Stockfish Testing Queue

Finished - 1120 tests

19-02-14 31m passed_protection diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 41791 W: 9259 L: 9217 D: 23315
sprt @ 10+0.1 th 1 Opposite effect: candidate passers only.
19-02-14 31m threat_strongqueen3 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 26079 W: 5770 L: 5805 D: 14504
sprt @ 10+0.1 th 1 In November, this was a 70K STC and 108K LTC yellow--back when we used [0, 5] bounds for both. Respin with new bounds. Master already has ThreatByRook[QUEEN] but applies this only to "weak" queens, but queens are always weak to threats by lesser pieces. Give ThreatByRook[QUEEN] / 2 bonus for a rook attack against a "strong" queen.
19-02-14 31m passed_protection diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 10366 W: 2217 L: 2330 D: 5819
sprt @ 10+0.1 th 1 My take on MJZ's idea. Don't apply this increased bonus for candidate passers, only for "real" ones.
19-02-13 31m passerUnsafe_ABP diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 14015 W: 3015 L: 3110 D: 7890
sprt @ 10+0.1 th 1 More complex take. Only consider squares defended by our non-passed pawns. If a square is defended by a passed pawn, we might trade two passers for one--which probably isn't a good idea. (Sometimes this is a useful tactic for promotion, but search should take care of that.)
19-02-13 31m passerUnsafe_ABP diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 33325 W: 7339 L: 7339 D: 18647
sprt @ 10+0.1 th 1 A really intuitive idea: a square in front of a passer is not unsafe if we attack it with a pawn. (I suspect this may have been tried before, but I'm curious and the framework is mostly empty.)
19-02-13 31m lmr_pawnPushVsK2 diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 34373 W: 7565 L: 7560 D: 19248
sprt @ 10+0.1 th 1 An insight to try to turn this 165K yellow green: we should only apply this if the opponent cannot castle. If they can, pressuring the king is likely of little use.
19-02-13 31m lmr_pawnPushVsK2 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 10942 W: 2340 L: 2450 D: 6152
sprt @ 10+0.1 th 1 Only if we have a friendly rook on the file to support the push.
19-02-12 31m MBP_neighbors diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 8537 W: 1825 L: 1947 D: 4765
sprt @ 10+0.1 th 1 Require that there be friendly pawns on neighboring files before giving MinorBehindPawn bonus with friendly pawns.
19-02-10 31m lmr_pawnPushVsK~9 diff
LLR: -2.96 (-2.94,2.94) [0.00,3.50]
Total: 30554 W: 4985 L: 5076 D: 20493
sprt @ 60+0.6 th 1 Speculative LTC for 165K yellow.
19-02-09 31m lmr_pawnPushVsK diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 62411 W: 13782 L: 13638 D: 34991
sprt @ 10+0.1 th 1 Comparatively small change to 165K yellow: NPM > RNB rather than RNN.
19-02-09 31m lmr_pawnPushVsK diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 25288 W: 5569 L: 5608 D: 14111
sprt @ 10+0.1 th 1 Code reorder of 165K yellow: apply to pawn captures too, not only pawn pushes.
19-02-09 31m lmr_pawnPushVsK diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 14736 W: 3184 L: 3276 D: 8276
sprt @ 10+0.1 th 1 Stricter NPM: NPM > QR.
19-02-09 31m lmr_pawnPushVsK diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 14373 W: 3087 L: 3181 D: 8105
sprt @ 10+0.1 th 1 Less strict NPM condition: NPM > RN.
19-02-09 31m lmr_pawnPushVsK diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 11295 W: 2406 L: 2515 D: 6374
sprt @ 10+0.1 th 1 165K yellow is very close...before speculative LTC, try a few more variations. @ElbertoOne's suggestion: 2 plies.
19-02-09 31m extend_passercreation diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 6156 W: 1267 L: 1401 D: 3488
sprt @ 10+0.1 th 1 Extend captures (except pawn moves) that create passed pawns.
19-02-09 31m extend_passercreation^ diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 3491 W: 704 L: 852 D: 1935
sprt @ 10+0.1 th 1 Extend captures that create passed pawns.
19-02-09 31m lmr_pawnPushVsK diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 165291 W: 36466 L: 35817 D: 93008
sprt @ 10+0.1 th 1 @vondele recommended trying a NPM condition. NPM > R + 2N.
19-02-09 31m lmr_pawnPushVsK diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 59580 W: 13099 L: 12970 D: 33511
sprt @ 10+0.1 th 1 Initial (very preliminary) results are promising--but I have many more ideas to test. No explicit NPM condition, but require that the pawn be in front of the king. The enemy king is more likely to be near relative rank 8 in the middlegame, making this tacitly a game-phase condition; however, it also corrects for the fact that I'm not sure I want to affect pawn pushes behind the enemy king in the endgame. Include the king's neighboring files in this version.
