Stockfish Testing Queue

Finished - 1419 tests

19-06-01 31m initatk1 diff
LLR: -2.94 (-2.94,2.94) [0.50,4.50]
Total: 11342 W: 2458 L: 2566 D: 6318
sprt @ 10+0.1 th 1 I'm curious what happens to @xoto's initatk1 if we (substantially) increase the effect size.
19-06-01 31m BishopConnectivity diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 7901 W: 1717 L: 1843 D: 4341
sprt @ 10+0.1 th 1 Endgame penalty S(0, -16) if our threats against enemy pieces are offset by an active bishop--increasing penalty for the more pieces this bishop is defending. (Effectively this is like an endgame-only Overload penalty, but of opposite effect.)
19-05-31 31m tweak_pawns diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 50017 W: 11116 L: 11082 D: 27819
sprt @ 10+0.1 th 1 Tweak the three pawn penalties Backward, Doubled, and Isolated.
19-05-31 31m tweak_pawns diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 20030 W: 4447 L: 4532 D: 11051
sprt @ 10+0.1 th 1 Take 2.
19-05-31 31m tweak_kd diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 22996 W: 5111 L: 5184 D: 12701
sprt @ 10+0.1 th 1 Attempt to polish king danger coefficients a bit.
19-05-31 31m tweak_kd diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 10415 W: 2303 L: 2427 D: 5685
sprt @ 10+0.1 th 1 Another attempt.
19-05-30 31m tweak_kd diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 5310 W: 1113 L: 1257 D: 2940
sprt @ 10+0.1 th 1 Misc. [0, 4] tweaks to king danger.
19-05-30 31m tweak_safecheck diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 40408 W: 9065 L: 9068 D: 22275
sprt @ 10+0.1 th 1 Inspired by @praveentml's sbb1, which showed that very small tweaks to SafeCheck values could hold some promise. Here's my attempt.
19-05-29 31m combo_052819 diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 23266 W: 3880 L: 3959 D: 15427
sprt @ 60+0.6 th 1 LTC for combo.
19-05-29 31m combo_052819 diff
LLR: 2.95 (-2.94,2.94) [0.00,4.00]
Total: 154838 W: 34860 L: 34078 D: 85900
sprt @ 10+0.1 th 1 Revert imbalance change.
19-05-29 31m combo_052819^ diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 63174 W: 14136 L: 14049 D: 34989
sprt @ 10+0.1 th 1 Since the two bishop-focused tweaks may interact, test them separately in the combo. Revert PSQT change.
19-05-29 31m combo_052819 diff
LLR: -2.94 (-2.94,2.94) [0.00,4.00]
Total: 26441 W: 5885 L: 5944 D: 14612
sprt @ 10+0.1 th 1 Large combo of recent promising [0, 4]s. (bugfix)
19-05-29 31m combo_052819 diff
LLR: -0.09 (-2.94,2.94) [0.00,4.00]
Total: 126 W: 31 L: 36 D: 59
sprt @ 10+0.1 th 1 Large combo of recent promising [0, 4]s.
19-05-28 31m tweak_outpost3 diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 62264 W: 10645 L: 10597 D: 41022
sprt @ 60+0.6 th 1 Speculative LTC for 90K yellow, low throughput.
19-05-28 31m tweak_outpost3 diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 43252 W: 9720 L: 9712 D: 23820
sprt @ 10+0.1 th 1 Double effect (eg += 2)
19-05-28 31m tweak_outpost3 diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 90969 W: 20419 L: 20221 D: 50329
sprt @ 10+0.1 th 1 EG += 1
19-05-28 31m tweak_outpost3^ diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 28004 W: 6236 L: 6289 D: 15479
sprt @ 10+0.1 th 1 EG -= 1
19-05-28 31m tweak_outpost3^^^ diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 26960 W: 5996 L: 6053 D: 14911
sprt @ 10+0.1 th 1 On an empty framework, uninformed search for small tweaks to known Elo-sensitive parameters, to include in combo with recent promising [0, 4]s. MG -= 1.
19-05-28 31m tweak_outpost3^^ diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 21334 W: 4711 L: 4791 D: 11832
sprt @ 10+0.1 th 1 MG += 1
19-05-27 31m tweak_KDC3 diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 27703 W: 6139 L: 6194 D: 15370
sprt @ 10+0.1 th 1 King danger constant: -1.
19-05-26 31m tweak_kingProtector2^^ diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 30003 W: 6747 L: 6792 D: 16464
sprt @ 10+0.1 th 1 MG += 1
19-05-26 31m tweak_kingProtector2 diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 29258 W: 6473 L: 6522 D: 16263
sprt @ 10+0.1 th 1 EG += 1
19-05-26 31m tweak_kingProtector2^^^ diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 16274 W: 3586 L: 3686 D: 9002
sprt @ 10+0.1 th 1 For an empty framework: KingProtector is a small, highly sensitive Score, so try some micro-tweaks to see if there's anything here. MG -= 1.
19-05-26 31m tweak_kingProtector2^ diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 11602 W: 2589 L: 2708 D: 6305
sprt @ 10+0.1 th 1 EG -= 1
19-05-25 31m passer_bb_attacks diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 13417 W: 2908 L: 3006 D: 7503
sprt @ 10+0.1 th 1 A few testers have tried to restore this part of evaluation. Here's my best guess--no attacks through friendly pawns or enemy doubly-attacked pieces.
19-05-23 31m combo_KFA diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 37904 W: 6436 L: 6467 D: 25001
sprt @ 60+0.6 th 1 LTC: Combo of KFA tweaks and @locutus2's promising [0, 4].
