Stockfish Testing Queue

Finished - 4995 tests

17-02-17 sg ttmove diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 26154 W: 4639 L: 4654 D: 16861
sprt @ 10+0.1 th 1 Try now at cut nodes the first killer if its the same as the countermove
17-02-17 sg killer diff
LLR: -2.99 (-2.94,2.94) [0.00,5.00]
Total: 5754 W: 954 L: 1056 D: 3744
sprt @ 10+0.1 th 1 Try killer moves idea from following post: http://talkchess.com/forum/viewtopic.php?topic_view=threads&p=705589&t=63090
17-02-15 sg ttmove diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 6547 W: 1119 L: 1216 D: 4212
sprt @ 10+0.1 th 1 if no ttmove was found try a good killer instead
17-02-15 sg simplify_iid diff
LLR: -2.95 (-2.94,2.94) [-3.00,1.00]
Total: 8060 W: 1333 L: 1502 D: 5225
sprt @ 10+0.1 th 1 Remove node type and eval condition.
17-02-15 sg iid diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 13900 W: 2477 L: 2543 D: 8880
sprt @ 10+0.1 th 1 Do IID only if no tt hit or IID depth > tt depth
17-02-14 sg fmh_update_stats diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 9768 W: 1723 L: 1807 D: 6238
sprt @ 10+0.1 th 1 Half update weight
17-02-14 sg fmh_update_stats diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 20610 W: 3693 L: 3731 D: 13186
sprt @ 10+0.1 th 1 Update also fmh1 stats for the case if we change the order of the current and last own move.
17-02-12 sg initiative diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 8513 W: 1478 L: 1567 D: 5468
sprt @ 10+0.1 th 1 less weight for game phase
17-02-12 sg initiative diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 13829 W: 2457 L: 2524 D: 8848
sprt @ 10+0.1 th 1 Add game phase
17-02-11 sg cmh2 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 19572 W: 3450 L: 3493 D: 12629
sprt @ 10+0.1 th 1 higher weight
17-02-11 sg cmh2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 12246 W: 2138 L: 2211 D: 7897
sprt @ 10+0.1 th 1 quarter weight
17-02-10 sg cmh2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 31738 W: 5649 L: 5640 D: 20449
sprt @ 10+0.1 th 1 I have done some experiments with my old cmh2 idea. First try half weight for cmh2.
17-02-08 sg lmr diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 23862 W: 4312 L: 4336 D: 15214
sprt @ 10+0.1 th 1 Try the opposite and subtract a small linear move count term. Cap value to avoid going negative.
17-02-08 sg lmr diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 22438 W: 4015 L: 4045 D: 14378
sprt @ 10+0.1 th 1 Half factor.
17-02-08 sg lmr diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 8233 W: 1391 L: 1481 D: 5361
sprt @ 10+0.1 th 1 Add small linear move count term to reduction formula.
17-02-06 sg nullmove diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 16365 W: 2869 L: 2925 D: 10571
sprt @ 10+0.1 th 1 Slightly lower eval bound for null move pruning
17-02-06 sg nullmove diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 21754 W: 3937 L: 3969 D: 13848
sprt @ 10+0.1 th 1 Slightly higher eval bound for null move pruning
17-02-05 sg nmp2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 2863 W: 442 L: 554 D: 1867
sprt @ 10+0.1 th 1 not in endgame. This solves also the mate in 21.
17-02-05 sg nmp2 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 13974 W: 2473 L: 2539 D: 8962
sprt @ 10+0.1 th 1 Also a piece have to be captured. This version also finds the mate in the position of my last test.
17-02-05 sg nmp2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 14099 W: 2510 L: 2575 D: 9014
sprt @ 10+0.1 th 1 No nullmove pruning if in last position, where you also at move no check occurs. This help to find a mate in 21 in following position: 8/p3P3/rp6/kp6/1p4PK/1P6/1P5p/4r2R w - - 0 1
17-02-04 sg nmp diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 10897 W: 1874 L: 1953 D: 7070
sprt @ 10+0.1 th 1 ok, try the three last positions
17-02-04 sg nmp diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 11688 W: 2061 L: 2137 D: 7490
sprt @ 10+0.1 th 1 no nullmove pruning if the last two positions (with same side to move) are in check
17-02-03 sg nmp diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 9664 W: 1720 L: 1804 D: 6140
sprt @ 10+0.1 th 1 no nullmove pruning if last position was in check
17-02-02 sg rookMob diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 3316 W: 548 L: 659 D: 2109
sprt @ 10+0.1 th 1 Retry first version but with MobilityBonus compensation (based on measurements)
17-02-01 sg null_history diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 22382 W: 3989 L: 4019 D: 14374
sprt @ 10+0.1 th 1 Try the opposite and relax the condition eval >= beta if eval is only a little bit smaller than beta and history value is under the 1% highest value to allow more pruning
17-02-01 sg null_history diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 11276 W: 1960 L: 2038 D: 7278
sprt @ 10+0.1 th 1 Do no nullmove pruning for about the 1% lowest values for depth >= 13
17-02-01 sg null_history diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 8862 W: 1500 L: 1588 D: 5774
sprt @ 10+0.1 th 1 Adjusting the move reduction depth shows no success. So try another usage of the nullmove history stats. Do no nullmove pruning for about the 1% lowest values.
