Stockfish Testing Queue

Finished - 52702 tests

13-06-16 uri another_reduction_try diff
LLR: -2.95 (-2.94,2.94)
Total: 21977 W: 4596 L: 4641 D: 12740
sprt @ 15+0.05 th 1 simplification of the code when I reduce less only first counter move and not second counter move(same lines of code but the new line is shorter)
13-06-16 rta SimpleEval_A1 diff
LLR: 2.96 (-2.94,2.94)
Total: 20674 W: 4505 L: 4336 D: 11833
sprt @ 15+0.05 th 1 Removes the semi-open eval element and replaced it with the rook on pawn bonus which has been expanded to attacks on file. Even if failed, I believe further inquiry would be wise.
13-06-16 mco cut_reduction diff
LLR: -2.97 (-2.94,2.94)
Total: 4307 W: 836 L: 932 D: 2539
sprt @ 15+0.05 th 1 Reduced moves shall fail-low, so a quick search is enough
13-06-16 mco cut_reduction diff
LLR: -2.96 (-2.94,2.94)
Total: 2373 W: 453 L: 555 D: 1365
sprt @ 15+0.05 th 1 Same as previous, but don't reduce twice
13-06-16 mco cut_reduction diff
LLR: -2.97 (-2.94,2.94)
Total: 3816 W: 724 L: 821 D: 2271
sprt @ 15+0.05 th 1 Another way to compensate cutNode reduction
13-06-16 tvi SkipIDForReductions diff
LLR: -2.95 (-2.94,2.94)
Total: 17571 W: 3628 L: 3686 D: 10257
sprt @ 15+0.05 th 1 Skip ID for reduced moves because we are effectively already doing a reduced
13-06-16 mco contempt diff
LLR: -2.96 (-2.94,2.94)
Total: 8571 W: 1640 L: 1700 D: 5231
sprt @ 60+0.05 th 1 Verify contempt found with CLOP after 10K games
13-06-16 rta SimpleEval_A1 diff
LLR: -2.96 (-2.94,2.94)
Total: 32738 W: 6190 L: 6123 D: 20425
sprt @ 60+0.05 th 1 Replaces semi-open file evaluation with rook on pawn bonus applied to files. Functionally, it tweaks some values to a small degree and does not give the semi-open bonus when a rook is already attacking a pawn. Also, and more importantly, is does not give the half open reward when pieces are blocking the rook such as a fellow rook or a minor piece.
13-06-17 tvi SkipIDForReductions diff
LLR: -2.97 (-2.94,2.94)
Total: 19523 W: 4057 L: 4110 D: 11356
sprt @ 15+0.05 th 1 Reduce ID reduction for reduced moves
13-06-17 jki cut2 diff
LLR: -2.97 (-2.94,2.94)
Total: 13163 W: 2735 L: 2806 D: 7622
sprt @ 15+0.05 th 1 Cut nodes: reduce 2 plies
13-06-17 mco null_cut diff
LLR: -2.96 (-2.94,2.94)
Total: 5050 W: 1016 L: 1110 D: 2924
sprt @ 15+0.05 th 1 Increase threshold of fail-low after null for cut nodes
13-06-17 mco major_on_pawn diff
LLR: 2.97 (-2.94,2.94)
Total: 57846 W: 12248 L: 11974 D: 33624
sprt @ 15+0.05 th 1 Original idea from Ryan that I have cherry-picked from the full patch
13-06-17 rta SimpleEval_B1 diff
LLR: 2.95 (-2.94,2.94)
Total: 13680 W: 2940 L: 2793 D: 7947
sprt @ 15+0.05 th 1 Attempts to remove the old Knight Check Bonus by increasing knight check bonus. Should fail quickly or pass in around 20k games.
13-06-17 tvi LMR2Phase diff
LLR: -2.95 (-2.94,2.94)
Total: 2631 W: 493 L: 593 D: 1545
sprt @ 15+0.05 th 1 Try LMR in two stages
13-06-17 gli killer_evasion diff
LLR: -2.95 (-2.94,2.94)
Total: 4128 W: 800 L: 896 D: 2432
sprt @ 15+0.05 th 1 Use killer/countermoves in check evasions
13-06-17 rta SimpleEval_B1 diff
LLR: -0.21 (-2.94,2.94)
Total: 44584 W: 8524 L: 8296 D: 27764
sprt @ 60+0.05 th 1 Test the removing of knight check bonus (punishment) while raising the bonus of knight attacks on the king. Did well in fast TC. Testing against base of master as it is the collection of now two ideas. If it fails with significantly higher losses than wins, this means that removing knight check bonus does not scale well. If it passes, my recommendation would be to push through both patches.
