Stockfish Testing Queue

Pending - 0 tests 0.0 hrs

None

Active - 0 tests

Finished - 463 tests

18-08-04 31m KingFau2 diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 217624 W: 37358 L: 36804 D: 143462
sprt @ 60+0.6 th 1 For @fauzi2: Test number 2, after 370k.
18-08-04 31m ROPattacks4 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 20572 W: 4507 L: 4537 D: 11528
sprt @ 10+0.1 th 1 Half effect, only pawns we block.
18-08-04 31m ROPattacks4^ diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 15953 W: 3560 L: 3612 D: 8781
sprt @ 10+0.1 th 1 Double effect appears to have been too much. Try extra half effect for blocked enemy pawns.
18-08-04 31m ROPattacks4 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 12217 W: 2674 L: 2745 D: 6798
sprt @ 10+0.1 th 1 Same, but only pawns we block.
18-08-04 31m ROPattacks4^ diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 8778 W: 1902 L: 1990 D: 4886
sprt @ 10+0.1 th 1 I had tried my best to avoid adding complexity in my previous tests, and thus intended [0, 4], but since I have been asked to use [0, 5] bounds, I'll try some ideas on this branch which certainly add complexity. Double effect for blocked pawns. The total effect of this patch is to dramatically change the meaning of RookOnPawn.
18-08-04 31m risk11 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 6528 W: 1074 L: 1171 D: 4283
sprt @ 60+0.6 th 1 For @snicolet: Only for depths 0,2,4.
18-08-04 31m risk11^^ diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 6495 W: 1084 L: 1181 D: 4230
sprt @ 60+0.6 th 1 For @snicolet: Only for depth < 1.
18-08-04 31m risk11~5 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 6918 W: 1131 L: 1227 D: 4560
sprt @ 60+0.6 th 1 Framework is going empty with 4 Elo-gaining LTCs to run. I hope it's OK if I start them. For @snicolet: Only for depth < 4.
18-08-02 31m RookOnPawn_attacks^ diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 100629 W: 17394 L: 17219 D: 66016
sprt @ 60+0.6 th 1 Speculative LTC for this 79K STC yellow. Since this mostly affects endgame bonus, it may perform better at LTC.
18-08-03 31m ROPattacks3^^ diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 56951 W: 12689 L: 12538 D: 31724
sprt @ 10+0.1 th 1 Rescheduled as [0, 5] as requested. +25%, and also introduce a 6 cp middlegame component in this endgame-only Score. When paired with a corresponding tweak to ThreatByMinor in the past, this middlegame component has fared well.
18-08-03 31m ROPattacks3 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 45908 W: 10161 L: 10066 D: 25681
sprt @ 10+0.1 th 1 Rescheduled as [0, 5] as requested. Increase the middlegame component to 6 (currently zero), but make no change to the endgame component.
18-08-03 31m ROPattacks3^^^^ diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 8060 W: 1755 L: 1847 D: 4458
sprt @ 10+0.1 th 1 Rescheduled as [0, 5] as requested, although I do not understand why the original [0, 4] is incorrect. After over 99K games, the speculative LTC is still running. Consistent with several long STC runs, this idea seems close to passing. Because it may change the interaction with ThreatByRook, try several different tweaks to that endgame-only Score as compensation. +25%.
18-08-03 31m ROPattacks3^^^ diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 7914 W: 1686 L: 1779 D: 4449
sprt @ 10+0.1 th 1 Rescheduled as [0, 5] as requested. -25%.
18-08-03 31m ROPattacks3^ diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 18676 W: 4099 L: 4138 D: 10439
sprt @ 10+0.1 th 1 -25%, and introduce the 6 cp middlegame component.
18-08-02 31m mob_initiative4^^^^ diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 99251 W: 17157 L: 16875 D: 65219
sprt @ 60+0.6 th 1 LTC for @snicolet, since framework is empty.
18-08-02 31m KingOpenDiagonal diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 9744 W: 2137 L: 2221 D: 5386
sprt @ 10+0.1 th 1 Also penalize during the endgame: S(10, 10).
18-08-02 31m KingOpenDiagonal^ diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 9306 W: 2045 L: 2131 D: 5130
sprt @ 10+0.1 th 1 My initial attempt in response to recent comments by Bryan on AlphaZero vs. Stockfish 8: Game 8. In r5k1/1bq1bppp/2p1p3/1pP1P3/1Pp1B1QP/2B3P1/5PK1/2R5 w - - 5 26, SF does not choose Qh5, provoking g6 and weakening Black's kingside. Give a middlegame-only S(10, 0) penalty if all of these conditions are met: (a) Our king ring intersects a long diagonal, (b) our king ring does not intersect the center of the board (so the long diagonal is unique), (c) we have no pawns blocking this intersection in our king ring, and (d) the opponent has the "correct" bishop to take advantage of this weakness. Notably, this applies regardless of whether the diagonal is currently attacked--in the aforementioned position, it's still a weakness despite the white e5-pawn.
