Stockfish Testing Queue

Finished - 40825 tests

15-08-30 hxi symhist diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 26807 W: 4989 L: 4998 D: 16820
sprt @ 15+0.05 th 1 take 2
15-08-30 rko qsearch_bishop_knight diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 2978 W: 503 L: 616 D: 1859
sprt @ 15+0.05 th 1 prune knight x bishpo and equal captures in qsearch
15-08-28 lbr tuned diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 69458 W: 10941 L: 10881 D: 47636
sprt @ 60+0.05 th 1 LTC: tuned ad-hoc eval terms locally (30k iterations in 4+0.04)
15-08-30 rko nullmove_modification diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 6010 W: 1053 L: 1153 D: 3804
sprt @ 15+0.05 th 1 Do nullmove even if eval < beta. second try.
15-08-30 rko razoring_margin diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 14285 W: 2657 L: 2721 D: 8907
sprt @ 15+0.05 th 1 Added razoring margin to return value
15-08-30 rko simplify_razoring diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 7585 W: 1417 L: 1510 D: 4658
sprt @ 15+0.05 th 1 only do razoring if depth == 1 and lower margin
15-08-30 jos bonus1 diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 9230 W: 1681 L: 1807 D: 5742
sprt @ 15+0.05 th 1 Values after 100k iterations.
15-08-30 rko nullmove_modification diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 4254 W: 735 L: 843 D: 2676
sprt @ 15+0.05 th 1 Do nullmove even if eval < beta
15-08-29 Roc PawnAttack2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 26055 W: 4861 L: 4873 D: 16321
sprt @ 15+0.05 th 1 Take 2 (re-re-fix)
15-08-30 Roc WeakBlockedPawns diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 1646 W: 300 L: 423 D: 923
sprt @ 15+0.05 th 1 There is a threat bonus for pieces attacking a weak pawn (not supported by a pawn). Increase the bonus for each pawn which is blocked. Test is run with options: w1=-2 w2=7 w3=2 w4=-1
15-08-28 jos bonus1 diff
ELO: 1.12 +-2.0 (95%) LOS: 86.6%
Total: 49133 W: 10371 L: 10212 D: 28550
40000 @ 7+0.04 th 1 Test some preliminary tuned values.
15-08-27 gli minsplitdepth2 diff
LLR: 2.95 (-2.94,2.94) [0.00,5.00]
Total: 20538 W: 2769 L: 2584 D: 15185
sprt @ 60+0.05 th 7 LTC! Check for scaling at higher threads: reduce minimumsplitdepth to 2. But use fewer slaves: STC
15-08-27 Voy FailHighRefutation diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 59622 W: 11115 L: 10982 D: 37525
sprt @ 15+0.05 th 1 take 2
15-08-29 Roc PawnAttack2 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 10584 W: 1964 L: 2044 D: 6576
sprt @ 15+0.05 th 1 Add one case to the PawnAttackThreat
15-08-26 jki conte diff
ELO: 181.68 +-4.2 (95%) LOS: 100.0%
Total: 20000 W: 11635 L: 2036 D: 6329
20000 @ 15+0.05 th 1 Contempt=20 vs. sf-dd
15-08-26 jki master diff
ELO: 178.87 +-4.0 (95%) LOS: 100.0%
Total: 20000 W: 11128 L: 1654 D: 7218
20000 @ 15+0.05 th 1 master vs. sf-dd
15-08-28 Voy JKhistgrav diff
LLR: -1.10 (-2.94,2.94) [0.00,5.00]
Total: 105 W: 5 L: 51 D: 49
sprt @ 15+0.05 th 1 To get a better understanding of jk's patch: Used the original histgrav, but don't clear the entry.
15-08-27 sg flanking_storm diff
LLR: 2.95 (-2.94,2.94) [0.00,5.00]
Total: 44914 W: 8580 L: 8257 D: 28077
sprt @ 15+0.05 th 1 Bonus for a flanking storming pawns pair like e4/h4 against a white king on g1. Base on an old suggestion from the forum.
