Stockfish Testing Queue

Finished - 46941 tests

15-09-30 Voy FHrevisit diff
LLR: -3.12 (-2.94,2.94) [0.00,5.00]
Total: 60726 W: 11060 L: 10933 D: 38733
sprt @ 15+0.05 th 1 Last shot at this idea. I believe that due to very aggressive move count pruning, it is much easier to "fail high" at low depths.
15-09-30 Mys KS5.3 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 47061 W: 8649 L: 8572 D: 29840
sprt @ 15+0.05 th 1 King safety and minimum pawn distance in endgame (revisit)
15-09-30 Voy FHrevisit diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 49789 W: 9089 L: 9001 D: 31699
sprt @ 15+0.05 th 1 Fixed version...
15-09-30 IIv reduction_tune diff
9856/10000 iterations
20000/20000 games played
20000 @ 30+0.05 th 1 Tuning first 6 move - session 2. Read more under comments.
15-09-30 Roc SkipThreats2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 11076 W: 1979 L: 2057 D: 7040
sprt @ 15+0.05 th 1 ST2_20150926_2
15-09-29 Voy FHrevisit diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 80734 W: 12175 L: 12000 D: 56559
sprt @ 60+0.05 th 1 LTC: This test came close to passing LTC...try a couple small tweaks...(fix bench)
15-09-30 Roc ThreatSimplified diff
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 19146 W: 2950 L: 2827 D: 13369
sprt @ 60+0.05 th 1 Fixed SPRT bound [-3,1] LTC
15-09-29 Roc ThreatSimplified diff
LLR: -0.08 (-2.94,2.94) [0.00,5.00]
Total: 71 W: 11 L: 14 D: 46
sprt @ 60+0.05 th 1 Simplification test: remove Q threats in evaluate-threats. LTC
15-09-29 Roc ThreatSimplified diff
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 13627 W: 2607 L: 2473 D: 8547
sprt @ 15+0.05 th 1 Simplification test: remove Q threats in evaluate-threats.
15-09-29 Voy ST diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 17110 W: 3087 L: 3139 D: 10884
sprt @ 15+0.05 th 1 Stat Tweak Idea.
15-09-29 Mys DP diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 26242 W: 4713 L: 4780 D: 16749
sprt @ 15+0.05 th 1 I would like to try a doubled tweak, which focuses on pawns where our castled king would be.
15-09-27 mco locking diff
LLR: 3.48 (-2.94,2.94) [-3.00,1.00]
Total: 47706 W: 7390 L: 7283 D: 33033
sprt @ 15+0.05 th 7 Regression test simplified locking scheme. This is not enough, also torture test with 20 core machine is necessary to verify we don't hang.
15-09-29 sg ext_check diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 13878 W: 2495 L: 2561 D: 8822
sprt @ 15+0.05 th 1 extend capture checks for depth < 4. Inspired by joergoster .
15-09-29 Roc CB2 diff
ELO: -56.31 +-22.3 (95%) LOS: 0.0%
Total: 417 W: 60 L: 127 D: 230
20000 @ 9+0.05 th 1 Quickly checked tuned parameters. See if the idea to remove "attackunits" checks (and replace with scores) have a chance against current master.
15-09-28 mbo lazy_smp2 diff
ELO: 8.76 +-5.0 (95%) LOS: 100.0%
Total: 5000 W: 735 L: 609 D: 3656
5000 @ 60+0.05 th 7 New version of Lazy SMP. 7 Threads. LTC.
15-09-29 sg ext_check diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 21905 W: 3989 L: 4020 D: 13896
sprt @ 15+0.05 th 1 extend capture checks always
15-09-29 Fis deltaIncrement diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 15503 W: 2768 L: 2872 D: 9863
sprt @ 15+0.05 th 1 Take 2 final try
15-09-29 Roc ThreatSimplified diff
ELO: 1.08 +-4.1 (95%) LOS: 69.5%
Total: 10000 W: 1855 L: 1824 D: 6321
10000 @ 15+0.05 th 1 Let's see if any gain if remove Q threats in evaluate-threats. Quick 10M test before trying a simplification
15-09-28 IIv reduction_tune diff
9928/10000 iterations
20000/20000 games played
20000 @ 30+0.05 th 1 Tuning first 6 moves. All tests were quite neutral, but promising because different functions give similar results. That's why I'm starting tuning 6 by 6 moves, and then 6 by 6 depths, starting from default values, until some progress will be noticeable.
15-09-28 Voy FHrevisit diff
LLR: 2.95 (-2.94,2.94) [0.00,5.00]
Total: 65529 W: 12166 L: 11759 D: 41604
sprt @ 15+0.05 th 1 This test came close to passing LTC...try a couple small tweaks...(fix bench)
15-09-29 Roc OnceUPawnAKnight diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 13600 W: 2458 L: 2525 D: 8617
sprt @ 15+0.05 th 1 When more pawns and less pieces, the Knight is more effective
15-09-29 Roc IndividualThreats diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 3332 W: 572 L: 684 D: 2076
sprt @ 15+0.05 th 1 IT_20150927_4
15-09-28 sni good_night2 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 26248 W: 4786 L: 4799 D: 16663
sprt @ 15+0.05 th 1 Remove blocked pawns condition
15-09-28 mco tune_pawn_cnt diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 20509 W: 3731 L: 3768 D: 13010
sprt @ 15+0.05 th 1 Test pawn endgame scaling tuned values after 80K games
15-09-28 mbo lazy_smp2 diff
ELO: 16.76 +-5.4 (95%) LOS: 100.0%
Total: 5000 W: 899 L: 658 D: 3443
5000 @ 15+0.05 th 7 New version of Lazy SMP. 7 Threads.
