Stockfish Testing Queue

Finished - 38773 tests

15-03-24 Fis phaseDivorceTune2 diff
37712/40000 iterations
61710/80000 games played
80000 @ 15+0.05 th 1 One more try to separate piece material values from phase values. This time include pawns in phase and tune everything simultaneously including Eg values. Pri -1
15-03-24 mco tt_vector diff
LLR: -2.96 (-2.94,2.94) [-3.00,1.00]
Total: 26215 W: 4941 L: 5147 D: 16127
sprt @ 15+0.05 th 1 Use a std::vector to implement TT ( big simplification): Take 2, double prefetch.
15-03-24 lan rook_psqt diff
LLR: 3.06 (-2.94,2.94) [0.00,4.00]
Total: 51630 W: 10098 L: 9749 D: 31783
sprt @ 15+0.05 th 1 Increase values of rook psqt according to rank. Suggested by Lyudmil Tsvetkov
15-03-24 sg king_safety diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 10640 W: 1978 L: 2055 D: 6607
sprt @ 15+0.05 th 1 More penalty for pawns attacking king zone. Inspired by this thread http://talkchess.com/forum/viewtopic.php?t=55743&postdays=0&postorder=asc&topic_view=&start=10
15-03-24 sg king_safety2 diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 20368 W: 3972 L: 4023 D: 12373
sprt @ 15+0.05 th 1 More penalty for pawns which attacks the king zone (Take 2)
15-03-24 sg king_safety3 diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 7312 W: 1330 L: 1416 D: 4566
sprt @ 15+0.05 th 1 More penalty for pawns which attacks the king zone (Take 3)
15-03-24 Voy cmh_lmr3 diff
LLR: 2.97 (-2.94,2.94) [-1.50,4.50]
Total: 15391 W: 3051 L: 2903 D: 9437
sprt @ 15+0.05 th 1 Use history for margin.
15-03-24 Fis failHLTime diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 13651 W: 2582 L: 2651 D: 8418
sprt @ 15+0.05 th 1 Use 20% less time when the search fails high or low.
15-03-25 jos knight_psqt diff
ELO: 0.79 +-2.5 (95%) LOS: 73.1%
Total: 30000 W: 6138 L: 6070 D: 17792
30000 @ 9+0.05 th 1 Measure new knight values.
15-03-25 mco spinlock_500 diff
ELO: 0.49 +-3.9 (95%) LOS: 59.7%
Total: 10000 W: 1621 L: 1607 D: 6772
10000 @ 15+0.05 th 7 Spinlock: Busy wait a bit before to yield
15-03-25 lan rook_psqt diff
LLR: -2.97 (-2.94,2.94) [0.00,6.00]
Total: 22970 W: 3605 L: 3602 D: 15763
sprt @ 60+0.05 th 1 LTC: Increase values of rook psqt according to rank. Suggested by Lyudmil Tsvetkov
15-03-25 Fis phaseDivorce2 diff
LLR: -3.25 (-2.94,2.94) [-1.50,4.50]
Total: 54974 W: 10738 L: 10706 D: 33530
sprt @ 15+0.05 th 1 Separate piece material values from phase values.
15-03-25 lan rook_psqt diff
LLR: -3.27 (-2.94,2.94) [0.00,4.00]
Total: 43007 W: 6761 L: 6798 D: 29448
sprt @ 60+0.05 th 1 Resubmitting the rook psqt test with the right test parameters for tweak:[0,4] (please don't shoot me yet :). Suggested by Lyudmil Tsvetkov. Test parameters also suggested by Nicklas Persson.
15-03-25 sni pawn diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 23541 W: 4733 L: 4775 D: 14033
sprt @ 15+0.05 th 1 Don't apply structural penalties to pawns in the enemy camp
15-03-26 Voy cmh-lmr2 diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 13422 W: 2561 L: 2631 D: 8230
sprt @ 15+0.05 th 1 10 pt margin. Test against latest master.
15-03-26 Voy cmh-lmr3 diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 58198 W: 11121 L: 11072 D: 36005
sprt @ 15+0.05 th 1 Use history as margin. Test against latest master.
15-03-26 mco ttMoveCmh diff
LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 11560 W: 2321 L: 2183 D: 7056
sprt @ 15+0.05 th 1 Simplify ttMoveCmh bonus
15-03-26 mco ttMoveCmh diff
LLR: -2.95 (-2.94,2.94) [-3.00,1.00]
Total: 47516 W: 7392 L: 7617 D: 32507
sprt @ 60+0.05 th 1 LTC: Simplify ttMoveCmh bonus
15-03-26 Voy cmh-lmr3 diff
LLR: 2.96 (-2.94,2.94) [0.00,6.00]
Total: 6479 W: 1099 L: 967 D: 4413
sprt @ 60+0.05 th 1 This might be TC sensitive. Lets see what happens at LTC.