15-08-28 jos outpost_bonus diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 8113 W: 1507 L: 1598 D: 5008
sprt @ 15+0.05 th 1 Only assign bonus if outpost piece is protected by a pawn. Final try.
15-08-27 lbr tuned diff
LLR: 2.94 (-2.94,2.94) [0.00,4.00]
Total: 39955 W: 7560 L: 7264 D: 25131
sprt @ 15+0.05 th 1 tuned ad-hoc eval terms locally (30k iterations in 4+0.04)
15-08-27 jos outpost_bonus diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 15731 W: 2966 L: 3023 D: 9742
sprt @ 15+0.05 th 1 Bonus for outposts, if the opponent has no minor piece which can trade with the outpost piece.
15-08-26 Roc KO diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 36060 W: 5660 L: 5645 D: 24755
sprt @ 60+0.05 th 1 LTC
15-08-27 aji minsplitdepth2 diff
LLR: 2.95 (-2.94,2.94) [0.00,5.00]
Total: 2587 W: 485 L: 353 D: 1749
sprt @ 15+0.05 th 7 Check for scaling at higher threads: reduce minimumsplitdepth to 2. But use fewer slaves: STC
15-08-27 jki histgrav3 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 3642 W: 642 L: 753 D: 2247
sprt @ 15+0.05 th 1 History gravity, other way around, 1024
15-08-26 Fis KM2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 24390 W: 4572 L: 4591 D: 15227
sprt @ 15+0.05 th 1 Bonus for knights mobility after two moves.
15-08-26 jki histgrav2 diff
LLR: 2.95 (-2.94,2.94) [0.00,5.00]
Total: 16933 W: 2685 L: 2501 D: 11747
sprt @ 60+0.05 th 1 LTC: History gravity, other way around
15-08-26 qua search_bonus1 diff
LLR: -1.66 (-2.94,2.94) [0.00,5.00]
Total: 20825 W: 3970 L: 3949 D: 12906
sprt @ 15+0.05 th 1 Small Bonus in search if more than 1 good move detected
15-08-26 aji minsplitdepth2 diff
LLR: 2.97 (-2.94,2.94) [0.00,5.00]
Total: 11351 W: 2008 L: 1838 D: 7505
sprt @ 15+0.05 th 3 reduce minimumsplitdepth to 2. But use fewer slaves: STC
15-08-26 Voy FailHighRefutation diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 1924 W: 325 L: 444 D: 1155
sprt @ 15+0.05 th 1 Penalty for prior move that caused a fail high cutoff.
15-08-25 mco lone_split diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 7532 W: 1158 L: 1251 D: 5123
sprt @ 15+0.05 th 7 Attempt with mixed approach: join at split and keep joining also while idle
15-08-25 Fis KMTune diff
35258/25000 iterations
72901/75000 games played
75000 @ 30+0.05 th 1 Tuning for secondary knight mobility.
15-08-24 mco lone_split^ diff
LLR: -5.27 (-2.94,2.94) [0.00,5.00]
Total: 24561 W: 3770 L: 3894 D: 16897
sprt @ 15+0.05 th 7 Very speculative patch: split without allocating! Let the idle threads to 'late join' as soon as possible: take 1
15-08-08 SC capture_tuning diff
74060/75000 iterations
149158/150000 games played
150000 @ 30+0.05 th 1 Reschedule tuning session for piece type dependent rank penalty from most promising values until now. Back then was stopped by icewulf.
15-08-25 jki histgrav2 diff
LLR: 2.96 (-2.94,2.94) [0.00,5.00]
Total: 47270 W: 9083 L: 8748 D: 29439
sprt @ 15+0.05 th 1 History gravity, other way around
15-08-26 mbo extension1 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 17411 W: 3254 L: 3304 D: 10853
sprt @ 15+0.05 th 1 Not exactly a recapture extension because of bugs... but a local test had +13 elo after 5000 games. Something must have been wrong with that test... Low throughput.