19-05-23 31m passer_sac diff
LLR: -2.95 (-2.94,2.94) [0.00,3.50]
Total: 66563 W: 11348 L: 11335 D: 43880
sprt @ 60+0.6 th 1 Speculative LTC for 178K yellow. Low throughput (500).
19-05-23 31m combo_KFA diff
LLR: 2.96 (-2.94,2.94) [0.00,4.00]
Total: 67526 W: 15420 L: 14983 D: 37123
sprt @ 10+0.1 th 1 Combo of KFA tweaks and @locutus2's promising [0, 4]
19-05-23 31m combo_KFA^ diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 9526 W: 2077 L: 2204 D: 5245
sprt @ 10+0.1 th 1 Combo of KFA tweaks and @protonspring's promising [0, 4]
19-05-22 31m passer_sac diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 178243 W: 40022 L: 39299 D: 98922
sprt @ 10+0.1 th 1 Same, but include phalanx pawns.
19-05-22 31m passer_sac^ diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 23239 W: 5140 L: 5189 D: 12910
sprt @ 10+0.1 th 1 Count as a full passer even if blocked; don't include phalanx pawns.
19-05-22 31m sb1~3 diff
LLR: -2.95 (-2.94,2.94) [0.00,3.50]
Total: 93008 W: 15854 L: 15766 D: 61388
sprt @ 60+0.6 th 1 Speculative LTC for 82K yellow. Low throughput (500).
19-05-21 31m passer_sac~4 diff
LLR: -2.96 (-2.94,2.94) [0.00,3.50]
Total: 41981 W: 7176 L: 7233 D: 27572
sprt @ 60+0.6 th 1 This 73K yellow appeared to replicate and potentially improve on the strong yellow STC. Attempt speculative LTC; please approve only if STCs fail. Low throughput (500).
19-05-22 31m passer_sac diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 52568 W: 11661 L: 11565 D: 29342
sprt @ 10+0.1 th 1 Count as a full passer, not a candidate passer. Include phalanxes.
19-05-22 31m passer_sac^ diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 39146 W: 8722 L: 8692 D: 21732
sprt @ 10+0.1 th 1 Count as a full passer, not a candidate passer. Don't include phalanxes.
19-05-22 31m combo_comp_lmr diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 5418 W: 1098 L: 1240 D: 3080
sprt @ 10+0.1 th 1 Combo of the best LTC [0, 4]s from the past week, from @protonspring (96K yellow) and @Vizvizdenec (87K yellow)
19-05-22 31m sb1 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 14868 W: 3266 L: 3357 D: 8245
sprt @ 10+0.1 th 1 I think these tweaks to defended and unsafe squares might work naturally together--let's see.
19-05-21 31m sb1 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 82299 W: 18507 L: 18261 D: 45531
sprt @ 10+0.1 th 1 A related idea: squares occupied by friendly non-pawns are defended.
19-05-21 31m sb1 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 30468 W: 6808 L: 6821 D: 16839
sprt @ 10+0.1 th 1 Part 2
19-05-21 31m sb1 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 8591 W: 1880 L: 2002 D: 4709
sprt @ 10+0.1 th 1 Just knights and bishops.
19-05-21 31m sb1 diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 16870 W: 3745 L: 3826 D: 9299
sprt @ 10+0.1 th 1 Just friendly rooks and queens.
19-05-21 31m sb1^ diff
LLR: -2.94 (-2.94,2.94) [0.50,4.50]
Total: 27528 W: 6162 L: 6189 D: 15177
sprt @ 10+0.1 th 1 Try to break the two changes in @praveentml's promising patch into two separate tests, to see if one is better than the combination as a whole. Part 1.
19-05-21 31m passer_sac^ diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 45862 W: 10257 L: 10193 D: 25412
sprt @ 10+0.1 th 1 Opposite effect. Candidate passer bonus *= 2/3; include phalanx in sacrifices.
19-05-21 31m passer_sac diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 21865 W: 4796 L: 4852 D: 12217
sprt @ 10+0.1 th 1 Opposite effect. Candidate passer bonus *= 2/3; do not include phalanx in sacrifices.
19-05-21 31m passer_sac diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 8135 W: 1721 L: 1845 D: 4569
sprt @ 10+0.1 th 1 Same, but without the phalanx inclusion. (If either test passes, I would be grateful to anyone who submits the LTC on my behalf.)
19-05-21 31m passer_sac^ diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 8539 W: 1855 L: 1978 D: 4706
sprt @ 10+0.1 th 1 Compensate for expanded candidate passer definition: increase divisor from 2 to 3.
19-05-21 31m passer_sac diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 73870 W: 16508 L: 16306 D: 41056
sprt @ 10+0.1 th 1 My take on @Lolligerhans's lol_mp_1. Include not just threatened sacrifices, but ones currently on the board (i.e., phalanx pawns that meet the same criteria as the support pawn pushes).
19-05-21 31m passer_phalanx diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 18567 W: 4149 L: 4221 D: 10197
sprt @ 10+0.1 th 1 Test the "| phalanx" from my promising (as of 25K games) passer_sac test, in isolation.
19-05-20 31m passer_bb diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 41104 W: 9159 L: 9119 D: 22826
sprt @ 10+0.1 th 1 @locutus2 and I have both recently tested promising changes to this line--but they surely interact. Try them both.
19-05-19 31m simplify_passer_bb2 diff
LLR: -2.94 (-2.94,2.94) [0.00,4.00]
Total: 56673 W: 9646 L: 9616 D: 37411
sprt @ 60+0.6 th 1 Retest as Elo-gainer as requested by @mcostalba. See discussion on PR #2154.