Stockfish Testing Queue

Finished - 40849 tests

16-11-09 fau doubled2 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 99340 W: 17586 L: 17297 D: 64457
sprt @ 10+0.1 th 1 Take 2
16-11-08 Voy improvingPruning diff
LLR: 2.95 (-2.94,2.94) [0.00,5.00]
Total: 54389 W: 9487 L: 9141 D: 35761
sprt @ 10+0.1 th 1 Increase factor if improving...
16-11-06 Roc PShelterTune diff
46415/50000 iterations
100000/100000 games played
100000 @ 10+0.1 th 1 3/4 Only StormDanger[1] (fix) (priority -1)
16-11-08 fau doubled diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 9390 W: 1608 L: 1693 D: 6089
sprt @ 10+0.1 th 1 Guardian's idea (Reduce the penalty for the doubled, if it is supported) But this time increasing the penalty itself for endgame (As suggested by my local tuning)
16-11-08 Gua Only-unsupported diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 33409 W: 5909 L: 5894 D: 21606
sprt @ 10+0.1 th 1 Use doubled penalty only if doubled is unsupported
16-11-08 atu repetition diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 15165 W: 2622 L: 2684 D: 9859
sprt @ 10+0.1 th 1 Implement the 3-fold repetition rule (using pre-root pseudostates)
16-11-07 Voy pt3 diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 118459 W: 14884 L: 14732 D: 88843
sprt @ 60+0.6 th 1 ltc: take 3
16-11-08 Elb flank_end diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 40056 W: 7088 L: 7045 D: 25923
sprt @ 10+0.1 th 1 Endings on a single flank are drawish, take 2
16-11-08 Gua Doubled-&-supported diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 77417 W: 13791 L: 13591 D: 50035
sprt @ 10+0.1 th 1 Reduce the penalty for the doubled, if it is supported
16-11-08 Elb flank_end diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 54056 W: 9435 L: 9336 D: 35285
sprt @ 10+0.1 th 1 Endings that take place on a single flank are drawish
16-11-08 Gua Doubled-&-unsupported-p diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 10566 W: 1787 L: 1868 D: 6911
sprt @ 10+0.1 th 1 increase the penalty for the doubled, if it is not supported
16-11-08 Fis 3stageQuiets diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 12861 W: 2219 L: 2290 D: 8352
sprt @ 10+0.1 th 1 Split quiets into 3 stages.
16-11-06 Roc PShelterTune diff
48104/50000 iterations
99776/100000 games played
100000 @ 20+0.2 th 1 2/4 Now only tuning the StormDanger[0]
16-11-06 Voy pt3 diff
LLR: 3.19 (-2.94,2.94) [0.00,4.00]
Total: 237009 W: 42093 L: 41135 D: 153781
sprt @ 10+0.1 th 1 take 3
16-11-07 vdv nodecount diff
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 32410 W: 5644 L: 5542 D: 21224
sprt @ 10+0.1 th 1 Test the make sure this doesn't regress actual play.
16-11-07 Elb lmr_stat diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 6928 W: 1185 L: 1318 D: 4425
sprt @ 10+0.1 th 1 Take 2: increase the divider
16-11-07 Elb lmr_stat diff
LLR: -2.94 (-2.94,2.94) [0.00,4.00]
Total: 2687 W: 400 L: 546 D: 1741
sprt @ 10+0.1 th 1 LMR stat tweak (run as [0,4] since it removes 2 lines)
16-11-07 Gua BishopPSQT1 diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 3551 W: 618 L: 764 D: 2169
sprt @ 10+0.1 th 1 b2/g2/b7/g7 large increase in bishop psqt
16-11-07 Elb lmrt1c diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 13384 W: 2336 L: 2405 D: 8643
sprt @ 10+0.1 th 1 A variation of VoyagerOne's lmrt idea (http://tests.stockfishchess.org/tests/view/581c96df0ebc5910626b9b33)
16-11-05 Roc PShelterTune diff
47316/50000 iterations
100000/100000 games played
100000 @ 20+0.2 th 1 After the last few months changes to king safety, I hope to see some changes here.
16-11-06 mco no_piecelist diff
LLR: -3.42 (-2.94,2.94) [-3.00,1.00]
Total: 33769 W: 4178 L: 4391 D: 25200
sprt @ 60+0.6 th 1 LTC: Retire piecelist: take 2 (added quick access to king square)
16-11-07 cib ss_hist diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 4244 W: 705 L: 812 D: 2727
sprt @ 10+0.1 th 1 ss->history tweak
16-11-06 Voy statS diff
LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 34327 W: 6080 L: 5982 D: 22265
sprt @ 10+0.1 th 1 Stat Simplification, with the hope of obtaining a bit of ELO
16-11-06 mco no_piecelist diff
LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 12666 W: 2285 L: 2151 D: 8230
sprt @ 10+0.1 th 1 Retire piecelist: take 2 (added quick access to king square)
16-11-06 Roc PawnShelterTuned1 diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 84514 W: 15125 L: 14996 D: 54393
sprt @ 10+0.1 th 1 SPSA tuned values after 46532 games.