13-06-18 gli killer_evasion diff
LLR: -2.95 (-2.94,2.94)
Total: 21367 W: 4471 L: 4518 D: 12378
sprt @ 15+0.05 th 1 Try using just killers in evasion move searches
13-06-18 uri aspiration_change diff
LLR: -2.95 (-2.94,2.94)
Total: 36661 W: 7703 L: 7706 D: 21252
sprt @ 15+0.05 th 1 replaced delta+=delta/2 by delta+=Value(16) that is simpler and wrote the code in a way that I do not write it twice
13-06-18 tvi IDScoreResearch diff
LLR: -2.95 (-2.94,2.94)
Total: 14064 W: 2889 L: 2957 D: 8218
sprt @ 15+0.05 th 1 Relax ID condition for reduces searches and research ID fail lows for these searches
13-06-18 mco major_on_pawn diff
LLR: 2.95 (-2.94,2.94)
Total: 9181 W: 1732 L: 1581 D: 5868
sprt @ 60+0.05 th 1 Rescheduled at long TC (if it takes too long I will stop it and move to slim_master). Original idea from Ryan that I have cherry-picked from the full patch
13-06-18 rta SimpleEval_C1 diff
LLR: -2.94 (-2.94,2.94)
Total: 12527 W: 2565 L: 2637 D: 7325
sprt @ 15+0.05 th 1 Long-shot attempt to remove the ever-strange bishop outpost bonus. Builds on-top and competes against slim_master.
13-06-18 mco contact_checks diff
LLR: -2.97 (-2.94,2.94)
Total: 15262 W: 3232 L: 3297 D: 8733
sprt @ 15+0.05 th 1 Retire bishop and knight contact checks (original idea from Ryan)
13-06-19 rta SimpleEval_D1 diff
LLR: -2.96 (-2.94,2.94)
Total: 2107 W: 435 L: 540 D: 1132
sprt @ 15+0.05 th 1 Tests on-top of and against slim_master. Replaces ugly and clunky tables with more lines, but MUCH easier tuning. Regardless of result, it would be a prime target for CLOP, once it is fully functional from within the framework.
13-06-19 rta SimpleEval_E1 diff
LLR: -2.95 (-2.94,2.94)
Total: 23095 W: 4811 L: 4853 D: 13431
sprt @ 15+0.05 th 1 Removes the slightly dubious same color bishop bonus mod from within outpost bonus term. In addition, it delays dividing terms until the last moment to slightly increase accuracy.
13-06-19 tvi IDScoreResearch diff
LLR: -2.96 (-2.94,2.94)
Total: 14793 W: 3100 L: 3166 D: 8527
sprt @ 15+0.05 th 1 Restore original passing idScore < alpha condition which has a much lower signature count
13-06-19 mco contact_checks diff
LLR: -2.95 (-2.94,2.94)
Total: 43931 W: 9268 L: 9250 D: 25413
sprt @ 15+0.05 th 1 Increase bishop KingAttack weight
13-06-19 mco contact_checks diff
LLR: 2.95 (-2.94,2.94)
Total: 11674 W: 2549 L: 2407 D: 6718
sprt @ 15+0.05 th 1 Increase knight KingAttack weight
13-06-19 tvi AllFail diff
LLR: -2.95 (-2.94,2.94)
Total: 24825 W: 5177 L: 5214 D: 14434
sprt @ 15+0.05 th 1 If an ALL node is in reality a CUT node then extend
13-06-19 gli contact_checks diff
LLR: -1.96 (-2.94,2.94)
Total: 52596 W: 9667 L: 9466 D: 33463
sprt @ 60+0.05 th 1 Long TC: Increase knight KingAttack weight
13-06-19 rap BishopPawns_Max5 diff
LLR: 2.96 (-2.94,2.94)
Total: 62219 W: 13320 L: 13032 D: 35867
sprt @ 15+0.05 th 1 Limit BisopOnPawns penatalize to 5 pawns maximum
13-06-19 mco contact_checks diff
LLR: 0.64 (-2.94,2.94)
Total: 28568 W: 5401 L: 5229 D: 17938
sprt @ 60+0.05 th 1 Failed (with 50%) at low TC, but king safety is strongly TC dependent so retry at long TC. Increase bishop KingAttack weight
13-06-19 rta ThreatsRevamp_A1 diff
LLR: -2.96 (-2.94,2.94)
Total: 3759 W: 749 L: 847 D: 2163
sprt @ 15+0.05 th 1 Gives a small reward for attacking a pawn defended piece. This is patch the prepares the way for future work on the semi-open file.
13-06-19 rta SimpleEval_G1 diff
LLR: -2.96 (-2.94,2.94)
Total: 1820 W: 315 L: 417 D: 1088
sprt @ 15+0.05 th 1 Removes the limitation on rook vs pawn along with the ideas of attacking rook on file bonuses. More paving for the removing of the semi-open file bonus and what not.
