Stockfish Testing Queue

Pending - 0 tests 0.0 hrs

None

Active - 0 tests

Finished - 1020 tests

18-07-20 pb0 altStrikeBack diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 5246 W: 1096 L: 1202 D: 2948
sprt @ 10+0.1 th 1 Take 4
18-07-20 pb0 altStrikeBack diff
ELO: -1.76 +-2.3 (95%) LOS: 6.4%
Total: 40000 W: 8787 L: 8990 D: 22223
40000 @ 10+0.1 th 1 Trying to minimize the overhead by packing the new move into killers no functional change
18-07-20 pb0 altStrikeBack diff
ELO: 1.04 +-2.3 (95%) LOS: 81.6%
Total: 40000 W: 8936 L: 8816 D: 22248
40000 @ 10+0.1 th 1 Want to measure the pure tradeoff of handling an possible alternative strike-back square in see. no functional change
18-07-20 pb0 altStrikeBack diff
ELO: -1.50 +-2.3 (95%) LOS: 9.7%
Total: 40000 W: 8799 L: 8972 D: 22229
40000 @ 10+0.1 th 1 Since I don't understand what's wrong with the patch, measure again the pure tradeoff of handling now also with queen en prise detecting logic. no functional change
18-07-20 pb0 altStrikeBack diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 11717 W: 2540 L: 2614 D: 6563
sprt @ 10+0.1 th 1 Take 2: handle direct check corner case
18-07-20 pb0 altStrikeBack diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 9657 W: 2131 L: 2215 D: 5311
sprt @ 10+0.1 th 1 Take 1: use SEE strikeBack for our Queen en prise
18-07-19 pb0 hanging1 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 4978 W: 1059 L: 1166 D: 2753
sprt @ 10+0.1 th 1 Idea by Brian: count as hanging pieces attacked by more than one, and defended only by the king.
18-07-19 pb0 pinOnBackRank diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 23830 W: 5301 L: 5314 D: 13215
sprt @ 10+0.1 th 1 Penalty for minor piece pinned at back rank inspired by Brians analisys of this game https://groups.google.com/d/msg/lczero/7Ffntxej6gc/eE3PDlYXAgAJ
18-07-19 pb0 pinOnBackRank diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 22618 W: 5056 L: 5075 D: 12487
sprt @ 10+0.1 th 1 Take 2: only endgame penalty
18-07-19 pb0 afterNull diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 16612 W: 3649 L: 3699 D: 9264
sprt @ 10+0.1 th 1 No capture stat bonus if previous move was null-move
18-07-18 pb0 sanityCheck diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 10009 W: 2186 L: 2268 D: 5555
sprt @ 5+0.05 th 3 Sanity check on 3 threads for offset removal
18-07-18 pb0 extraPenalty2 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 10737 W: 2284 L: 2363 D: 6090
sprt @ 10+0.1 th 1 Take 4
18-07-18 pb0 extraPenalty2 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 15292 W: 3374 L: 3430 D: 8488
sprt @ 10+0.1 th 1 Take 3
18-07-18 pb0 extraPenalty2 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 11748 W: 2577 L: 2651 D: 6520
sprt @ 10+0.1 th 1 Take 2, fixed logic
18-07-17 pb0 threatQR diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 13452 W: 2957 L: 3022 D: 7473
sprt @ 10+0.1 th 1 Take 1
18-07-17 pb0 extraPenalty2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 5396 W: 1161 L: 1266 D: 2969
sprt @ 10+0.1 th 1 Extra penalty for the first capture in previous ply when it gets refuted
18-07-16 pb0 lmr_cmh_diff diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 46037 W: 7922 L: 7858 D: 30257
sprt @ 60+0.6 th 1 Fix captured piece condition, LTC for Stephan
18-07-16 pb0 updCaptStat diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 12609 W: 2787 L: 2856 D: 6966
sprt @ 10+0.1 th 1 I guess this was not tried yet, since we penalize capture moves that fail to create a cutoff even at quiet move cutoffs.
18-07-16 pb0 goodCaptureOffset diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 47327 W: 10641 L: 10617 D: 26069
sprt @ 10+0.1 th 1 Take 3
18-07-16 pb0 updCaptStat diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 40202 W: 8944 L: 8876 D: 22382
sprt @ 10+0.1 th 1 Retry this idea, now that we update with stat_bonus(depth + ONE_PLY)
18-07-16 pb0 updCaptStat diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 12607 W: 2831 L: 2900 D: 6876
sprt @ 10+0.1 th 1 Take 2 (changed logic)
18-07-16 pb0 resetTTmove diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 8441 W: 1782 L: 1872 D: 4787
sprt @ 10+0.1 th 1 Take 2
18-07-16 pb0 resetTTmove diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 2962 W: 629 L: 747 D: 1586
sprt @ 10+0.1 th 1 Reset TT-Move on expected all-nodes it fails low with a value < alpha
18-07-13 pb0 goodCaptureOffset^ diff
LLR: -2.94 (-2.94,2.94) [0.00,4.00]
