Stockfish Testing Queue

Finished - 45450 tests

14-03-20 jos iid_tweak diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 38473 W: 7161 L: 7166 D: 24146
sprt @ 15+0.05 th 1 As we arrive here with under 1% as PvNode && !ttMove, I simplified the precondition. As this noticeably affects functionality, I test in standard mode. Take 2.
14-03-20 sni LMR_with_eval diff
LLR: 2.97 (-2.94,2.94) [-1.50,4.50]
Total: 18245 W: 3461 L: 3308 D: 11476
sprt @ 15+0.05 th 1 Recurse to qsearch() in LMR if reduced depth is less than ONE_PLY
14-03-20 sni LMR_with_eval diff
LLR: -2.96 (-2.94,2.94) [0.00,6.00]
Total: 16790 W: 2566 L: 2592 D: 11632
sprt @ 60+0.05 th 1 LTC : recurse to qsearch() in LMR if reduced depth is less than ONE_PLY
14-03-21 Mys rBeta_tweak1 diff
LLR: -1.59 (-2.94,2.94) [-1.50,4.50]
Total: 3954 W: 700 L: 746 D: 2508
sprt @ 15+0.05 th 1 Introduce a factor of - depth / 8 to the singular extension margin.
14-03-21 Fis tt_move diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 14221 W: 2602 L: 2670 D: 8949
sprt @ 15+0.05 th 1 Value tt entries w/ moves more. Test w/ 16MB to simulate more normal hash pressure.
14-03-21 sni LMR_with_eval diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 721 W: 81 L: 183 D: 457
sprt @ 15+0.05 th 1 Increase LMR reduction for positions which seem to be easy to prove/disprove (100). Low priority.
14-03-21 sni LMR_with_eval diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 1714 W: 267 L: 367 D: 1080
sprt @ 15+0.05 th 1 Increase LMR reduction for positions which seem to be easy to prove/disprove (200). Low priority.
14-03-21 sni LMR_with_eval diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 2779 W: 457 L: 555 D: 1767
sprt @ 15+0.05 th 1 Increase LMR reduction for positions which seem to be easy to prove/disprove (300). Low priority.
14-03-21 sni LMR_with_eval diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 5315 W: 967 L: 1059 D: 3289
sprt @ 15+0.05 th 1 Increase LMR reduction for positions which seem to be easy to prove/disprove (400). Low priority.
14-03-21 Mys rBeta_tweak2 diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 14877 W: 2742 L: 2808 D: 9327
sprt @ 15+0.05 th 1 Try the same concept as aspiration delta. Split rBeta margin according to depth. Likely last take.
14-03-21 sg rook_ending diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 9248 W: 1705 L: 1786 D: 5757
sprt @ 15+0.05 th 1 drawish rooke nding (lower scale factor) Take 5
14-03-21 Fis tt_move diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 25574 W: 4697 L: 4736 D: 16141
sprt @ 15+0.05 th 1 Value tt entries w/ moves more. Test w/ 16MB to simulate more normal hash pressure. Take 2
14-03-21 rst capt_ord diff
LLR: -2.96 (-2.94,2.94) [0.00,6.00]
Total: 7212 W: 1058 L: 1127 D: 5027
sprt @ 60+0.05 th 1 Speculative LTC: Capture move ordering. Prioritize moves into the center of the board for pawn, knight and bishop and off the center for rook and queen, king. Take 4
14-03-21 jos fut_tweak diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 21856 W: 4020 L: 4068 D: 13768
sprt @ 15+0.05 th 1 Do not prune ttMoves.
14-03-21 rst capt_ord diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 10144 W: 1838 L: 1917 D: 6389
sprt @ 15+0.05 th 1 Capture move ordering. Prioritize moves into the center of the board. Final take 8.
14-03-21 sni LMR_with_eval diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 4009 W: 686 L: 780 D: 2543
sprt @ 15+0.05 th 1 Don't even try LMR if eval >= alpha + 200. Low priority.
