Stockfish Testing Queue

Finished - 1117 tests

19-01-13 31m hinderMinorEdge diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 35483 W: 7748 L: 7738 D: 19997
sprt @ 10+0.1 th 1 Halve HinderMinor if the king is on the edge.
19-01-13 31m hinderMinorEdge diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 23652 W: 5190 L: 5238 D: 13224
sprt @ 10+0.1 th 1 Double HinderMinor in the key positions (one-tenth of HinderMinor positions) where the king is not on the edge. Test against PR #1945 (@ElbertoOne's passed patch).
19-01-13 31m tweak_hinderMinor^ diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 20818 W: 4519 L: 4601 D: 11698
sprt @ 10+0.1 th 1 Test these parameter tweaks in isolation against PR #1945. S(15, 15).
19-01-13 31m tweak_hinderMinor diff
LLR: -2.94 (-2.94,2.94) [0.00,4.00]
Total: 19718 W: 4312 L: 4398 D: 11008
sprt @ 10+0.1 th 1 Test these parameter tweaks in isolation against PR #1945. S(15, 23).
19-01-13 31m hinderMinorEdge diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 65832 W: 14451 L: 14292 D: 37089
sprt @ 10+0.1 th 1 A more modest +50% (rather than double).
19-01-13 31m hinderMinorEdge diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 62001 W: 13714 L: 13572 D: 34715
sprt @ 10+0.1 th 1 Previous analysis showed that eliminating king-on-edge cases reduced the proportion of positions where HinderMinor is applied by a factor of 10. More detailed analysis shows that the middle game is especially affected (20x) compared to the endgame (5x). Therefore, compensate middlegame only. Increase HinderMinor's mg component by 50%, eliminating mg->eg gradient: S(15, 15). Test against PR #1945 (@ElbertoOne's passed patch).
19-01-13 31m simplify_candidatePasse diff
LLR: -2.95 (-2.94,2.94) [-3.00,1.00]
Total: 15937 W: 3360 L: 3555 D: 9022
sprt @ 10+0.1 th 1 Try 5/8, splitting the difference between my two December STC greens (1/2 and 3/4).
19-01-13 31m hinderMinorEdge diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 7230 W: 1564 L: 1693 D: 3973
sprt @ 10+0.1 th 1 Eliminating the edge case reduces the proportion of positions HinderMinor is applied in by a factor of 10--yet it also appears to be consistently about 1 Elo better. Since this bonus is untuned in the first place, perhaps it can be increased in these key positions (about 1% of positions). Double HinderMinor to S(20, 30). Test against PR #1945 (@ElbertoOne's passed patch).
19-01-13 31m hinderMinorEdge diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 40888 W: 8971 L: 8934 D: 22983
sprt @ 10+0.1 th 1 So far, not applying HinderMinor when the king is on the edge appears reasonably strong (about +1 Elo). But this includes corners, and not applying HinderMinor when the king was specifically in a corner was terrible (about -4 Elo). What happens if we don't apply HinderMinor for a king on the edge of the board, unless it's in a corner (in which case we do)? Test against PR #1945 (@ElbertoOne's passed patch).
19-01-13 31m hinderMinorEdge^ diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 49062 W: 10716 L: 10640 D: 27706
sprt @ 10+0.1 th 1 Don't apply HinderMinor if the K is on the edge. Idea originally formulated by @protonspring. Test against PR #1945 (@ElbertoOne's passed patch).
19-01-13 31m DominatedKnight diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 36028 W: 7961 L: 7947 D: 20120
sprt @ 10+0.1 th 1 Best result so far is slight positive score, so increase effect size to S(50, 50). Also include a simplification/bugfix in logic: the previous attempts inadvertently gave DominatedKnight penalty for pinned knights.
19-01-13 31m DominatedKnight diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 23432 W: 5088 L: 5137 D: 13207
sprt @ 10+0.1 th 1 S(60, 60).
19-01-13 31m hinderMinorEdge diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 9015 W: 1931 L: 2051 D: 5033
sprt @ 10+0.1 th 1 Don't apply HinderMinor if the K is in the corner. Test against PR #1945 (@ElbertoOne's passed patch).
19-01-12 31m simplify_candidatePasse diff
LLR: -2.96 (-2.94,2.94) [-3.00,1.00]
Total: 55008 W: 12034 L: 12308 D: 30666
sprt @ 10+0.1 th 1 This passed on Dec. 17, but I never ran the LTC, which in hindsight I regret. Merge new master and retry. http://tests.stockfishchess.org/tests/view/5c176cb20ebc5902ba123156
19-01-12 31m DominatedKnight diff
LLR: -2.94 (-2.94,2.94) [0.50,4.50]
Total: 39089 W: 8604 L: 8575 D: 21910
sprt @ 10+0.1 th 1 S(40, 40).
