Stockfish Testing Queue

Finished - 37193 tests

15-05-12 Mys KS2 diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 11259 W: 2099 L: 2175 D: 6985
sprt @ 15+0.05 th 1 Squares influenced by our king and not attacked by the enemy are ok.
15-05-11 sg remove_gains diff
LLR: -0.39 (-2.94,2.94) [-3.00,1.00]
Total: 30000 W: 5660 L: 5734 D: 18606
sprt @ 15+0.05 th 1 Simplification: Remove gains
15-05-11 sto mouse_dont_escape diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 16049 W: 3008 L: 3071 D: 9970
sprt @ 15+0.05 th 1 Use 12% more time when ahead by 76 to 153 cp.
15-05-11 Fis behindTime diff
LLR: -3.62 (-2.94,2.94) [-1.50,4.50]
Total: 31529 W: 5874 L: 5920 D: 19735
sprt @ 15+0.05 th 1 Use up to an additional 20% more time if we are 100 centipawns or more behind.
15-05-10 Voy MaxHistory diff
LLR: 3.43 (-2.94,2.94) [-1.50,4.50]
Total: 34319 W: 6565 L: 6352 D: 21402
sprt @ 15+0.05 th 1 From 250 to 175.
15-05-11 SC score_evasions_6 diff
ELO: -3.23 +-3.0 (95%) LOS: 1.9%
Total: 20000 W: 3906 L: 4092 D: 12002
20000 @ 10+0.05 th 1 Try different formulas for evasions scoring. Take 6: use regression formula instead of MVV/LVA for good captures.
15-05-11 SC score_evasions_7 diff
ELO: 2.02 +-3.1 (95%) LOS: 90.1%
Total: 20000 W: 4134 L: 4018 D: 11848
20000 @ 10+0.05 th 1 Try different formulas for evasions scoring. Take 7: use regression formula based on current m.value().
15-05-11 mco matefind diff
LLR: -1.63 (-2.94,2.94) [-3.00,1.00]
Total: 10590 W: 1614 L: 1714 D: 7262
sprt @ 60+0.05 th 1 LTC: Another attempt at improving matefind without introducing regressions. This time set depth at 5 plies to see if can be a regression at tournament TC.
15-05-11 jhe mf_new diff
ELO: -8.55 +-3.0 (95%) LOS: 0.0%
Total: 20000 W: 3580 L: 4072 D: 12348
20000 @ 15+0.05 th 1 Better mate threat detection in null move (testable at low TC).
15-05-11 Fis TTflipEval2 diff
LLR: -4.27 (-2.94,2.94) [-1.50,4.50]
Total: 13001 W: 2422 L: 2540 D: 8039
sprt @ 15+0.05 th 1 Probe the TT for an eval() of a position with flipped side to move. 2MB Take 2
15-05-10 SC score_evasions diff
ELO: -10.90 +-3.1 (95%) LOS: 0.0%
Total: 20000 W: 3768 L: 4395 D: 11837
20000 @ 10+0.05 th 1 Try different formulas for evasions scoring. Take 5: decide whether a capture is bad based on magic formula rather than pos.see_sign(m). This is the most promising!
15-05-10 sg gains diff
LLR: 3.12 (-2.94,2.94) [-1.50,4.50]
Total: 33434 W: 6437 L: 6237 D: 20760
sprt @ 15+0.05 th 1 Tweak gains update formula
15-05-10 SC score_evasions diff
ELO: -42.56 +-3.1 (95%) LOS: 0.0%
Total: 20000 W: 2973 L: 5411 D: 11616
20000 @ 10+0.05 th 1 Try different formulas for evasions scoring. Take 4: never use pos.see(m) for captures.
15-05-10 mco matefind diff
LLR: 3.02 (-2.94,2.94) [-3.00,1.00]
Total: 18787 W: 3078 L: 2951 D: 12758
sprt @ 60+0.05 th 1 LTC: Another attempt at improving matefind without introducing regressions
15-05-10 cru LinQuadTuned diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 6935 W: 1299 L: 1387 D: 4249
sprt @ 15+0.05 th 1 STC test of tuned material imbalance parameters
15-05-10 sni connected diff
LLR: -3.36 (-2.94,2.94) [-1.50,4.50]
Total: 4679 W: 847 L: 955 D: 2877
sprt @ 15+0.05 th 1 Same values for endgame and midgame in connected array. Take 2 : use the old midgame values, not the old endgame values.