14-03-21 sni LMR_with_eval diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 3324 W: 562 L: 658 D: 2104
sprt @ 15+0.05 th 1 Don't even try LMR if eval >= alpha + 150. Low priority.
14-03-21 sni LMR_with_eval diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 1725 W: 269 L: 369 D: 1087
sprt @ 15+0.05 th 1 Don't even try LMR if eval >= alpha + 100. Low priority.
14-03-21 Fis tt_replace diff
LLR: -2.59 (-2.94,2.94) [-1.50,4.50]
Total: 128000 W: 23662 L: 23422 D: 80916
sprt @ 15+0.05 th 1 If neither tt entry is current and depths are very close prefer the more recent. Test w/ 16MB to simulate more normal hash pressure.
14-03-21 Fis tt_replace diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 4720 W: 831 L: 924 D: 2965
sprt @ 15+0.05 th 1 If neither tt entry is current and depths are very close prefer the more recent. Test w/ 16MB to simulate more normal hash pressure. Take 2
14-03-21 sg rook_ending diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 9060 W: 1641 L: 1723 D: 5696
sprt @ 15+0.05 th 1 drawish if pawns on same wing. inspired by http://tests.stockfishchess.org/tests/view/53083ba20ebc5901df50fb94 (Take 6)
14-03-21 sg rook_ending diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 7012 W: 1249 L: 1336 D: 4427
sprt @ 15+0.05 th 1 drawish if max 6 pawns (Take 7)
14-03-21 sg rook_ending diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 32853 W: 6121 L: 6140 D: 20592
sprt @ 15+0.05 th 1 drawish if no backward pawn at weaker side (Take 8)
14-03-22 Fis tt_replace diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 12052 W: 2140 L: 2214 D: 7698
sprt @ 15+0.05 th 1 Refresh an entries generation automatically on probe. Minor simplification as well. Test w/ 16MB to simulate more normal hash pressure.
14-03-22 sni isolated_passed diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 3461 W: 650 L: 748 D: 2063
sprt @ 15+0.05 th 1 Remove the malus for passed isolated pawns
14-03-22 sni isolated_passed diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 7822 W: 1411 L: 1496 D: 4915
sprt @ 15+0.05 th 1 Halve the isolation malus for passed isolated pawns
14-03-22 sni isolated_passed diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 2142 W: 364 L: 464 D: 1314
sprt @ 15+0.05 th 1 Don't isolate A and H pawns (take 1)
14-03-22 sni isolated_passed diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 5899 W: 1098 L: 1188 D: 3613
sprt @ 15+0.05 th 1 Don't isolate A and H pawns (take 2)
14-03-22 Mys rBeta_tweak3 diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 15548 W: 2889 L: 2954 D: 9705
sprt @ 15+0.05 th 1 Finally try a smaller factor.
14-03-22 tzi increase_mobility diff
LLR: -3.00 (-2.94,2.94) [-1.50,4.50]
Total: 8369 W: 1560 L: 1645 D: 5164
sprt @ 15+0.05 th 1 Increase global weight for mobility
14-03-22 sni isolated_passed diff
LLR: -2.94 (-2.94,2.94) [-1.50,4.50]
Total: 41078 W: 7953 L: 7948 D: 25177
sprt @ 15+0.05 th 1 Weak (=unsupported) pawns penalty
14-03-22 hwi futility_margin_67 diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 16576 W: 3035 L: 3097 D: 10444
sprt @ 15+0.05 th 1 Test futility margin seemed to work in previous test.
14-03-22 Fis tt_replace diff
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 18814 W: 3600 L: 3475 D: 11739
sprt @ 15+0.05 th 1 Refresh an entries generation automatically on probe. Test as a small simplification. Using 32MB to simulate more normal hash pressure.
14-03-22 sni isolated_passed diff
LLR: 2.96 (-2.94,2.94) [-1.50,4.50]
Total: 26751 W: 5234 L: 5056 D: 16461
sprt @ 15+0.05 th 1 Weak (=unsupported) pawns penalty. Take 2.
