Stockfish Testing Queue

Finished - 45441 tests

16-08-24 pb0 capture_stat diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 13972 W: 2581 L: 2646 D: 8745
sprt @ 10+0.1 th 1 Captures Move heuristic, take 2
16-08-24 SC pawnFields diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 10604 W: 1906 L: 1986 D: 6712
sprt @ 10+0.1 th 1 Bugfix only asymmetry. Forgot to push origin...
16-08-24 SC pawnFields diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 15081 W: 2797 L: 2857 D: 9427
sprt @ 10+0.1 th 1 Bugfix (thanks to Stefan G.). Only score.
16-08-24 sg hanging diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 9543 W: 1794 L: 1878 D: 5871
sprt @ 10+0.1 th 1 Don't prune moves of hanging pieces
16-08-24 tvi id diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 6672 W: 1208 L: 1305 D: 4159
sprt @ 10+0.1 th 1 Another IID variant
16-08-24 pb0 capture_stat2 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 5133 W: 922 L: 1026 D: 3185
sprt @ 10+0.1 th 1 Captures Move heuristic
16-08-24 SC pawnFields diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 14303 W: 2641 L: 2705 D: 8957
sprt @ 10+0.1 th 1 Only score. Take 4.
16-08-24 SC pawnFields diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 9121 W: 1655 L: 1741 D: 5725
sprt @ 10+0.1 th 1 Only asymmetry. Take 3.
16-08-24 Elb opb_mt diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 9495 W: 1291 L: 1414 D: 6790
sprt @ 10+0.1 th 1 Take 2: get rid of popcount to improve performance
16-08-24 eel More_recursiveIID diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 7735 W: 1424 L: 1517 D: 4794
sprt @ 10+0.1 th 1 Have some evidence that recursion starts to work on very long searches, the question is if it is very bad at STC. Loco Loco's version was yellow
16-08-23 sg hanging diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 11784 W: 2166 L: 2241 D: 7377
sprt @ 10+0.1 th 1 try half additional margin
16-08-23 cru eval_space_+800 diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 48401 W: 9072 L: 9060 D: 30269
sprt @ 10+0.1 th 1 Interpolating evaluate_space calls between two yellow STC attempts. +800. (Original idea 22 by S. Nicolet).
16-08-23 luc no_space diff
ELO: -7.44 +-3.0 (95%) LOS: 0.0%
Total: 20000 W: 3797 L: 4225 D: 11978
20000 @ 10+0.1 th 1 How much is space evaluation worth today, at STC? Last time we checked was in 2013 (?) and we're testing many patches...
16-08-23 Voy rcf diff
LLR: -2.96 (-2.94,2.94) [-3.00,1.00]
Total: 28588 W: 5240 L: 5448 D: 17900
sprt @ 10+0.1 th 1 See if we can now remove "correction factor" since rpCapture patch may help balance the scale. Test based off and against rpCapture.
16-08-23 luc tweak_space_score diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 21048 W: 3902 L: 3986 D: 13160
sprt @ 10+0.1 th 1 Since our pawns are now also counted, should now space evaluation decrease more slowly as a function of piece count? Take 1
16-08-23 Elb opb_mt diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 19656 W: 2849 L: 2943 D: 13864
sprt @ 10+0.1 th 1 Another opposite bishops endgame parameter tweak. This one seems to improve a lot of evals for positions that were reported by Mindbreaker1 here: https://groups.google.com/forum/#!topic/fishcooking/vuoEKP2yOCE%5B501-525%5D
16-08-23 SC pawnFields diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 18029 W: 3361 L: 3408 D: 11260
sprt @ 10+0.1 th 1 Take 2 with double effect.
16-08-23 sg hanging diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 6431 W: 1122 L: 1220 D: 4089
sprt @ 10+0.1 th 1 greater futility margin if pieces hanging
16-08-23 SC pawnFields diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 20207 W: 3821 L: 3858 D: 12528
sprt @ 10+0.1 th 1 Evaluate as an asset to control many more squares of the same color than the opponent. Take 1.
16-08-22 Roc DrawHeuristic diff
ELO: -6.74 +-5.2 (95%) LOS: 0.5%
Total: 5000 W: 682 L: 779 D: 3539
5000 @ 60+0.6 th 1 Take 2, 5000 games control at LTC, low throughput.
16-08-23 sg hanging diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 1984 W: 313 L: 430 D: 1241
sprt @ 10+0.1 th 1 no futility pruning if hanging pieces exists
16-08-23 Voy rpCaptureClean diff
LLR: 2.95 (-2.94,2.94) [0.00,5.00]
Total: 8208 W: 1153 L: 1008 D: 6047
sprt @ 60+0.6 th 1 Retest rpCapture as bench changed when not using piece/square in stack.
16-08-23 cru eval_space_+1500 diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 27252 W: 5085 L: 5147 D: 17020
sprt @ 10+0.1 th 1 Try 3: now increase by 1500, two yellow patches for +500 and +1000. (Idea 22 by S. Nicolet.)
16-08-23 SC smoothScaleFactor diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 9632 W: 1726 L: 1810 D: 6096
sprt @ 10+0.1 th 1 Take 3
16-08-23 SC smoothScaleFactor diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 7627 W: 1369 L: 1462 D: 4796
sprt @ 10+0.1 th 1 Take 2.