19-02-09 31m lmr_pawnPushVsK diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 44440 W: 9675 L: 9621 D: 25144
sprt @ 10+0.1 th 1 Just on the enemy king files, not the neighboring ones.
19-02-09 31m lmr_pawnPushVsK diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 38664 W: 8517 L: 8491 D: 21656
sprt @ 10+0.1 th 1 Less reduction for pawn pushes near the enemy king.
19-02-09 31m lmr_pawnPushVsK diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 18969 W: 4078 L: 4149 D: 10742
sprt @ 10+0.1 th 1 Relative rank restriction, this time only on the enemy king file. (If the event that any of these STCs pass, I'll likely be asleep and would be grateful to anyone who starts the LTC.)
19-02-09 31m lmr_pawnPushVsK diff
LLR: -2.94 (-2.94,2.94) [0.50,4.50]
Total: 14369 W: 3109 L: 3202 D: 8058
sprt @ 10+0.1 th 1 King file only and NPM restriction.
19-02-08 31m comment_semiopenFiles diff
LLR: -2.96 (-2.94,2.94) [-3.00,1.00]
Total: 51765 W: 11359 L: 11627 D: 28779
sprt @ 10+0.1 th 1 Surprised by recent failure. Try two small fixes: (a) change the type of semiopen_file() to match, removing an unnecessary conversion; (b) reorder the fields in the struct. Since they are stored in order of declaration, it's possible that having a Bitboard (64 bits) after the ints (32 bits) was causing some sort of sub-optimal memory layout, though I'm not sure...
19-02-08 31m comment_semiopenFiles diff
LLR: -2.96 (-2.94,2.94) [-3.00,1.00]
Total: 57789 W: 12481 L: 12759 D: 32549
sprt @ 10+0.1 th 1 There has been a lot of confusion about semiopenFiles. It is an "int" to save space only but is actually a Bitboard in effect; combined with a lack of documentation, it has tripped up numerous users in the past year (including me). Check that there's no regression in making it a true Bitboard as it arguably should be, and add a detailed comment.
19-02-08 31m SiberianOutpost3 diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 13952 W: 3032 L: 3128 D: 7792
sprt @ 10+0.1 th 1 Rank 6 only.
19-02-08 31m SiberianOutpost3 diff
LLR: -2.94 (-2.94,2.94) [0.50,4.50]
Total: 13957 W: 3038 L: 3133 D: 7786
sprt @ 10+0.1 th 1 Don't double an occupied Outpost if it is more than three squares from the nearest non-pawn enemy.
19-02-08 31m BishopBackward diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 18314 W: 4090 L: 4164 D: 10060
sprt @ 10+0.1 th 1 Also require ~attackedBy2[Them].
19-02-08 31m BishopBackward diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 14462 W: 3100 L: 3193 D: 8169
sprt @ 10+0.1 th 1 @Alayan-stk-2's idea: require pos.non_pawn_material(Us) > BishopValueMg.
19-02-08 31m BishopBackward diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 44282 W: 9731 L: 9677 D: 24874
sprt @ 10+0.1 th 1 S(35, 35)
19-02-08 31m SliderNoBlocker diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 33138 W: 7221 L: 7222 D: 18695
sprt @ 10+0.1 th 1 Exclude our king blockers.
19-02-08 31m SliderNoPawn diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 29400 W: 6510 L: 6529 D: 16361
sprt @ 10+0.1 th 1 Logical follow-up to @Vizvezdenec's 83K yellow, SliderNoPawn1. Exclude pawns that are stopped by enemy double pawn attacks.
19-02-08 31m BishopBackward diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 56116 W: 12250 L: 12139 D: 31727
sprt @ 10+0.1 th 1 S(15, 15) for enemy pawns that are (i) bishop-attacked, (ii) bishop-defended, (iii) not pawn-defended, and (iv) blocked by us.