15-09-28 IIv reduction_check diff
ELO: -2.80 +-6.0 (95%) LOS: 18.2%
Total: 4594 W: 810 L: 847 D: 2937
5000 @ 15+0.05 th 1 A quick elo check of tuned linear model for the first 12 moves.
15-09-28 Voy FHrevisit diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 22984 W: 4143 L: 4170 D: 14671
sprt @ 15+0.05 th 1 Take 2: Based on passed tests...I believe that history gravity discontinuity (at depth 23) may cause the regression at LTC
15-09-27 mco tune_pawn_cnt diff
39522/10000 iterations
79825/80000 games played
80000 @ 30+0.05 th 1 Tune pawn count scaling
15-09-26 Mys CB diff
LLR: 3.20 (-2.94,2.94) [0.00,5.00]
Total: 52518 W: 8107 L: 7782 D: 36629
sprt @ 60+0.05 th 1 Scoring checks - Respin of an almost successful previous patch, sizable bonuses.
15-09-27 Fis 4men diff
ELO: 2.42 +-2.4 (95%) LOS: 97.7%
Total: 30000 W: 5607 L: 5398 D: 18995
30000 @ 15+0.05 th 1 Test elo of internal 3 & 4 men syzygy. STC (Fixed Linux compile.)
15-09-28 sni good_knight diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 19434 W: 3494 L: 3537 D: 12403
sprt @ 15+0.05 th 1 Take 2, bonus=10
15-09-27 IIv reduction_tune diff
9898/10000 iterations
20000/20000 games played
20000 @ 30+0.05 th 1 Tuning first 12 moves. After this session, a quick elo check will show should I continue or restart tuning from default (log) values.
15-09-26 sg passed_pawns diff
LLR: 2.97 (-2.94,2.94) [0.00,5.00]
Total: 32301 W: 5101 L: 4858 D: 22342
sprt @ 60+0.05 th 1 LTC: try tuned values. The are only small changes, but lets see. Perhaps a further tuning session is needed.
15-09-27 aji ybwx_ext1 diff
ELO: 0.12 +-6.7 (95%) LOS: 51.3%
Total: 3000 W: 437 L: 436 D: 2127
3000 @ 15+0.05 th 7 quick sanity check : STC see comments for more info
15-09-28 Voy lmr diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 4818 W: 814 L: 919 D: 3085
sprt @ 15+0.05 th 1 Take 2
15-09-28 Voy lmr diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 15295 W: 2736 L: 2796 D: 9763
sprt @ 15+0.05 th 1 Hopefully smarter logic for lmr.
15-09-27 Voy lmrDouble diff
LLR: -2.08 (-2.94,2.94) [0.00,5.00]
Total: 5231 W: 920 L: 986 D: 3325
sprt @ 15+0.05 th 1 More reduction if history stats are very low.
15-09-27 sni good_knight diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 6213 W: 1118 L: 1217 D: 3878
sprt @ 15+0.05 th 1 Bonus for good knight with blocked pawns. Take 1, bonus=25
15-09-27 Voy Checkers diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 11454 W: 2037 L: 2114 D: 7303
sprt @ 15+0.05 th 1 Increase Hstats for checkers.
15-09-27 jos null1 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 8556 W: 1457 L: 1546 D: 5553
sprt @ 15+0.05 th 1 Make sure at least some moves are made at higher iterations before allowing to do a null move.
15-09-26 mbo lazy_smp diff
ELO: -7.55 +-5.0 (95%) LOS: 0.1%
Total: 4879 W: 589 L: 695 D: 3595
5000 @ 60+0.05 th 7 Lazy SMP. Estimate strength for 7 Threads LTC.
15-09-27 mbo lazy_smp2 diff
ELO: -13.52 +-6.5 (95%) LOS: 0.0%
Total: 3625 W: 535 L: 676 D: 2414
5000 @ 15+0.05 th 3 New version of Lazy SMP.
15-09-27 sni weights diff
LLR: -2.66 (-2.94,2.94) [0.00,4.00]
Total: 21913 W: 3989 L: 4055 D: 13869
sprt @ 15+0.05 th 1 Increase KingSafety weight
15-09-27 Fis 4men diff
ELO: 4.70 +-9.2 (95%) LOS: 84.0%
Total: 1996 W: 381 L: 354 D: 1261
30000 @ 15+0.05 th 1 Test elo of internal 3 & 4 men syzygy. STC
15-09-27 jos threat_per_piece diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 4457 W: 750 L: 856 D: 2851
sprt @ 15+0.05 th 1 Evaluate threats per piece type, and not only for Minor/Major. Pre-tuned values.
15-09-27 Voy SmarterLMR diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 4100 W: 716 L: 824 D: 2560
sprt @ 15+0.05 th 1 Take 2...
15-09-19 sni king_separation diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 110885 W: 16959 L: 16670 D: 77256
sprt @ 60+0.05 th 1 LTC: Try king separation bonus, take 2
15-09-25 mbo lazy_smp diff
ELO: -3.34 +-5.5 (95%) LOS: 11.7%
Total: 5000 W: 793 L: 841 D: 3366
5000 @ 15+0.05 th 7 Get another rough estimate for Lazy SMP. I'd like to see how it scales to 7 threads. If the results are much worse, we can abort the test.
15-09-26 Voy SmarterLMR diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 14300 W: 2557 L: 2621 D: 9122
sprt @ 15+0.05 th 1 CMH and History has their own average move score to assist in reduction.
15-09-27 Roc IndividualThreats diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 4835 W: 853 L: 958 D: 3024
sprt @ 15+0.05 th 1 Add individual threats instead of regrouping by Minor or Major.