16-11-06 Voy pt2 diff
LLR: -2.94 (-2.94,2.94) [0.00,4.00]
Total: 7669 W: 1325 L: 1455 D: 4889
sprt @ 10+0.1 th 1 stc
16-11-06 sg cutnode diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 23986 W: 4212 L: 4236 D: 15538
sprt @ 10+0.1 th 1 Try movecount = 20. Take 2
16-11-06 sg cutnode diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 17635 W: 3040 L: 3091 D: 11504
sprt @ 10+0.1 th 1 If after 29 moves no cutoff occured then set cutnode to false.
16-11-06 Gua patch-2 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 14738 W: 2582 L: 2645 D: 9511
sprt @ 10+0.1 th 1 Can we not use doubled penalty!?
16-11-05 lan numasf diff
ELO: 2.96 +-1.9 (95%) LOS: 99.9%
Total: 40000 W: 6205 L: 5864 D: 27931
40000 @ 10+0.1 th 3 Testing Thomas Zipproth's implementation of numa on a 3 threads. There was no regression on 1 thread.
16-11-06 hex lazy_quiet diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 21962 W: 3873 L: 3906 D: 14183
sprt @ 10+0.1 th 1 Partition simplified
16-11-05 Gua patch-1 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 22104 W: 3879 L: 3911 D: 14314
sprt @ 10+0.1 th 1 IsolatedDoubled penalty without new constants and 50% reducing
16-11-05 Voy pt diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 18768 W: 3309 L: 3355 D: 12104
sprt @ 10+0.1 th 1 pruning tweak
16-11-05 Fis partition diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 8899 W: 1523 L: 1611 D: 5765
sprt @ 10+0.1 th 1 Also put the 0 values ahead of the negative ones.
16-11-05 Voy lmrt1b diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 20766 W: 3623 L: 3661 D: 13482
sprt @ 10+0.1 th 1 stc
16-11-05 mco no_piecelist diff
LLR: -2.95 (-2.94,2.94) [-3.00,1.00]
Total: 113491 W: 19779 L: 20124 D: 73588
sprt @ 10+0.1 th 1 Respin an old patch from Lucas to retire piece lists
16-11-05 fau isolatedlast diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 11825 W: 1982 L: 2098 D: 7745
sprt @ 10+0.1 th 1 Observing "Gu" tests on isolated, I noticed that the ones who was faring the best was the ones with increased values for middlegame, so I try this last tuning and if it fails, it means that these values are perfectly tuned that any change on them would just harm the strength.
16-11-05 hex lazy_quiet diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 33620 W: 5980 L: 5964 D: 21676
sprt @ 10+0.1 th 1 Partition moves before pick_best
16-11-04 lan numasf diff
ELO: -0.09 +-2.0 (95%) LOS: 46.6%
Total: 40000 W: 6997 L: 7007 D: 25996
40000 @ 10+0.1 th 1 Testing Thomas Zipproth's implementation of numa on a single thread. If there is no regression we may commit on testing with big machines. Previous test was erroneous. Thanks, Ajith and Joerg for noticing
16-11-04 hex lazy_quiet diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 22554 W: 3989 L: 4019 D: 14546
sprt @ 10+0.1 th 1 Use pick_best instead of insert_sort if we have not enough quiet moves
16-11-04 Voy lmrt3 diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 40159 W: 7058 L: 7015 D: 26086
sprt @ 10+0.1 th 1 Variant 3
16-11-04 Gua patch-2 diff
LLR: -3.47 (-2.94,2.94) [0.00,5.00]
Total: 10672 W: 1768 L: 1870 D: 7034
sprt @ 10+0.1 th 1 IsolatedDoubled penalty
16-11-04 Voy lmrt2 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 13529 W: 2366 L: 2434 D: 8729
sprt @ 10+0.1 th 1 Take 2
16-11-04 Fis quietSort diff
LLR: -2.96 (-2.94,2.94) [-3.00,1.00]
Total: 4560 W: 714 L: 877 D: 2969
sprt @ 10+0.1 th 1 Always just sort the first partition and save time.
16-11-04 Elb strong_outposts diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 9261 W: 1589 L: 1675 D: 5997
sprt @ 10+0.1 th 1 Use tuned values
16-11-04 Gua patch-1 diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 5812 W: 957 L: 1093 D: 3762
sprt @ 10+0.1 th 1 Try to combine two yellow pawn patch FauziPawnsTuning + Reduce pawn structure score by 20% in midgame
16-11-04 Elb tune_strongoutposts diff
18464/20000 iterations
37858/40000 games played
40000 @ 10+0.1 th 1 As the queue is relatively empty, I would like to tune an extended version of a previous attempt which ended yellow http://tests.stockfishchess.org/tests/view/57b475fc0ebc591c761f6488
16-11-04 Voy lmrt diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 48656 W: 8712 L: 8632 D: 31312
sprt @ 10+0.1 th 1 stc
16-11-04 Voy ht2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 32789 W: 5678 L: 5667 D: 21444
sprt @ 10+0.1 th 1 stc
16-11-03 Roc PValueTune diff
13933/20000 iterations
28337/40000 games played
40000 @ 10+0.1 th 1 A pawn value experiment.