Stockfish Testing Queue

Pending - 0 tests 0.0 hrs

None

Active - 0 tests

Finished - 27 tests

25-07-17 An statFormula1 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 27892 W: 5041 L: 5047 D: 17804
sprt @ 10+0.1 th 1 Trying a formula derived from a set of test positions
23-07-17 An statComparisonByDepth2 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 35837 W: 6392 L: 6366 D: 23079
sprt @ 10+0.1 th 1 Final take. Passing at 1+.01s, and the last test failing with M=35*depth, suggests this patch is only good for low depths, so invert the scaling.
23-07-17 An statComparisonByDepth diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 23097 W: 4066 L: 4094 D: 14937
sprt @ 10+0.1 th 1 Add a depth component to try to scale it for longer time controls
23-07-17 An statComparisonTweak diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 7672 W: 1320 L: 1450 D: 4902
sprt @ 10+0.1 th 1 STC. Passed [0,5] locally at 1+.01s. See if change works with higher TC.
23-07-17 An statReductions2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 4986 W: 848 L: 952 D: 3186
sprt @ 10+0.1 th 1 Take 2. Less aggressive.
23-07-17 An statReductions diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 10303 W: 1773 L: 1855 D: 6675
sprt @ 10+0.1 th 1 Slowly increase LMR as more moves go through the LMR process without needing to be researched.
22-07-17 An limitCutnodeLMR diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 21547 W: 3813 L: 3847 D: 13887
sprt @ 10+0.1 th 1 Slightly reduce LMR for cutnodes in some cases.
21-07-17 An reduceCaptures2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 10161 W: 1798 L: 1880 D: 6483
sprt @ 10+0.1 th 1 Take 2 STC
21-07-17 An reduceCaptures diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 7607 W: 1297 L: 1390 D: 4920
sprt @ 10+0.1 th 1 Try reducing captures given a good tt move in cut nodes. Do this until any move upsets the ttMove, or any reduced capture is worthy of a research.
20-07-17 An ttCutNodes diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 30156 W: 5318 L: 5317 D: 19521
sprt @ 10+0.1 th 1 LMR based on ttCapture is a wash for cutnodes. Add more conditions for cutnodes.
20-07-17 An ttMoveHoldsFix diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 19809 W: 3548 L: 3589 D: 12672
sprt @ 10+0.1 th 1 STC
20-07-17 An ttLMRCapture diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 20526 W: 3655 L: 3693 D: 13178
sprt @ 10+0.1 th 1 STC
19-07-17 An ttDepthForLMR2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 20997 W: 3732 L: 3768 D: 13497
sprt @ 10+0.1 th 1 Take 2
19-07-17 An ttDepthForLMR diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 13723 W: 2448 L: 2515 D: 8760
sprt @ 10+0.1 th 1 TTE's from qsearch may not give a proper indication of a capture beating quiets
18-07-17 An razorMargins4 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 9654 W: 1663 L: 1747 D: 6244
sprt @ 10+0.1 th 1 Fixed a mistake. The assumption was that if there is a ttCapture, then a capture is likely the best, and qsearch will return a good score, so we can lower razoring margins. However, if the ttCapture came from the qsearch, then a quiet could still be best.
18-07-17 An razorMargins2 diff
LLR: -1.34 (-2.94,2.94) [0.00,5.00]
Total: 31602 W: 5583 L: 5508 D: 20511
sprt @ 10+0.1 th 1 Separate razor margins for searches with a ttCapture. Looked promising after local testing.
18-07-17 An razorMargins3 diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 8185 W: 1455 L: 1545 D: 5185
sprt @ 10+0.1 th 1 One more take, adding in the Depth 1 value from VoyagerOne's (currently) yellow test.
17-07-17 An razorMargins diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 7826 W: 1375 L: 1467 D: 4984
sprt @ 10+0.1 th 1 For depth 1 razoring, use a lower margin if the table is noisy. For depth 2 & 3, use a higher margin if the table is noisy.
16-07-17 An ttNonCheckingCapture diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 23884 W: 4231 L: 4256 D: 15397
sprt @ 10+0.1 th 1 Instead of increasing LMR for ttCaptures, only increase for ttCaptures that do not give check
16-07-17 An ttHoldsOrIsDeep diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 22023 W: 3910 L: 3942 D: 14171
sprt @ 10+0.1 th 1 Increase R for LMR if the table move was a capture, but only if the capture holds best OR came from a greater or equal draft search.
15-07-17 An noRazorWhenTactical diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 25937 W: 4647 L: 4662 D: 16628
sprt @ 10+0.1 th 1 Do not Razor above depth 1 if the table move exists and was tactical. (Fixed version)
15-07-17 An ttMoveCount diff
LLR: -3.32 (-2.94,2.94) [0.00,5.00]
Total: 17987 W: 3144 L: 3209 D: 11634
sprt @ 10+0.1 th 1 Reduce the thresh hold to set moveCountPruning if the table move was noisy
15-07-17 An anyTTMHoldsMoveCount diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 3428 W: 556 L: 666 D: 2206
sprt @ 10+0.1 th 1 Take 3, Final Attempt. Reduce threshold for move count pruning by 1 if the table move holds as the best move.
15-07-17 An ttHoldsMoveCountImp diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 4038 W: 664 L: 772 D: 2602
sprt @ 10+0.1 th 1 Reduce moveCountPruning threshold if position is improving, table is noisy, and table holds best. May try one more version without table = noisy condition.
03-07-17 An ttCaptureDepthBonus2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 23606 W: 4197 L: 4222 D: 15187
sprt @ 10+0.1 th 1 Increase R for LMR if table move is tactical and came from a higher depth (dropped the still best req.)
03-07-17 An ttCaptureDepthBonus diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 25874 W: 4643 L: 4658 D: 16573
sprt @ 10+0.1 th 1 Increase R if ttMove holds best, is tactical, and came from a greater draft than the current search
01-07-17 An ttCaptureHolds diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 35467 W: 6348 L: 6323 D: 22796
sprt @ 10+0.1 th 1 If table move was a capture, increase LMR for quiet moves, but only if the table move is still the best move.