15-05-10 Voy Betta diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 19025 W: 3743 L: 3798 D: 11484
sprt @ 15+0.05 th 1 Small pruning tweak, since cmh is not being cleared.
15-05-10 sni connected diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 9993 W: 1792 L: 1871 D: 6330
sprt @ 15+0.05 th 1 Equal values for endgame and midgame in connected array.
15-05-10 mco matefind diff
ELO: 1.75 +-3.0 (95%) LOS: 87.7%
Total: 20000 W: 3830 L: 3729 D: 12441
20000 @ 15+0.05 th 1 Another attempt at improving matefind without introducing regressions
15-05-10 Hai improving diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 17180 W: 3169 L: 3229 D: 10782
sprt @ 15+0.05 th 1 First two ply are always 'improving'
15-05-07 Voy Anglerfish(Again) diff
LLR: -2.87 (-2.94,2.94) [-1.00,4.00]
Total: 99084 W: 18861 L: 18721 D: 61502
sprt @ 15+0.05 th 1 I now have to retest patch again...
15-05-03 hxi imbal diff
LLR: -2.96 (-2.94,2.94) [-3.00,1.00]
Total: 52470 W: 8360 L: 8595 D: 35515
sprt @ 60+0.05 th 1 LTC: imbalance vs psqt
15-05-10 vin null_king diff
ELO: -8.13 +-3.1 (95%) LOS: 0.0%
Total: 18771 W: 3351 L: 3790 D: 11630
20000 @ 15+0.05 th 1 Quick measure of the effect of importing just the "potential threat" code from Stockfish MF. Does it cost a lot of Elo or is it a possible base for experiment?
15-05-10 Mys KS diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 6041 W: 1129 L: 1219 D: 3693
sprt @ 15+0.05 th 1 Similar to attack units, defending units should mitigate part of the attack.
15-05-09 lbr decay diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 36569 W: 5807 L: 5789 D: 24973
sprt @ 60+0.05 th 1 LTC: decay gains between moves
15-05-09 cru linear_quadratic_parame diff
11394/10000 iterations
20000/20000 games played
20000 @ 15+0.05 th 1 Stopped earlier tuning to fix a mistake in SetRange. Few parameters remaining to tune. Planning to add tuned diagonals to Linear.
15-05-10 Voy RainbowTrout(fixed) diff
LLR: -0.06 (-2.94,2.94) [-1.50,4.50]
Total: 28 W: 4 L: 6 D: 18
sprt @ 15+0.05 th 1 Add a random value (-Depth/2, Depth/2) to our history bonuses. The idea here is to prevent the search being "stuck" in a tree subset. Using MC method may encourage the search to look at more interesting paths.
15-05-10 lbr history diff
LLR: -2.95 (-2.94,2.94) [-0.50,4.50]
Total: 3265 W: 592 L: 708 D: 1965
sprt @ 15+0.05 th 1 double history max. now that we don't clear stats, maybe it's better?
15-05-09 Voy BloodfinTetra diff
LLR: -1.39 (-2.94,2.94) [-1.50,4.50]
Total: 5103 W: 955 L: 991 D: 3157
sprt @ 15+0.05 th 1 Add extra bonus to history stats if prior move was a capture.
15-05-09 sni good_knight_bad_bishop diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 20561 W: 3882 L: 3933 D: 12746
sprt @ 15+0.05 th 1 Use blocked pawns to detect good knights and bad bishops
15-05-09 Roc SafeCheckBonus diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 10002 W: 1862 L: 1941 D: 6199
sprt @ 15+0.05 th 1 Bonus on safe slider distant check when King must move.
15-05-09 Voy RainbowTrout diff
LLR: -0.15 (-2.94,2.94) [-1.50,4.50]
Total: 75 W: 10 L: 15 D: 50
sprt @ 15+0.05 th 1 Add a random value (-Depth, Depth) to our history bonuses. The idea here is to prevent the search being "stuck" in a tree subset. Using MC method will encourage the search to look at more interesting paths.
15-05-09 cru linear_quadratic_parame diff
501/10000 iterations
1009/20000 games played
20000 @ 15+0.05 th 1 Most parameters have converged, some to values in excess of 20% different from old values. Five parameters are still drifting, so tuning these last five.