15-03-26 jos bishop_psqt diff
ELO: -0.67 +-2.6 (95%) LOS: 30.3%
Total: 30000 W: 6321 L: 6379 D: 17300
30000 @ 9+0.05 th 1 Measure new bishop values.
15-03-27 sni pawn2 diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 13023 W: 2511 L: 2582 D: 7930
sprt @ 15+0.05 th 1 Tweak backward pawns values
15-03-27 lbr opposed diff
LLR: -3.04 (-2.94,2.94) [-0.50,4.50]
Total: 36101 W: 6921 L: 6923 D: 22257
sprt @ 15+0.05 th 1 test this one for a small elo gain
15-03-27 sni quiet diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 9573 W: 1805 L: 1885 D: 5883
sprt @ 15+0.05 th 1 Use pawn attacks to help quiet moves ordering.
15-03-27 Voy cmh-lmr3 diff
LLR: 0.11 (-2.94,2.94) [-1.50,4.50]
Total: 136616 W: 26316 L: 25947 D: 84353
sprt @ 15+0.05 th 1 Since there is not much tests going on...I want to retry this test that failed yellow last time. I am interested to see if noise played a role in this.
15-03-28 lbr mobility^ diff
LLR: -2.97 (-2.94,2.94) [-0.50,4.50]
Total: 5525 W: 1017 L: 1125 D: 3383
sprt @ 15+0.05 th 1 safe rook mobility
15-03-28 lbr mobility diff
LLR: -2.95 (-2.94,2.94) [-3.50,0.50]
Total: 17811 W: 3359 L: 3566 D: 10886
sprt @ 15+0.05 th 1 remove safe mobility for queens
15-03-28 aji hanging_pawn_threats_A diff
LLR: 2.96 (-2.94,2.94) [-1.50,4.50]
Total: 105539 W: 20389 L: 20001 D: 65149
sprt @ 15+0.05 th 1 Give a reduced bonus for threats by hanging pawns: STC
15-03-28 mco nodestime diff
ELO: 0.12 +-2.7 (95%) LOS: 53.5%
Total: 26267 W: 5281 L: 5272 D: 15714
20000 @ 30+0.05 th 1 Another run with nodestime mode after fixing some bugs. This is just a test run to verify there are no time losses with virtual speed set at 600K nps
15-03-28 n_p SPSA-LMR diff
530/30000 iterations
681/60000 games played
60000 @ 15+0.05 th 1 SPSA-session on LMR parameters. Now that several search patches more have passed since Stefan's last tuning of these parameters, another attempt at tuning these is due. Included "cmh-lmr3".
15-03-28 aji hanging_pawn_threats_A diff
LLR: 2.95 (-2.94,2.94) [0.00,6.00]
Total: 9629 W: 1577 L: 1432 D: 6620
sprt @ 60+0.05 th 1 Give a reduced bonus for threats by hanging pawns: LTC
15-03-28 mco nodestime diff
ELO: 4.86 +-13.9 (95%) LOS: 75.3%
Total: 1000 W: 217 L: 203 D: 580
1000 @ 30+0.05 th 1 Poor man noise measuring: we expect this short series ends with an almost 0 ELO result: trial 1
15-03-28 mco nodestime diff
ELO: 8.69 +-13.9 (95%) LOS: 89.0%
Total: 1000 W: 221 L: 196 D: 583
1000 @ 30+0.05 th 1 Poor man noise measuring: we expect this short series ends with an almost 0 ELO result: trial 2
15-03-28 mco nodestime diff
ELO: -3.13 +-12.9 (95%) LOS: 31.8%
Total: 1000 W: 176 L: 185 D: 639
1000 @ 30+0.05 th 1 Poor man noise measuring: we expect this short series ends with an almost 0 ELO result: trial 3
15-03-28 mco nodestime diff
ELO: -10.43 +-13.6 (95%) LOS: 6.6%
Total: 1000 W: 185 L: 215 D: 600
1000 @ 30+0.05 th 1 Poor man noise measuring: we expect this short series ends with an almost 0 ELO result: trial 4
15-03-28 mco nodestime diff
ELO: -4.17 +-13.4 (95%) LOS: 27.2%
Total: 1000 W: 189 L: 201 D: 610
1000 @ 30+0.05 th 1 Poor man noise measuring: we expect this short series ends with an almost 0 ELO result: trial 5
15-03-28 mco nodestime diff
ELO: 1.45 +-6.0 (95%) LOS: 68.1%