18-08-02 31m ROPattacks2 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 10707 W: 2321 L: 2400 D: 5986
sprt @ 10+0.1 th 1 The other test uses the best compensation from RookOnPawn_attacks, +25%. But this branch further narrows the bonus, so try further (50%) compensation.
18-08-02 31m ROPattacks2^ diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 8104 W: 1785 L: 1877 D: 4442
sprt @ 10+0.1 th 1 RookOnPawn_attacks has given me 3 long yellow runs (59K, 64K, 79K) in a row. I think there's good reason to attempt speculative LTC, since this is a primarily endgame-based bonus, but first let's search for a green if possible. It makes sense that only considering weak pawns doesn't usually work with ROP--we were using pseudo-attacks, so there could be many blockers in the way, rendering the pawns' weakness irrelevant. However, here we consider direct attacks, and it makes sense to adding this condition.
18-08-01 31m RookOnPawn_attacks^ diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 79060 W: 17668 L: 17518 D: 43874
sprt @ 10+0.1 th 1 +50% compensation was a 59K yellow; +100% was 11K red. Try +25%. (This patch is a [0, 4] change, right?)
18-08-01 31m RookOnPawn_attacks diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 64207 W: 14275 L: 14185 D: 35747
sprt @ 10+0.1 th 1 +75%.
18-07-31 31m simplify_RookOnPawn2b diff
LLR: -2.95 (-2.94,2.94) [-3.00,1.00]
Total: 167759 W: 36865 L: 37362 D: 93532
sprt @ 10+0.1 th 1 Despite the very close similarity in value between RookOnPawn and ThreatByRook[PAWN], some controversy arose on my PR to eliminate the former, because of the differences in the conditions where these bonuses are applied. Although I still think that PR is a good idea, I recognize that one of the larger differences is that RookOnPawn is based on pseudo-attacks. A recent neutral (+0.23 Elo) result on a test of mine suggests, however, that perhaps RookOnPawn may be replaced with "real" attacks without regression--so merge new master and attempt this unification.
18-07-31 31m RookOnPawn_attacks diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 59060 W: 13129 L: 13059 D: 32872
sprt @ 10+0.1 th 1 Partially compensate by increasing RookOnPawn by 50%.
18-07-31 31m RookOnPawn_attacks diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 11532 W: 2542 L: 2661 D: 6329
sprt @ 10+0.1 th 1 Double RookOnPawn outright to compensate, since the other tweak in this test seems to reduce its effect size by a factor larger than 3.
18-07-31 31m RookOnPawn_attacks diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 35522 W: 7909 L: 7932 D: 19681
sprt @ 10+0.1 th 1 Dramatically restrict RookOnPawn to only apply to pawns we attack. I don't expect this to pass, but if it's close to neutral, it implies an opportunity to simplify RookOnPawn. (If RookOnPawn can be restricted to just direct attacks, there are fewer remaining differences between RookOnPawn and a ThreatByRook/ThreatByRank approach.)
18-07-31 31m RookOnPawnBlock2 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 9843 W: 2087 L: 2170 D: 5586
sprt @ 10+0.1 th 1 I committed a logic error in my first version; I intended to use our king-pinned pieces but used our king blockers generally. Fix and try again.
18-07-30 31m RookOnPawnBlock4 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 32181 W: 7197 L: 7169 D: 17815
sprt @ 10+0.1 th 1 A simpler version: apply the no pawn attacks restriction to both cases.
18-07-30 31m RookOnPawnBlock5 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 10815 W: 2330 L: 2408 D: 6077
sprt @ 10+0.1 th 1 No RookOnPawn bonus through our own king.
18-07-29 31m RookOnPawnBlock3 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 45332 W: 10128 L: 10034 D: 25170
sprt @ 10+0.1 th 1 Require further that our pawns along the file are not attacked by enemy pawns, because if they are, they can capture and allow our rook to attack.
18-07-29 31m RookOnPawnBlock4 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 16014 W: 3532 L: 3584 D: 8898
sprt @ 10+0.1 th 1 I have two related tests which appear to be promising but not quite good enough to pass. Try applying both conditions together. Block the RookOnPawn bonus with (a) our pawns on the file which cannot capture, (b) our enemy-blocked pawns.
18-07-29 31m RookOnPawnBlock3^ diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 10765 W: 2316 L: 2395 D: 6054
sprt @ 10+0.1 th 1 Don't give RookOnPawn bonus through our own pawns along the file, since it will be difficult for them to move out of the way.