15-05-09 Roc BishopMobV4 diff
LLR: -2.97 (-2.94,2.94) [0.00,6.00]
Total: 10614 W: 1669 L: 1722 D: 7223
sprt @ 60+0.05 th 1 Opponent Bishop long-term mobility is penalized by same color pawns. This test considers our wedges (example c3-d4-e3) as an additional long term structural hurdle (for example, for a Bg7). Based on an idea by Fauzi.
15-05-09 lbr decay^ diff
LLR: -2.96 (-2.94,2.94) [-0.50,4.50]
Total: 25526 W: 4899 L: 4935 D: 15692
sprt @ 15+0.05 th 1 decay history between moves
15-05-08 Voy Razorfish diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 38949 W: 7471 L: 7473 D: 24005
sprt @ 15+0.05 th 1 Only clear stats at captures, always clear History stats. I believe this will be much stronger than Anglerfish...
15-05-09 lbr decay diff
LLR: 2.96 (-2.94,2.94) [-0.50,4.50]
Total: 20745 W: 4116 L: 3914 D: 12715
sprt @ 15+0.05 th 1 decay gains between moves
15-05-08 cru linear_quadratic_parame diff
11931/10000 iterations
21000/20000 games played
20000 @ 15+0.05 th 1
15-05-08 SC MVV_evasions diff
ELO: -12.81 +-3.1 (95%) LOS: 0.0%
Total: 20000 W: 3684 L: 4421 D: 11895
20000 @ 10+0.05 th 1 Try different formulas for evasions scoring. Take 3: pos.see().
15-05-08 SC MVV_evasions diff
ELO: -16.38 +-3.1 (95%) LOS: 0.0%
Total: 20000 W: 3631 L: 4573 D: 11796
20000 @ 10+0.05 th 1 Try different formulas for evasions scoring. Take 2: history with LVA correction.
15-05-08 SC MVV_evasions diff
ELO: -11.24 +-3.1 (95%) LOS: 0.0%
Total: 20000 W: 3709 L: 4356 D: 11935
20000 @ 10+0.05 th 1 Try different formulas for evasions scoring. Take 1: same formula as for captures.
15-05-08 Voy Razorfish2 diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 11030 W: 2056 L: 2132 D: 6842
sprt @ 15+0.05 th 1 Never clear stats, except for history.
15-05-08 jos RookOnAdjacentOpenFile diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 15719 W: 2956 L: 3020 D: 9743
sprt @ 15+0.05 th 1 Final try with slightly increased mg bonus, no eg bonus and fixed logic.
15-05-08 Mys rook_ks diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 8914 W: 1630 L: 1712 D: 5572
sprt @ 15+0.05 th 1 Wondering if without enemy rooks is our king actually safer as well?
15-05-07 aji kingside_safe_checks diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 26635 W: 5161 L: 5195 D: 16279
sprt @ 15+0.05 th 1 Add bonus only for kingside safe checks when attackercount=0: STC
15-05-07 cru linear_quadratic_parame diff
12536/10000 iterations
20000/20000 games played
20000 @ 15+0.05 th 1 1. Parameters that did not move or moved slightly are fixed and not tuned; 2. parameters that moved by a lot are re-tuned with an extrapolated starting value; 3. Parameters that were zeros before (i.e., not tuned) are now tuned.
15-05-07 jos RookOnAdjacentOpenFile diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 23340 W: 4450 L: 4494 D: 14396
sprt @ 15+0.05 th 1 Bonus only for file a or file h, also lower endgame bonus. Take 2.
15-05-08 Hai capture_score diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 9737 W: 1879 L: 1959 D: 5899
sprt @ 15+0.05 th 1 Take 2 on Rocky640's king square idea
15-05-08 sni bishop3 diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 5600 W: 1084 L: 1176 D: 3340
sprt @ 15+0.05 th 1 Penalty for each blocked pawn on the same color as one of our bishop.
15-05-07 SC MVV_evasions diff
LLR: -2.96 (-2.94,2.94) [-3.00,1.00]
Total: 14748 W: 2809 L: 2996 D: 8943
sprt @ 15+0.05 th 1 1) In scoring evasions value of captured piece should not play any role, so MVV does not apply (you can always just capture one piece) 2) LVA is (according to Lucas) already taken into account by move generator and stable sorting. Maybe is possible to brutally simplify evasions scoring.