19-01-13 31m simplify_candidatePasse diff
LLR: -2.96 (-2.94,2.94) [-3.00,1.00]
Total: 14313 W: 3091 L: 3285 D: 7937
sprt @ 10+0.1 th 1 Different parameter tweak.
19-01-13 31m fawn7 diff
LLR: -3.47 (-2.94,2.94) [0.50,4.50]
Total: 8949 W: 1898 L: 2047 D: 5004
sprt @ 10+0.1 th 1 In king danger, include among weak squares any enemy pawn attacks in the king ring that we do not pawn defend. (fixed bench--sorry!)
19-01-12 31m fawn diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 11749 W: 2545 L: 2651 D: 6553
sprt @ 10+0.1 th 1 I know this patch was criticized for its code style. But (at the very least) to inform future development, I would like to double the effect size and see whether this 76K STC yellow continues to demonstrate potential. (Not sure what to do if it somehow passes.)
19-01-12 31m DominatedKnight diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 10171 W: 2186 L: 2300 D: 5685
sprt @ 10+0.1 th 1 Revert to the best version so far and increase effect size. S(30, 30).
19-01-12 31m DominatedKnight diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 38006 W: 8295 L: 8273 D: 21438
sprt @ 10+0.1 th 1 Same as stopped run, but with improvement (removal of popcount) by @Vizvezdenec.
19-01-12 31m PseudoTrappedRook^ diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 39259 W: 8594 L: 8565 D: 22100
sprt @ 10+0.1 th 1 Inspired by @Vizvezdenec's recent useful/useless knight tests to retry this approach to trapped rook positions we currently do not explicitly handle. S(50, 50) penalty if a zero-mobility minor blocks a neighboring friendly rook along a rank.
19-01-12 31m Windmill2 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 34776 W: 7677 L: 7669 D: 19430
sprt @ 10+0.1 th 1 Windmill if the opponent has a discovered check that (a) captures material, as in previous attempts, or (b) cannot be captured by our king and makes a threat against our queen.
19-01-12 31m Windmill diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 17734 W: 3824 L: 3901 D: 10009
sprt @ 10+0.1 th 1 Require that the king be unable to capture the double check piece.
19-01-12 31m DominatedKnight diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 15881 W: 3454 L: 3540 D: 8887
sprt @ 10+0.1 th 1 Penalty if our knight, on the edge of the board, has no attacks within our mobilityArea that are not attacked by the enemy bishop. S(20, 20).
19-01-12 31m Windmill diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 11053 W: 2387 L: 2497 D: 6169
sprt @ 10+0.1 th 1 Try to turn this very-nearly-passed patch green. Double checks may also be used in a windmill-like fashion, so include them.
19-01-12 31m DominatedKnight diff
LLR: 0.02 (-2.94,2.94) [0.50,4.50]
Total: 5 W: 2 L: 1 D: 2
sprt @ 10+0.1 th 1 S(20, 20); require that the knight be on the edge of the board.
19-01-12 31m DominatedKnight diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 6832 W: 1435 L: 1566 D: 3831
sprt @ 10+0.1 th 1 S(10, 10) was Elo-neutral; try doubling it.
19-01-12 31m DominatedKnight diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 22330 W: 4938 L: 4992 D: 12400
sprt @ 10+0.1 th 1 Rework @protonspring's patch to be a per-knight penalty rather than per-bishop bonus. The problem with the per-bishop approach is that the three attackedBy[Them][KNIGHT] squares could be attacked by different knights, in which case neither are dominated.
19-01-12 31m PseudoTrappedRook^ diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 15810 W: 3442 L: 3528 D: 8840
sprt @ 10+0.1 th 1 S(100, 100)
19-01-12 31m PseudoTrappedRook diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 13509 W: 2944 L: 3042 D: 7523
sprt @ 10+0.1 th 1 S(50, 0)
19-01-12 31m PseudoTrappedRook diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 9170 W: 1976 L: 2095 D: 5099
sprt @ 10+0.1 th 1 Larger penalty (50) seems unexpectedly much stronger (than 25). In case this is not a fluke, try S(75, 75).
19-01-12 31m PseudoTrappedRook diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 12176 W: 2662 L: 2766 D: 6748
sprt @ 10+0.1 th 1 Half effect, S(25, 25).
