Stockfish Testing Queue

Finished - 33316 tests

15-12-15 Voy spt' diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 17279 W: 3218 L: 3315 D: 10746
sprt @ 10+0.1 th 1 ver. 2
15-12-15 mbo norazor diff
LLR: -2.95 (-2.94,2.94) [-3.00,1.00]
Total: 4184 W: 701 L: 866 D: 2617
sprt @ 10+0.1 th 1 Remove razoring.
15-12-15 Voy sp diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 3906 W: 673 L: 782 D: 2451
sprt @ 10+0.1 th 1 Don't SEE prune if main killer move is played first.
15-12-15 Kal new_space diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 15574 W: 2925 L: 2983 D: 9666
sprt @ 10+0.1 th 1 evaluate_space only triggers with >5 pawns on starting squares
15-12-14 Roc nttune diff
42713/50000 iterations
100000/100000 games played
100000 @ 20+0.2 th 1 Let's see if we can still improve on this, tuning at 20+0.2
15-12-14 pec tuned_nfl1 diff
LLR: -2.94 (-2.94,2.94) [0.00,4.00]
Total: 13217 W: 2447 L: 2558 D: 8212
sprt @ 10+0.1 th 1 tuned parameters
15-12-14 pec tune_nfl diff
24082/20000 iterations
40000/40000 games played
40000 @ 20+0.2 th 1 One more tuning session from place where last one ended now with a little longer tc to see if it is possible to get some gain. In last session MaxRatio and FL were consitently decreasing, so I was wondering if they were getting to better local optimum
15-12-14 Roc HangingType diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 17295 W: 3152 L: 3203 D: 10940
sprt @ 10+0.1 th 1 Using different scores for Hanging pawns or pieces
15-12-14 Voy pr diff
LLR: -2.92 (-2.94,2.94) [0.00,5.00]
Total: 9592 W: 1755 L: 1838 D: 5999
sprt @ 10+0.1 th 1 Apply penalty refutation if prior move is main killer. (fixed bench)
15-12-14 jos TW1 diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 19751 W: 3719 L: 3807 D: 12225
sprt @ 10+0.1 th 1 Increased endgame weights for Mobility and Pawn Structure. Testrun for lantonov's modified SPSA tuner.
15-12-13 lbr pins diff
LLR: -2.96 (-2.94,2.94) [1.00,5.00]
Total: 44939 W: 6382 L: 6343 D: 32214
sprt @ 60+0.6 th 1 Failed yellow at STC. But this goes towards increasing king safety score, which has often been seen to scale. so give it a chance at LTC, with low throughput. This is cheating statistics, however. So, use stricter bounds (1,5) for LTC to compensate. Stricter bounds will also make it fail faster/cheaper, if it doesn't have the assumed scaling property.
15-12-13 IIv rft_tune diff
39141/40000 iterations
80000/80000 games played
80000 @ 20+0.2 th 1 This tuning is inspired by recent VoyagerOne's patches which were successful on STC, but unsuccessful on LTC.
15-12-14 Roc SimplifyAligned diff
LLR: -2.96 (-2.94,2.94) [-3.00,1.00]
Total: 33601 W: 6411 L: 6631 D: 20559
sprt @ 10+0.1 th 1 Simplify the rook aligned bonus
15-12-10 Roc TestNewThreat diff
LLR: -1.49 (-2.94,2.94) [0.00,4.00]
Total: 109058 W: 15765 L: 15532 D: 77761
sprt @ 60+0.6 th 1 LTC (60+0.6) [0,4] (sorry !) for tuned parameters after 117000 games
15-12-13 Roc TestNewThreat diff
LLR: 2.95 (-2.94,2.94) [0.00,4.00]
Total: 21046 W: 3200 L: 2989 D: 14857
sprt @ 60+0.6 th 1 Since the 1st green test is struggling at LTC, LTC on the 2nd green test (tuned + 10% on eg threats) Last try on this,
15-12-13 Voy hppd' diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 4018 W: 715 L: 860 D: 2443
sprt @ 10+0.1 th 1 May have to reduce the depth a bit.
15-12-13 Voy hppd diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 2918 W: 491 L: 639 D: 1788
sprt @ 10+0.1 th 1 Use predicted depth for history pruning.
15-12-13 lbr pins diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 79706 W: 15327 L: 15202 D: 49177
sprt @ 10+0.1 th 1 pins
15-12-13 sni outpost17 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 6694 W: 1245 L: 1342 D: 4107
sprt @ 10+0.1 th 1 Double outpost value if opponent has pawn(s) behind or on the side of the square
15-12-13 sni outpost17 diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 5985 W: 1111 L: 1211 D: 3663
sprt @ 10+0.1 th 1 Double outpost value if opponent has pawn(s) behind the square
15-12-13 sni outpost16 diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 38130 W: 7294 L: 7252 D: 23584
sprt @ 10+0.1 th 1 Prefer outposts near the opponent king (take 2, same bonus for mg and eg)
15-12-13 IIv PV_vs_nonPV diff
LLR: -0.08 (-2.94,2.94) [0.00,4.00]
Total: 329 W: 38 L: 41 D: 250
sprt @ 60+0.6 th 1 LTC (with at most 128K games)
15-12-12 Voy hpfi diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 59220 W: 11550 L: 11411 D: 36259
sprt @ 10+0.1 th 1 One more shot...