14-03-22 sni isolated_passed diff
LLR: 2.96 (-2.94,2.94) [-1.50,4.50]
Total: 12107 W: 2411 L: 2272 D: 7424
sprt @ 15+0.05 th 1 Weak (=unsupported) pawns penalty. Take 3.
14-03-23 Fis tt_replace diff
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 20731 W: 3217 L: 3096 D: 14418
sprt @ 60+0.05 th 1 Refresh an entries generation automatically on probe. Test as a small simplification. Using 64MB for LTC to simulate more normal hash pressure.
14-03-23 sni LMR_with_eval diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 5589 W: 974 L: 1064 D: 3551
sprt @ 15+0.05 th 1 Retire cutnode. Rather drastic simplification : remove the recursive cutnode parameter of search(), instead use an estimation of cut probability based on eval inside LMR.
14-03-23 mco swap_extmove diff
LLR: -2.95 (-2.94,2.94) [0.00,6.00]
Total: 12249 W: 2251 L: 2292 D: 7706
sprt @ 15+0.05 th 1 Test for speed optimization
14-03-23 mco tt_replacement diff
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 32046 W: 6020 L: 5918 D: 20108
sprt @ 15+0.05 th 1 TT replace simplification: test for no-regression (it is a no functional change)
14-03-23 sni isolated_passed diff
LLR: 1.05 (-2.94,2.94) [-1.50,4.50]
Total: 2350 W: 472 L: 428 D: 1450
sprt @ 15+0.05 th 1 LTC for : weak (=unsupported) pawns penalty. Take 3.
14-03-23 sni isolated_passed diff
LLR: -1.17 (-2.94,2.94) [0.00,6.00]
Total: 25708 W: 4204 L: 4124 D: 17380
sprt @ 60+0.05 th 1 LTC for weak (=unsupported) pawns penalty, take 2. This a low priority run, because take 3 has passed STC also and is tested at LTC.
14-03-23 sni isolated_passed diff
LLR: 2.96 (-2.94,2.94) [0.00,6.00]
Total: 9204 W: 1605 L: 1458 D: 6141
sprt @ 60+0.05 th 1 LTC for : weak (=unsupported) pawns penalty. Take 3.
14-03-23 jos fut_tweak diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 17546 W: 3204 L: 3263 D: 11079
sprt @ 15+0.05 th 1 A random shot at futility margin, based on search stats.
14-03-23 rst probcut_after_null_agai diff
LLR: 2.96 (-2.94,2.94) [-1.50,4.50]
Total: 63601 W: 11912 L: 11641 D: 40048
sprt @ 15+0.05 th 1 allow probcut after null move when do_null_move() clears capturedType
14-03-23 mco new_see diff
LLR: -2.96 (-2.94,2.94) [-3.00,1.00]
Total: 26235 W: 4812 L: 5016 D: 16407
sprt @ 15+0.05 th 1 Rewrite new SEE: simplification, test for no-regressions
14-03-23 jhe bishop_op diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 26643 W: 4900 L: 4936 D: 16807
sprt @ 15+0.05 th 1 Tune bishop outposts.
14-03-23 hxi pawn_bonus diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 10190 W: 1863 L: 1942 D: 6385
sprt @ 15+0.05 th 1 Doubled, isolated and backward pawn penalty adjustment
14-03-23 Fis tt_replace diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 10893 W: 1982 L: 2059 D: 6852
sprt @ 15+0.05 th 1 If neither tt entry is current and depths are close prefer the more recent. Test w/ 32MB to simulate more normal hash pressure. Take 3
14-03-24 rst probcut_after_null_agai diff
LLR: -2.95 (-2.94,2.94) [0.00,6.00]
Total: 12161 W: 1868 L: 1914 D: 8379
sprt @ 60+0.05 th 1 allow probcut after null move when do_null_move() clears capturedType
14-03-24 rst retire_cutnode_2 diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 5652 W: 989 L: 1079 D: 3584
sprt @ 15+0.05 th 1 Retire cutnode. Rather drastic simplification : remove the recursive cutnode parameter of search(), instead use an estimation of cut probability based on eval inside LMR. Take 2.