1 th 1 Take 4: use weights = (3,10)
15-10-23 jhe log2_depth diff
ELO: 0.97 +-5.7 (95%) LOS: 63.1%
Total: 5000 W: 881 L: 867 D: 3252
5000 @ 10+0.1 th 3 Correct Depth Rounding.
15-10-23 tvi NearDraw diff
LLR: 2.95 (-2.94,2.94) [0.00,5.00]
Total: 12221 W: 2426 L: 2245 D: 7550
sprt @ 10+0.1 th 1 Penalty/Bonus for approaching draw, triple bonus
15-10-23 sni NearDraw_in_eval diff
LLR: -1.78 (-2.94,2.94) [0.00,5.00]
Total: 5987 W: 1088 L: 1138 D: 3761
sprt @ 10+0.1 th 1 Try to see if it's possible to simulate Tom Vijlbrief's "NearDraw" anti-shuffling idea in eval level too.
15-10-23 sni NearDraw_in_eval diff
LLR: -1.21 (-2.94,2.94) [0.00,5.00]
Total: 1196 W: 211 L: 258 D: 727
sprt @ 10+0.1 th 1 Try to see if it's possible to simulate Tom Vijlbrief's "NearDraw" anti-shuffling idea in eval level too (double bonus)
15-10-23 Voy PlyToHistory3 diff
LLR: -1.95 (-2.94,2.94) [0.00,5.00]
Total: 4409 W: 821 L: 885 D: 2703
sprt @ 10+0.1 th 1 Version 3
15-10-23 jhe log2_depth diff
ELO: -6.12 +-5.5 (95%) LOS: 1.4%
Total: 5000 W: 763 L: 851 D: 3386
5000 @ 10+0.1 th 7 Use log2 thread depth calculation.
15-10-23 Voy PlyToHistory diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 9750 W: 1788 L: 1872 D: 6090
sprt @ 10+0.1 th 1 New History Stat Idea: Ply-To Square stats.
15-10-23 tvi NearDraw diff
LLR: -2.74 (-2.94,2.94) [0.00,5.00]
Total: 9155 W: 1721 L: 1798 D: 5636
sprt @ 10+0.1 th 1 Penalty/Bonus for approaching draw, take 2, halve bonus
15-10-23 tvi NearDraw diff
LLR: -3.19 (-2.94,2.94) [0.00,5.00]
Total: 20881 W: 4002 L: 4046 D: 12833
sprt @ 10+0.1 th 1 Penalty/Bonus for approaching draw, take 2
15-10-23 Voy Hpt2 diff
LLR: -2.94 (-2.94,2.94) [0.00,4.00]
Total: 7942 W: 1468 L: 1598 D: 4876
sprt @ 10+0.1 th 1 Just use CMH for History Pruning
15-10-23 tvi NearDraw diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 14148 W: 2676 L: 2740 D: 8732
sprt @ 10+0.1 th 1 Penalty/Bonus for approaching draw
15-10-23 sg dynamicHP diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 8055 W: 1503 L: 1594 D: 4958
sprt @ 10+0.1 th 1 Use dynamic depth bound for history pruning. Test against passed history pruning patch from VoyagerOne
15-10-21 sni pawn_span2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 90886 W: 17917 L: 17633 D: 55336
sprt @ 10+0.1 th 1 Continue with Stefan's idea of pawn span
15-10-23 sni ab_window diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 9604 W: 1843 L: 1927 D: 5834
sprt @ 10+0.1 th 1 Use transposition table to reduce the (alpha,beta) window at PV nodes.
15-10-23 sni ab_window diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 4442 W: 829 L: 937 D: 2676
sprt @ 10+0.1 th 1 Take 2: more conservative, only reduce window when tte->depth()==depth.
15-10-22 IIv delta_values diff
LLR: -1.89 (-2.94,2.94) [0.00,4.00]
Total: 50594 W: 9757 L: 9678 D: 31159
sprt @ 10+0.1 th 1 New delta values, Take2
15-10-23 Fis 4men_TBDepthTTBias diff
LLR: -3.13 (-2.94,2.94) [0.00,4.00]
Total: 44838 W: 8630 L: 8638 D: 27570
sprt @ 10+0.1 th 1 Increase depth bias for TB nodes in the TT. 4MB STC
15-10-22 sni pawn_span2 diff
LLR: -3.09 (-2.94,2.94) [0.00,5.00]
Total: 33933 W: 6565 L: 6546 D: 20822
sprt @ 10+0.1 th 1 Take 3: use weights = (6,4)
15-10-23 Voy ttDepth diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 10034 W: 1924 L: 2006 D: 6104
sprt @ 10+0.1 th 1 Use tt's depth for scoring history stats.
15-10-22 mco ybw_removal diff
LLR: -2.95 (-2.94,2.94) [-3.00,1.00]
Total: 36125 W: 6017 L: 6230 D: 23878
sprt @ 10+0.1 th 7 Use std::mutex now that is no more in hot path. Test with 7 threads to increase pressure.
15-10-23 Voy MCPt diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 10531 W: 1986 L: 2066 D: 6479
sprt @ 10+0.1 th 1 Use History stats to assist in move count pruning.
15-10-22 lbr min_thinking_time diff
LLR: 0.14 (-2.94,2.94) [-3.00,1.00]
Total: 31311 W: 4606 L: 4644 D: 22061
sprt @ 60+0.4 th 1 LTC: Attempt to correct minimum thinking time logic. Apply as a minimum, instead of an add-on, and use a dynamic value, instead of a hardcoded one.
15-10-22 Voy HPt diff
LLR: -2.98 (-2.94,2.94) [0.00,4.00]
Total: 10693 W: 2028 L: 2150 D: 6515
sprt @ 10+0.1 th 1 Set depth at 4 for History Pruning.