15-12-10 Roc TestNewThreat diff
LLR: 2.96 (-2.94,2.94) [0.00,4.00]
Total: 45239 W: 8913 L: 8591 D: 27735
sprt @ 10+0.1 th 1 Before going LTC, use green test values, but increase eg threats by 10%
15-12-13 IIv PV_vs_nonPV diff
LLR: 2.95 (-2.94,2.94) [0.00,4.00]
Total: 11768 W: 2347 L: 2147 D: 7274
sprt @ 10+0.1 th 1 This small change, which is still 5 * (best - start) from the tuning session, is a check whether is this close to optimum.
15-12-12 lbr safety diff
LLR: -2.96 (-2.94,2.94) [-3.00,1.00]
Total: 10058 W: 1501 L: 1669 D: 6888
sprt @ 60+0.6 th 1 simplify king safety: last try
15-12-12 Roc MidgameInitiative diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 11520 W: 2245 L: 2321 D: 6954
sprt @ 10+0.1 th 1 Take 1
15-12-12 Voy hpf3 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 54390 W: 10467 L: 10352 D: 33571
sprt @ 10+0.1 th 1 Last shot at this...
15-12-12 sni outpost16 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 13685 W: 2632 L: 2698 D: 8355
sprt @ 10+0.1 th 1 Prefer outposts near the opponent king
15-12-12 pb0 distinct_iter_paths7 diff
LLR: 2.95 (-2.94,2.94) [0.00,5.00]
Total: 2493 W: 462 L: 331 D: 1700
sprt @ 5+0.1 th 20 test with many threads
15-12-12 lbr safety diff
LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 14750 W: 3027 L: 2894 D: 8829
sprt @ 10+0.1 th 1 simplify king safety: last try
15-12-10 sni bishops4 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 29071 W: 5606 L: 5603 D: 17862
sprt @ 10+0.1 th 1 Opposite bishop scaling by number of pieces
15-12-12 jos dangerous^ diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 10690 W: 2020 L: 2099 D: 6571
sprt @ 10+0.1 th 1 Less LMR for dangerous moves.
15-12-11 lbr tune diff
37162/40000 iterations
76538/80000 games played
80000 @ 20+0.2 th 1 forgot to call Eval::init() after ucinewgame
15-12-12 Fis interpolateSpace diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 7715 W: 1450 L: 1543 D: 4722
sprt @ 10+0.1 th 1 Phase out space evaluation gradually.
15-12-11 pb0 distinct_iter_paths7 diff
LLR: 2.95 (-2.94,2.94) [0.00,5.00]
Total: 19739 W: 2910 L: 2721 D: 14108
sprt @ 20+0.2 th 7 LTC, Distinct iteration paths approach with generalized half-density formula supporting up to 128 threads.
15-12-11 IIv PV_tune diff
19557/20000 iterations
40000/40000 games played
40000 @ 20+0.2 th 1 Tuning almost passed patch (PV_vs_nonPV).
15-12-12 Voy hpi diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 42384 W: 8155 L: 8093 D: 26136
sprt @ 10+0.1 th 1 reduce history pruning depth if improving.
15-12-11 pec tuned_nfl diff
LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 37231 W: 5333 L: 5236 D: 26662
sprt @ 60+0.6 th 1 LTC. Simplification(removal of hard stops ) + ponder off bug fix. rescheduled at 60+0.6
15-12-11 Voy PSQTt2 diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 35218 W: 6740 L: 6773 D: 21705
sprt @ 10+0.1 th 1 Reduce all psqt EG values by 10% except for King and Rook tables.
15-12-09 IIv PV_vs_nonPV2 diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 63957 W: 12172 L: 12104 D: 39681
sprt @ 10+0.1 th 1 Last try before tuning this idea.
15-12-03 IIv PV_vs_nonPV diff
LLR: -2.94 (-2.94,2.94) [0.00,4.00]
Total: 212885 W: 31218 L: 30758 D: 150909
sprt @ 60+0.4 th 1 Trying the average between master and its neutral counterpart: LTC.
15-12-10 Fis tuneSpace diff
24639/25000 iterations
50000/50000 games played
50000 @ 10+0.1 th 1 Phase out space evaluation gradually. Tune.
15-12-11 Voy hpf2 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 19850 W: 2880 L: 2929 D: 14041
sprt @ 60+0.6 th 1 Use Proper LTC
15-12-07 Fis fixMultiPV diff
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 55460 W: 9665 L: 9601 D: 36194
sprt @ 5+0.1 th 7 See https://github.com/official-stockfish/Stockfish/pull/519
15-12-11 Voy PSQTt' diff
LLR: -1.79 (-2.94,2.94) [0.00,4.00]
Total: 4582 W: 841 L: 921 D: 2820
sprt @ 10+0.1 th 1 Reduce King's EG values by 10%.
15-12-10 pb0 distinct_iter_paths7 diff
LLR: 2.96 (-2.94,2.94) [0.00,5.00]
Total: 6026 W: 1047 L: 901 D: 4078
sprt @ 5+0.1 th 7 Distinct iteration paths approach with generalized half-density formula supporting up to 128 threads.
15-12-10 sni tune_bishops5 diff
9901/10000 iterations
20000/20000 games played
20000 @ 10+0.1 th 1 Tuning for opposite bishops scale factor (fixed bench)
15-12-08 Voy hpf2 diff
LLR: -1.91 (-2.94,2.94) [0.00,5.00]
Total: 41908 W: 5568 L: 5506 D: 30834
sprt @ 100+1 th 1 LTC: Take 2. Less aggressive pruning at higher rootDepth.
15-12-11 IIv low_depths_lmr diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 7038 W: 1270 L: 1366 D: 4402
sprt @ 10+0.1 th 1 Checking for progress after tuning.