Total: 13881 W: 2296 L: 2405 D: 9180
sprt @ 60+0.6 th 1 Take 1, LTC
18-07-13 pb0 goodCaptureOffset^ diff
LLR: 2.95 (-2.94,2.94) [0.00,4.00]
Total: 90971 W: 20286 L: 19762 D: 50923
sprt @ 10+0.1 th 1 Take 1
18-07-13 pb0 goodCaptureOffset diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 46799 W: 7939 L: 7942 D: 30918
sprt @ 60+0.6 th 1 Take 2, LTC
18-07-13 pb0 goodCaptureOffset diff
LLR: 2.95 (-2.94,2.94) [0.00,4.00]
Total: 43703 W: 9798 L: 9461 D: 24444
sprt @ 10+0.1 th 1 Take 2
18-07-12 pb0 probcutOffset diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 20222 W: 4424 L: 4456 D: 11342
sprt @ 10+0.1 th 1 Use escape detection ONLY in normal LMR
18-07-12 pb0 probcutOffset diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 18263 W: 4015 L: 4056 D: 10192
sprt @ 10+0.1 th 1 Use escape detection also in normal LMR
18-07-12 pb0 probcutOffset^ diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 14920 W: 3227 L: 3285 D: 8408
sprt @ 10+0.1 th 1 No shallow cm based pruning if move escapes a capture (fixed logic)
18-07-12 pb0 probcutOffset diff
LLR: 0.13 (-2.94,2.94) [0.00,5.00]
Total: 673 W: 145 L: 136 D: 392
sprt @ 10+0.1 th 1 No shallow countermove based pruning if move escapes a capture
18-07-11 pb0 probcutOffset diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 31863 W: 7005 L: 6979 D: 17879
sprt @ 10+0.1 th 1 Another shot at this idea
18-07-11 pb0 probcutOffset diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 19708 W: 4306 L: 4340 D: 11062
sprt @ 10+0.1 th 1 Only if sum of cm-histories is negative
18-07-10 pb0 probCut5 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 11734 W: 2624 L: 2698 D: 6412
sprt @ 10+0.1 th 1 Take 7
18-07-09 pb0 probCut5 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 42500 W: 9387 L: 9309 D: 23804
sprt @ 10+0.1 th 1 Take 6
18-07-09 pb0 threadDist diff
LLR: -2.95 (-2.94,2.94) [-3.00,1.00]
Total: 35350 W: 7718 L: 7953 D: 19679
sprt @ 5+0.05 th 3 Alternative threading scheme. Also testing corner case of 3 threads, where odd rootDepth's only get searched by the main thread.
18-07-09 pb0 probCut5 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 15782 W: 3463 L: 3517 D: 8802
sprt @ 10+0.1 th 1 Take 4, only at cut-nodes with assist from nullmove-detection
18-07-09 pb0 probCut5 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 5452 W: 1143 L: 1248 D: 3061
sprt @ 10+0.1 th 1 Take 5, only at very shallow depth's with negative CMPrune stat values
18-07-06 pb0 simpleHalfDensity diff
ELO: -0.03 +-2.9 (95%) LOS: 49.0%
Total: 20000 W: 3525 L: 3527 D: 12948
20000 @ 10+0.1 th 43 Trying a simpler half-density approach, 43 threads (= TCEC conditions)
18-07-06 pb0 simpleHalfDensity diff
LLR: -2.96 (-2.94,2.94) [-3.00,1.00]
Total: 20716 W: 4110 L: 4310 D: 12296
sprt @ 5+0.05 th 8 Trying a simpler half-density approach. Test against passed patch.
18-07-06 pb0 probCut5^ diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 45513 W: 10140 L: 10046 D: 25327
sprt @ 10+0.1 th 1 Another shot at this idea
18-07-06 pb0 probCut5 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 10922 W: 2396 L: 2474 D: 6052
sprt @ 10+0.1 th 1 Take 2, only on all-nodes
18-07-05 pb0 pawnthreats diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 33354 W: 7431 L: 7397 D: 18526
sprt @ 10+0.1 th 1 Take 2, don't shallow prune sharp pawn pushes to much
18-07-05 pb0 pawnthreats diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 33457 W: 7368 L: 7335 D: 18754
sprt @ 10+0.1 th 1 Don't shallow prune sharp pawn pushes to much
18-07-05 pb0 tweakAdvancedPawnPush diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 10425 W: 2276 L: 2356 D: 5793
sprt @ 10+0.1 th 1 also consider legal pushes threatening non-pawn enemies
18-07-04 pb0 probcutOffset diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 11902 W: 2642 L: 2715 D: 6545
sprt @ 10+0.1 th 1 Take 10
18-07-04 pb0 probcutOffset diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 31195 W: 6971 L: 6948 D: 17276
sprt @ 10+0.1 th 1 Take 10: this is a nonfunctional optimization (speedup) of take 7
18-07-02 pb0 probcutOffset diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 305753 W: 52669 L: 51549 D: 201535
sprt @ 60+0.6 th 1 Take 7, LTC
18-07-03 pb0 probcutOffset diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 27164 W: 6013 L: 6010 D: 15141
sprt @ 10+0.1 th 1 Take 8
18-07-02 pb0 probcutOffset diff
LLR: 2.96 (-2.94,2.94) [0.00,5.00]
Total: 23032 W: 5212 L: 4966 D: 12854
sprt @ 10+0.1 th 1 Take 7