18-07-29 31m hanging diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 67323 W: 11463 L: 11399 D: 44461
sprt @ 60+0.6 th 1 Since the framework is empty, LTC for IIvec. Hanging (57,32).
18-07-29 31m RookOnPawnBlock diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 46713 W: 10414 L: 10314 D: 25985
sprt @ 10+0.1 th 1 Only consider our pawns that the enemy blocks.
18-07-29 31m RookOnPawnBlock diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 24509 W: 5400 L: 5410 D: 13699
sprt @ 10+0.1 th 1 Also apply locutus2's idea to the previous version.
18-07-29 31m RookOnPawnBlock2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 14850 W: 3273 L: 3331 D: 8246
sprt @ 10+0.1 th 1 No RookOnPawn through our king blockers.
18-07-29 31m RookOnPawnBlock diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 11508 W: 2513 L: 2588 D: 6407
sprt @ 10+0.1 th 1 Disallow enemy pawn attacks, or else our pawn could capture. locutus2's idea.
18-07-29 31m RookOnPawnBlock diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 7371 W: 1587 L: 1682 D: 4102
sprt @ 10+0.1 th 1 Don't give RookOnPawn bonus through our own blocked pawns.
18-07-28 31m ThreatByRank_kf diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 18221 W: 4046 L: 4087 D: 10088
sprt @ 10+0.1 th 1 Half ThreatByRank bonus for rook attacks against enemy pawns not only if they are king-pinned, but also if they are merely on the king file. This includes rook attacks along a rank, or attacks on the pawn when there are still other pieces blocking an immediate check.
18-07-27 31m ThreatByRank_blocked^ diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 17051 W: 3809 L: 3856 D: 9386
sprt @ 10+0.1 th 1 Divide by 6 rather than 4.
18-07-27 31m ThreatByRank_blocked diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 8167 W: 1793 L: 1885 D: 4489
sprt @ 10+0.1 th 1 Restore the version currently performing well, but merge in sg's 88K yellow parameter tweak, which should interact directly with this patch.
18-07-27 31m ThreatByRank_blocked^^ diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 6241 W: 1359 L: 1460 D: 3422
sprt @ 10+0.1 th 1 42K games in to a STC, it looks like I'm getting close (Elo: +1.2 to +1.4). Try a few small variations. Here, divide by 3 rather than 4.
18-07-27 31m ThreatByRank_blocked diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 46213 W: 10414 L: 10315 D: 25484
sprt @ 10+0.1 th 1 I'm running out of ideas on this branch...try half effect.
18-07-27 31m ThreatByRank_blocked diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 17517 W: 3922 L: 3967 D: 9628
sprt @ 10+0.1 th 1 Surprisingly, the minors-only version (-0.69 Elo) was worse that the minors-and-rooks version (+0.76 Elo). Perhaps a rooks-only version could be an Elo gain...
18-07-27 31m ThreatByRank_blocked diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 15339 W: 3412 L: 3468 D: 8459
sprt @ 10+0.1 th 1 Revert to the original effect size, but still only apply it to minors.
18-07-27 31m ThreatByRank_blocked diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 5861 W: 1239 L: 1342 D: 3280
sprt @ 10+0.1 th 1 Double effect, but only for the minor case; inspired by @snicolet's pinned_threat results. No effect for rooks.
18-07-27 31m ThreatByRank_blocked diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 3645 W: 778 L: 893 D: 1974
sprt @ 10+0.1 th 1 Double the effect of the just-completed yellow STC.
18-07-27 31m ThreatByRank_blocked diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 29749 W: 6713 L: 6696 D: 16340
sprt @ 10+0.1 th 1 Include @locutus2's bugfix and retry this STC. Only include the blocked enemy pawns that we block. Enemy pieces blocking enemy pawns may move out of the way.
18-07-27 31m ThreatByRank_blocked^ diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 15834 W: 3524 L: 3577 D: 8733
sprt @ 10+0.1 th 1 Thanks to @locutus2 who spotted my error and corrected me. Retry this version with his bugfix included. Include blocked pawns in ThreatByRank with half effect, alongside the recently added king-pinned pawns.
18-07-26 31m simplify_RookOnPawn2a diff
LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 54991 W: 9381 L: 9316 D: 36294
sprt @ 60+0.6 th 1 LTC after merging new master. Retire RookOnPawn score, replacing it with ThreatByRook[PAWN], and add middlegame components to ThreatByRook[PAWN] and ThreatByMinor[PAWN] to compensate.
18-07-26 31m ThreatByRank_blocked diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 32166 W: 7182 L: 7154 D: 17830
sprt @ 10+0.1 th 1 An idea taking inspiration from both @locutus2's recent passed patch and @Rocky640's tests with Overload for blocked pawns. Include blocked pawns in ThreatByRank with half effect, alongside the recently added king-pinned pawns.