19-02-08 31m BishopBackward diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 33383 W: 7318 L: 7318 D: 18747
sprt @ 10+0.1 th 1 S(25, 25)
19-02-08 31m BishopBackward diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 16445 W: 3544 L: 3627 D: 9274
sprt @ 10+0.1 th 1 Also inspired by TCEC 14 Superfinal Game 11. In the position below, one of the most striking features is the relative activity of the light-squared bishops. Give a bonus for bishop attacks on bishop-defended backward pawns. This combines the most striking features of Overload (relative immobility of the defending bishop, susceptibility to built-up attacks) with two insights specific to this case: (a) because the pawn is backward in the chain, the defender is probably defending from behind and the attacker in front--making this a proxy for "good"/"bad" bishops; (b) because the pawn is backward, it has much lower mobility and is much weaker than was typical for Overload. S(15, 15). 2b1q2r/k3r1b1/4pRP1/1p1pP2p/p1pP1N1B/PnP4B/1P3QRK/8 b - - 4 45
19-02-08 31m SiberianOutpost2 diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 12462 W: 2698 L: 2801 D: 6963
sprt @ 10+0.1 th 1 Only for outposts on relative rank 6, as in TCEC 14 Superfinal Game 11.
19-02-08 31m SiberianOutpost2 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 11805 W: 2516 L: 2622 D: 6667
sprt @ 10+0.1 th 1 More restrictive definition with full effect: Outpost on A/B file and opposing king on G/H file, or vice versa.
19-02-08 31m SiberianOutpost diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 11654 W: 2471 L: 2578 D: 6605
sprt @ 10+0.1 th 1 Half effect.
19-02-07 31m SiberianOutpost diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 8494 W: 1780 L: 1902 D: 4812
sprt @ 10+0.1 th 1 Use distance to enemy king rather than enemy king flank.
19-02-07 31m SiberianOutpost diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 8856 W: 1894 L: 2015 D: 4947
sprt @ 10+0.1 th 1 Name inspired by @Kurtbusch's SiberianQueen. Don't double the Outpost bonus for occupied outposts if the position of the outpost is terrible: if its attacks are neither on the enemy king flank nor enemy pieces (excluding pawn-defended pawns). Based on TCEC 14 Superfinal Game 11.
19-02-07 31m TBM_passer diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 30711 W: 6638 L: 6652 D: 17421
sprt @ 10+0.1 th 1 No restriction by rank at all.
19-02-07 31m BishopPawns_pawnattacks diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 15315 W: 3356 L: 3445 D: 8514
sprt @ 10+0.1 th 1 Bugfix: exclude our pawn-defended, bishop-defended pawns.
19-02-07 31m TBM_passer^^ diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 15108 W: 3262 L: 3352 D: 8494
sprt @ 10+0.1 th 1 Require relative_rank(Them, new passer) >= rank 5 (previously, rank 6).
19-02-07 31m TBM_passer^ diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 13330 W: 2888 L: 2987 D: 7455
sprt @ 10+0.1 th 1 Require relative_rank(Them, new passer) >= rank 4 (previously, rank 6).
19-02-07 31m TBM_passer diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 38282 W: 8373 L: 8349 D: 21560
sprt @ 10+0.1 th 1 Idea from Bryan's analysis of the recent TCEC 14 Superfinal Game 11, LC0 1-0 SF. No ThreatByMinor or corresponding ThreatByRank if the capture would create an advanced enemy passer.
19-02-07 31m BishopPawns_pawnattacks diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 6616 W: 1369 L: 1500 D: 3747
sprt @ 10+0.1 th 1 An intuitive idea based on the same position: exclude coordinated pawn/bishop attacks from BishopPawns. Specifically, subtract from pe->pawns_on_same_color_squares(Us, s) a popcount of squares attacked by our pawns and the bishop in question.
19-02-07 31m BishopPawns_bothsides^ diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 4478 W: 892 L: 1034 D: 2552
sprt @ 10+0.1 th 1 Inspired by what appears to be an oversight in move 45 of the same game (FEN below), where both sides have equal BishopPawns penalties but Black's bishop-pawn interaction is clearly terrible while White's is productive. Try halving BishopPawns for a bishop that can see both sides of the board (White's half and Black's half). 2b1q2r/k3r1b1/4pRP1/1p1pP2p/p1pP1N1B/PnP4B/1P3QRK/8 b - - 4 45
19-02-07 31m BishopPawns_bothsides diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 3692 W: 723 L: 868 D: 2101
sprt @ 10+0.1 th 1 Double BishopPawns for a bishop that cannot.
19-02-05 31m extend_perpetual diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 12491 W: 2676 L: 2779 D: 7036
sprt @ 10+0.1 th 1 Extend only if the enemy king has no surrounding pieces to defend it.
19-02-04 31m extend_perpetual diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 42876 W: 9341 L: 9295 D: 24240
sprt @ 10+0.1 th 1 Queen checks only, since these are the most likely to be perpetuals.
19-02-04 31m extend_perpetual diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 23210 W: 5046 L: 5096 D: 13068
sprt @ 10+0.1 th 1 Derivative of @MJZ1977's test. Exclude pawn checks, since these are not reversible moves and are unlikely to result in perpetuals.