15-08-25 Voy MobilityFormula diff
LLR: -2.95 (-2.94,2.94) [-3.00,1.00]
Total: 36735 W: 6835 L: 7058 D: 22842
sprt @ 15+0.05 th 1 Used Excel to best fit mobility using a log equation: a*log(mobSq+1)-b. Doing this will be significantly easier to tune.
15-08-26 Voy cmhWeight diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 11262 W: 2169 L: 2246 D: 6847
sprt @ 15+0.05 th 1 Take 2...
15-08-23 mco ajithcj diff
LLR: 2.96 (-2.94,2.94) [0.00,5.00]
Total: 45369 W: 7465 L: 7166 D: 30738
sprt @ 15+0.05 th 7 Respin ajithcj patch with a more aggressive increase of max slaves per splitpoint. Also tested with 7 threads instead of 9, should be faster.
15-08-25 Voy cmhWeight diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 7133 W: 1328 L: 1423 D: 4382
sprt @ 15+0.05 th 1 Higher CMH weight near the leaves.
15-08-25 Voy SelfAdjBonuses diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 13913 W: 2572 L: 2637 D: 8704
sprt @ 15+0.05 th 1 Take 3
15-08-23 aji max_slaves_splitlevel diff
LLR: -0.87 (-2.94,2.94) [0.00,5.00]
Total: 257 W: 21 L: 55 D: 181
sprt @ 15+0.05 th 15 Dynamically vary max slaves per splitpoint using split level: STC and 16 threads
15-08-24 mco lone_split diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 3119 W: 435 L: 545 D: 2139
sprt @ 15+0.05 th 7 Very speculative patch: split without allocating! Let the idle threads to 'late join' as soon as possible: take 2
15-08-25 Voy lmr diff
LLR: -3.14 (-2.94,2.94) [0.00,5.00]
Total: 10093 W: 1851 L: 1941 D: 6301
sprt @ 15+0.05 th 1 lmr tweak.
15-08-23 jki conte diff
ELO: 1.04 +-3.0 (95%) LOS: 75.0%
Total: 20000 W: 3987 L: 3927 D: 12086
20000 @ 15+0.05 th 1 A quick test for contempt implementation. Contempt = 5cp.
15-08-21 Voy DynamicSplit diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 15963 W: 2566 L: 2626 D: 10771
sprt @ 15+0.05 th 5 Dynamically split threads....similar to LMR.
15-08-25 Roc KO diff
LLR: 2.95 (-2.94,2.94) [0.00,5.00]
Total: 40268 W: 7846 L: 7540 D: 24882
sprt @ 15+0.05 th 1 Bonus for outpost Knight attacking pawn.
15-08-23 jki conte diff
ELO: -3.77 +-3.1 (95%) LOS: 0.8%
Total: 20000 W: 3996 L: 4213 D: 11791
20000 @ 15+0.05 th 1 A quick test for contempt implementation. Contempt = 10cp.
15-08-19 lbr epcastle diff
LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 97183 W: 15120 L: 15115 D: 66948
sprt @ 60+0.05 th 1 LTC: Prune castling. SEE now handles castling correctly, so there is no reason to single out these moves.
15-08-23 jki conte diff
ELO: -1.29 +-3.2 (95%) LOS: 21.7%
Total: 20000 W: 4452 L: 4526 D: 11022
20000 @ 15+0.05 th 1 A quick test for contempt implementation. Contempt = 20cp.
15-08-22 Voy SelfAdjBonuses diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 79696 W: 15049 L: 14826 D: 49821
sprt @ 15+0.05 th 1 Self adjusting bonuses: If a reduce search fails, adj bonuses if history or cmh <0.