19-01-11 31m combo_190110^ diff
LLR: -3.20 (-2.94,2.94) [0.00,4.00]
Total: 41331 W: 9053 L: 9068 D: 23210
sprt @ 10+0.1 th 1 The KDC and F tweaks don't seem to perform well together. Try dropping one at a time from the large combo. Remove F from the combo.
19-01-11 31m fawn6^ diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 32602 W: 7226 L: 7229 D: 18147
sprt @ 10+0.1 th 1 Another attempt to indirectly handle fawn pawns. Currently, when evaluating support for the purposes of Connected bonus, we simply count the supporting pawns. But a once-supported A/H-file pawn is much stronger than a once-supported pawn elsewhere, because it is less vulnerable to pawn attacks. Consider this special case to be doubly supported.
19-01-11 31m fawn6 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 29203 W: 6334 L: 6355 D: 16514
sprt @ 10+0.1 th 1 Half effect: give the arithmetic average of the once-supported and twice-supported bonuses, because this case is somewhere between the two.
19-01-11 31m fawn5 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 20607 W: 4543 L: 4605 D: 11459
sprt @ 10+0.1 th 1 Because my previous test turned a highly promising patch into a terrible one, try the opposite.
19-01-11 31m fawn5 diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 7694 W: 1663 L: 1790 D: 4241
sprt @ 10+0.1 th 1 @snicolet's fawn2 was extraordinarily close to passing (Dec. 31, STC 18K green, LTC 147K yellow). It was applied quite broadly, but based on recent tests it appears that fawn pawns must be opposed. Try narrowing this test but otherwise using the same parameters.
19-01-11 31m fawn4 diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 16337 W: 3538 L: 3622 D: 9177
sprt @ 10+0.1 th 1 Yet another potential approach to fawn pawns.
19-01-11 31m fawn3 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 15835 W: 3419 L: 3505 D: 8911
sprt @ 10+0.1 th 1 Check whether the side which has the fawn pawn also has a queen or rook.
19-01-11 31m fawn3 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 16038 W: 3490 L: 3575 D: 8973
sprt @ 10+0.1 th 1 Take 2.
19-01-11 31m combo_190110 diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 18218 W: 3971 L: 4063 D: 10184
sprt @ 10+0.1 th 1 The KDC and F tweaks don't seem to perform well together. Try dropping one at a time from the large combo. Remove KDC from the combo.
19-01-11 31m fawn3 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 8978 W: 1939 L: 2059 D: 4980
sprt @ 10+0.1 th 1 A different take. Enemy pawns in our king ring on the A or H files, which are neither attacked by our pawns nor weak. S(30, 30).
19-01-11 31m combo_190110 diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 17360 W: 3821 L: 3917 D: 9622
sprt @ 10+0.1 th 1 Merge new master and bugfix--I incorrectly implemented @snicolet's fawn2 (due to a mistake on my part when manually fixing a merge conflict with an earlier combo).
19-01-09 31m fawn diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 76128 W: 16610 L: 16402 D: 43116
sprt @ 10+0.1 th 1 Half effect for the a-file.
19-01-10 31m fawn2 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 24131 W: 5276 L: 5321 D: 13534
sprt @ 10+0.1 th 1 Try again, ignoring castling rights. Half penalty regardless of king position along the back rank, full penalty if on the outer two files on the same side as the fawn pawn.
19-01-10 31m combo_KDC_F diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 18585 W: 4022 L: 4113 D: 10450
sprt @ 10+0.1 th 1 Just the two most recent and most promising [0, 4]s, both by @snicolet.
19-01-10 31m combo_190110 diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 18012 W: 3969 L: 4062 D: 9981
sprt @ 10+0.1 th 1 Combination of eight promising tweaks. I have erred on the side of inclusion for anything that looked promising in recent (since Dec. 14) combo or standalone attempts.
19-01-10 31m fawn2 diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 9452 W: 2001 L: 2119 D: 5332
sprt @ 10+0.1 th 1 I have been asked to try to maintain symmetry here--yet the asymmetric version (1/2 effect for queenside) is my best so far. Maybe the key is king position after all--we are more likely to be castled or able to castle kingside. Full effect if on the outer two files, half effect if able to castle there.
19-01-10 31m hinderBishop diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 10142 W: 2182 L: 2296 D: 5664
sprt @ 10+0.1 th 1 The opposite: knight only.
19-01-09 31m hinderBishop diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 17422 W: 3727 L: 3806 D: 9889
sprt @ 10+0.1 th 1 Much closer to @ElbertoOne's test. However, knights are harder to block than bishops--so I wonder if this penalty ought to only be applied to bishops.