17-01-31 sg rookMob diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 7719 W: 1347 L: 1439 D: 4933
sprt @ 10+0.1 th 1 includes squares with opponent non-pawn pieces
17-01-31 sg rookMob diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 5389 W: 952 L: 1055 D: 3382
sprt @ 10+0.1 th 1 My try on rook mobility. Count by knight or bishop attacked squares if defended.
17-01-31 sg null_history diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 20463 W: 3707 L: 3745 D: 13011
sprt @ 10+0.1 th 1 Try negative correction
17-01-31 sg null_history diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 16668 W: 3011 L: 3065 D: 10592
sprt @ 10+0.1 th 1 no correction
17-01-31 sg null_history diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 7118 W: 1217 L: 1312 D: 4589
sprt @ 10+0.1 th 1 Collect nullmove history stats indexed by last move to adjust reduction like in LMR
17-01-29 sg passed_pawns diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 20956 W: 3783 L: 3868 D: 13305
sprt @ 10+0.1 th 1 Also score passed pawns with friendly pawn in front like candidate passers (move condition from pawns.cpp to evaluate.cpp).
17-01-29 sg move_order diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 4265 W: 746 L: 854 D: 2665
sprt @ 10+0.1 th 1 Bonus for checks in quiet move ordering
17-01-28 sg mcp diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 18093 W: 3223 L: 3271 D: 11599
sprt @ 10+0.1 th 1 Try additive formula
17-01-28 sg mcp diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 9811 W: 1692 L: 1776 D: 6343
sprt @ 10+0.1 th 1 Try only a quarter effect size
17-01-28 sg mcp diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 8352 W: 1434 L: 1524 D: 5394
sprt @ 10+0.1 th 1 Move count pruning depending on distance to PV
17-01-25 sg smp_nullmove diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 10648 W: 1696 L: 1777 D: 7175
sprt @ 5+0.05 th 7 Disable null move pruning for some threads if at least 7 threads are used.
17-01-25 sg double_pawn_threat diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 6760 W: 1158 L: 1254 D: 4348
sprt @ 10+0.1 th 1 Bonus for the threat of creating an opponent passed pawn.
17-01-24 sg passed_pawns2a diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 21044 W: 3718 L: 3803 D: 13523
sprt @ 10+0.1 th 1 Quarter bonus for both was neutral and quarter bonus only for endgame seems bad (current running test). So try quarter bonus only for midgame. (Park with prio -1). Schedule test now with the correct SPRT[0,4].
17-01-24 sg passed_pawns2a diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 17761 W: 3128 L: 3224 D: 11409
sprt @ 10+0.1 th 1 Quarter bonus only for endgame. Test against my passed patch.
17-01-23 sg passed_pawns3 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 14790 W: 1864 L: 1934 D: 10992
sprt @ 60+0.6 th 1 Speculative LTC: we have a yellow result at STC and because we now check the whole path for safety i hope for a better scaling in deep searches. Low throughput Now check each square on promotion path for safety for candidate passers. Test against my passed patch.
17-01-23 sg passed_pawns3 diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 48353 W: 8759 L: 8679 D: 30915
sprt @ 10+0.1 th 1 Now check each square on promotion path for safety for candidate passers. Test against my passed patch.
17-01-22 sg passed_pawns2a diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 49659 W: 8960 L: 8948 D: 31751
sprt @ 10+0.1 th 1 Try only quarter bonus for candidate passers. Test against my passed patch.
17-01-21 sg passed_pawns2 diff
LLR: 2.95 (-2.94,2.94) [0.00,5.00]
Total: 77576 W: 10387 L: 10014 D: 57175
sprt @ 60+0.6 th 1 LTC: Half bonus if two pushes are necessary to become passed.
17-01-20 sg passed_pawns2 diff
LLR: 2.96 (-2.94,2.94) [0.00,5.00]
Total: 53569 W: 9925 L: 9570 D: 34074
sprt @ 10+0.1 th 1 Half bonus if two pushes are necessary to become passed.
17-01-20 sg passed_pawns diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 31245 W: 5716 L: 5707 D: 19822
sprt @ 10+0.1 th 1 Try for my first STC passed version a half bonus.
17-01-19 sg passed_pawns diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 127680 W: 16665 L: 16369 D: 94646
sprt @ 60+0.6 th 1 LTC: A more compact and general version which includes the case of 2 attackers and 2 defenders. Bench is the same but their differs for my 1000 position file.
17-01-20 sg passed_pawns2 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 22543 W: 3999 L: 4029 D: 14515
sprt @ 10+0.1 th 1 LTC of my other version struggles. So test a even more general version which also checks two pawn pushes to passed state.
17-01-19 sg passed_pawns diff
LLR: 0.95 (-2.94,2.94) [0.00,5.00]
Total: 34338 W: 6310 L: 6123 D: 21905
sprt @ 10+0.1 th 1 Treat pawns which are able to become passed like real passed pawns