16-08-23 Elb holes diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 10524 W: 1872 L: 1952 D: 6700
sprt @ 10+0.1 th 1 Take 3: now only give penalty in endgame
16-08-23 Mys tst2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 9558 W: 1511 L: 1596 D: 6451
sprt @ 10+0.1 th 3 reset delta higher for helper threads
16-08-23 Elb holes diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 4099 W: 737 L: 846 D: 2516
sprt @ 10+0.1 th 1 Try the concept of holes in pawn structure (take 2)
16-08-23 Mys tst2 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 14530 W: 2330 L: 2396 D: 9804
sprt @ 10+0.1 th 3 Finally test the inverse
16-08-23 Elb holes diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 7733 W: 1366 L: 1458 D: 4909
sprt @ 10+0.1 th 1 Try the concept of holes in pawn structure
16-08-23 sg hanging diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 14444 W: 2683 L: 2746 D: 9015
sprt @ 10+0.1 th 1 less razoring if hanging pieces exists
16-08-23 luc spaceeval_threshold diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 12820 W: 2371 L: 2484 D: 7965
sprt @ 10+0.1 th 1 Try evaluating space even with less material, take 2 (more than doubled threshold, as suggested by Rocky).
16-08-23 Elb knight_outpost2 diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 20657 W: 3714 L: 3800 D: 13143
sprt @ 10+0.1 th 1 Another attempt at knight outposts
16-08-22 luc spaceeval_threshold diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 26620 W: 4975 L: 4985 D: 16660
sprt @ 10+0.1 th 1 Empty queue... Try evaluating space even with much less material!
16-08-22 lbr tuned diff
LLR: -2.96 (-2.94,2.94) [-3.00,1.00]
Total: 30976 W: 5018 L: 5221 D: 20737
sprt @ 20+0.2 th 1 This was tuned in 20+0.2, and failed STC 10+0.1. would things have been any different running STC in 20+0.2 ?
16-08-22 luc files_space diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 11629 W: 2147 L: 2222 D: 7260
sprt @ 10+0.1 th 1 Add a bonus for pawn structure in space evaluation... Quick and dirty implementation.
16-08-22 tvi yellow diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 43031 W: 8081 L: 8088 D: 26862
sprt @ 10+0.1 th 1 test merging two yellows
16-08-22 Fis pawnAsymmetry diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 14199 W: 2615 L: 2679 D: 8905
sprt @ 10+0.1 th 1 Take 2
16-08-22 Roc DrawHeuristic diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 10123 W: 1953 L: 2035 D: 6135
sprt @ 10+0.1 th 1 Take 2
16-08-22 jhe efficient_rep diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 5839 W: 1060 L: 1161 D: 3618
sprt @ 10+0.1 th 1 Efficient Repetition
16-08-22 cru sn_eval_space_cap diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 96743 W: 18119 L: 17939 D: 60685
sprt @ 10+0.1 th 1 [Try3: increasing value by 1000 as 500 was yellow.] S. Nicolet idea 22: expand space area (changing the cap if necessary), or lower (by steps of 500) the material threshold for which evaluate_space() is called.
16-08-22 Mys tst diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 37398 W: 6869 L: 6897 D: 23632
sprt @ 10+0.1 th 1 Tweak rook psqt to avoid edges where a king is trappable
16-08-22 lan 3-fold_repetition diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 1506 W: 227 L: 347 D: 932
sprt @ 10+0.1 th 1 Bug fix for the 3-fold repetition draw. Test as a Elo-gaining patch
16-08-22 Voy rpCapture diff
LLR: 2.96 (-2.94,2.94) [0.00,5.00]
Total: 8992 W: 1287 L: 1138 D: 6567
sprt @ 60+0.6 th 1 LTC: Take 2
16-08-22 Roc DrawHeuristic diff
ELO: -6.32 +-6.7 (95%) LOS: 3.3%
Total: 2088 W: 195 L: 233 D: 1660
5000 @ 60+0.6 th 1 Experiment: return a draw score earlier in simple KBP, KNP or KP endings when "no progress". Quick LTC run, low throughput.
16-08-22 sg fmh2 diff
LLR: -3.07 (-2.94,2.94) [0.00,5.00]
Total: 18361 W: 3379 L: 3430 D: 11552
sprt @ 10+0.1 th 1 use only half bonus for fmh2 update
16-08-22 pb0 lmr_bad_captures diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 10766 W: 1954 L: 2033 D: 6779
sprt @ 10+0.1 th 1 Don't decrease LMR reduction for captures with a very bad SEE-val (bugfix)
16-08-22 pb0 lmr_bad_captures diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 7138 W: 1279 L: 1374 D: 4485
sprt @ 10+0.1 th 1 Don't decrease LMR reduction for captures with a very bad SEE-val
16-08-22 Elb pawn_dist diff
LLR: -2.94 (-2.94,2.94) [0.00,4.00]
Total: 11204 W: 2003 L: 2121 D: 7080
sprt @ 10+0.1 th 1 Take 2
16-08-21 jos endgame_fix1 diff
LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 157564 W: 29000 L: 29116 D: 99448
sprt @ 10+0.1 th 1 Based on my WDL-statistics remove KRKN and KRKB, and add KBBsK for draw detection of 2 or more bishops on squares of the same color against the lone king. Test as bugfix.