13-06-19 rta SimpleEval_F2 diff
LLR: -2.95 (-2.94,2.94)
Total: 48839 W: 10285 L: 10253 D: 28301
sprt @ 15+0.05 th 1 Removes the complexity of the trapped rook term by reducing the mobility modifier.
13-06-19 mco major_on_pawn diff
LLR: -2.96 (-2.94,2.94)
Total: 7941 W: 1644 L: 1730 D: 4567
sprt @ 15+0.05 th 1 Reduce eg openfile bonus instead of adding bonus for rook on a closed file
13-06-20 uri aspiration_try diff
LLR: -2.96 (-2.94,2.94)
Total: 20079 W: 4209 L: 4260 D: 11610
sprt @ 15+0.05 th 1 tried reducing alpha after fail high. The reason is not uncommon to see cases of fail low after fail high so I hope to make the fail low faster by reducing alpha and the bench is lower
13-06-20 rap QueenOnPawn diff
LLR: -2.96 (-2.94,2.94)
Total: 31025 W: 6405 L: 6425 D: 18195
sprt @ 15+0.05 th 1 Give bonus for queen pseudo-attack pawns
13-06-20 gli less_prune_multicore diff
LLR: -2.97 (-2.94,2.94)
Total: 6735 W: 1328 L: 1417 D: 3990
sprt @ 15+0.05 th 3 Try doing no movecount pruning in nodes descendant from split points
13-06-20 gli static_cutall diff
LLR: -2.97 (-2.94,2.94)
Total: 15303 W: 3221 L: 3286 D: 8796
sprt @ 15+0.05 th 1 Try using the more "static" definition of cut/all nodes
13-06-20 gli queen_pin diff
LLR: -2.96 (-2.94,2.94)
Total: 5590 W: 1171 L: 1264 D: 3155
sprt @ 15+0.05 th 1 Give queen bonus for pin/xray as well
13-06-20 gli pinned_eval diff
LLR: -2.95 (-2.94,2.94)
Total: 3667 W: 698 L: 795 D: 2174
sprt @ 15+0.05 th 1 Don't count attacks from pinned pawns
13-06-21 uri aspiration_try diff
LLR: -2.95 (-2.94,2.94)
Total: 40631 W: 8489 L: 8481 D: 23661
sprt @ 15+0.05 th 1 reduced alpha by smaller value in this time(note that my change did not fail by a big margin and it cause me to think that reducing alpha after fail high may be a good idea) I think also to increase beta after fail low but test one change at a time
13-06-21 tvi AllFailHigh diff
LLR: -2.97 (-2.94,2.94)
Total: 18103 W: 3773 L: 3830 D: 10500
sprt @ 15+0.05 th 1 Extend any LMR that is researched in an ALL node
13-06-21 eel razorCut diff
LLR: 2.97 (-2.94,2.94)
Total: 12456 W: 2661 L: 2517 D: 7278
sprt @ 15+0.05 th 1 Reschedule with different GCC signature, that differs from my MSVC compile. Very aggresive razoring. I would like more data than I have now.
13-06-21 uri aspiration_trynew diff
LLR: -2.96 (-2.94,2.94)
Total: 18979 W: 3987 L: 4041 D: 10951
sprt @ 15+0.05 th 1 reducing alpha by Value(8) after marco's post because Value(4) may be too small.
13-06-21 mco razor_cut diff
LLR: -2.95 (-2.94,2.94)
Total: 4160 W: 824 L: 920 D: 2416
sprt @ 15+0.05 th 1 Much more aggressive razoring (from Eelco). Only the base razor_margin change.
13-06-21 gli static_cutall diff
LLR: -2.96 (-2.94,2.94)
Total: 13776 W: 2837 L: 2906 D: 8033
sprt @ 15+0.05 th 1 Try "static" cut/all definition with marco's patch to save cut/all in TT
13-06-21 mco razor_cut diff
LLR: -2.96 (-2.94,2.94)
Total: 29983 W: 6228 L: 6251 D: 17504
sprt @ 15+0.05 th 1 Lower razoring agressivness in case of ALL nodes (from Eelco): second part of the Eelco idea of asymmetric razoring according to node type.
13-06-21 mco razor_cut diff
LLR: -2.97 (-2.94,2.94)
Total: 42973 W: 9128 L: 9113 D: 24732
sprt @ 15+0.05 th 1 Asymmetric razoring take 3: simplify formula preserving more or less Eelco's margins
13-06-21 rta KingAttacking_A1 diff
LLR: -2.95 (-2.94,2.94)
Total: 898 W: 168 L: 278 D: 452
sprt @ 15+0.05 th 1 Modifies king safety with the addition of pawn attacks on king. Largest help for non-queen attacks on the king.