Total: 5289 W: 1103 L: 1081 D: 3105
1000 @ 30+0.05 th 1 Poor man noise measuring: we expect this short series ends with an almost 0 ELO result: trial 6
15-03-28 mco nodestime diff
ELO: -1.39 +-13.0 (95%) LOS: 41.7%
Total: 1000 W: 181 L: 185 D: 634
1000 @ 30+0.05 th 1 Poor man noise measuring: we expect this short series ends with an almost 0 ELO result: trial 7
15-03-28 mco nodestime diff
ELO: 3.47 +-13.7 (95%) LOS: 69.1%
Total: 1000 W: 207 L: 197 D: 596
1000 @ 30+0.05 th 1 Poor man noise measuring: we expect this short series ends with an almost 0 ELO result: trial 8
15-03-28 mco nodestime diff
ELO: -7.30 +-13.5 (95%) LOS: 14.5%
Total: 1000 W: 186 L: 207 D: 607
1000 @ 15+0.05 th 1 Poor man noise measuring: standard TC run as a comparison. Series should end with bigger difference from 0 ELO: trial 1
15-03-28 mco nodestime diff
ELO: -5.91 +-13.5 (95%) LOS: 19.6%
Total: 1000 W: 189 L: 206 D: 605
1000 @ 15+0.05 th 1 Poor man noise measuring: standard TC run as a comparison. Series should end with bigger difference from 0 ELO: trial 2
15-03-28 mco nodestime diff
ELO: 7.64 +-13.3 (95%) LOS: 87.0%
Total: 1000 W: 202 L: 180 D: 618
1000 @ 15+0.05 th 1 Poor man noise measuring: standard TC run as a comparison. Series should end with bigger difference from 0 ELO: trial 3
15-03-28 mco nodestime diff
ELO: -2.78 +-13.0 (95%) LOS: 33.8%
Total: 1000 W: 179 L: 187 D: 634
1000 @ 15+0.05 th 1 Poor man noise measuring: standard TC run as a comparison. Series should end with bigger difference from 0 ELO: trial 4
15-03-28 mco nodestime diff
ELO: 7.30 +-13.2 (95%) LOS: 86.0%
Total: 1000 W: 200 L: 179 D: 621
1000 @ 15+0.05 th 1 Poor man noise measuring: standard TC run as a comparison. Series should end with bigger difference from 0 ELO: trial 5
15-03-28 mco nodestime diff
ELO: -5.56 +-13.3 (95%) LOS: 20.6%
Total: 1000 W: 182 L: 198 D: 620
1000 @ 15+0.05 th 1 Poor man noise measuring: standard TC run as a comparison. Series should end with bigger difference from 0 ELO: trial 6
15-03-28 mco nodestime diff
ELO: -8.69 +-14.0 (95%) LOS: 11.2%
Total: 1000 W: 200 L: 225 D: 575
1000 @ 15+0.05 th 1 Poor man noise measuring: standard TC run as a comparison. Series should end with bigger difference from 0 ELO: trial 7
15-03-28 mco nodestime diff
ELO: -9.38 +-13.2 (95%) LOS: 8.2%
Total: 1000 W: 175 L: 202 D: 623
1000 @ 15+0.05 th 1 Poor man noise measuring: standard TC run as a comparison. Series should end with bigger difference from 0 ELO: trial 8
15-03-28 sni quiet2 diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 9356 W: 1760 L: 1841 D: 5755
sprt @ 15+0.05 th 1 Use pawn attacks to help quiet moves ordering. Take 2
15-03-28 n_p SPSA-LMR diff
28787/30000 iterations
58235/60000 games played
60000 @ 15+0.05 th 1 SPSA-session on LMR parameters. Now that several search patches more have passed since Stefan's last tuning of these parameters, another attempt at tuning these is due. Included "cmh-lmr3".
15-03-29 lbr singular diff
LLR: -3.88 (-2.94,2.94) [-1.00,4.00]
Total: 55875 W: 10713 L: 10730 D: 34432
sprt @ 15+0.05 th 1 increase singular margin
15-03-29 mco nodestime diff
ELO: 1.39 +-13.6 (95%) LOS: 57.9%
Total: 1000 W: 201 L: 197 D: 602
1000 @ 30+0.05 th 1 nodestime noise measuring: sample 2