Stockfish Testing Queue

Finished - 1777 tests

18-02-09 pro ps_inlinepickbest diff
LLR: -0.21 (-2.94,2.94) [0.00,5.00]
Total: 3094 W: 689 L: 683 D: 1722
sprt @ 10+0.1 th 1 not a simplification per se, but testing an inlined pick_best moved into MovePicker for comparison against my ps_pickbest10.
18-02-09 pro ps_inlinepickbest diff
LLR: -0.54 (-2.94,2.94) [-3.00,1.00]
Total: 2692 W: 611 L: 647 D: 1434
sprt @ 10+0.1 th 1 Compare a simple inlined pick_best moved into MovePicker with my ps_pickbest10 branch (consolidate all pick_best segments). Did I do this right?
18-02-08 pro ps_pickbest10 diff
LLR: -2.95 (-2.94,2.94) [-3.00,1.00]
Total: 18813 W: 4117 L: 4320 D: 10376
sprt @ 10+0.1 th 1 fixed bug. non-functional simplification. consolidate all possible stages.
18-02-08 pro ps_qrecap_ttm diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 9969 W: 2155 L: 2237 D: 5577
sprt @ 10+0.1 th 1 If the ttMove recaptures the recapture square in QSEARCH_RECAP, return it first. Consistent behavior as with other stages.
18-02-08 pro ps_pickbest10 diff
LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 25053 W: 5584 L: 5469 D: 14000
sprt @ 10+0.1 th 1 simplification. Consolidated pick_best segments.
18-02-07 pro ps_skipQuiets diff
LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 38757 W: 6732 L: 6641 D: 25384
sprt @ 60+0.6 th 1 simplification. LTC. If skipQuiets is set, don't keep returning quiet moves. . . even if they are valued above zero.
18-02-06 pro ps_skipQuiets diff
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 34780 W: 7680 L: 7584 D: 19516
sprt @ 10+0.1 th 1 if skipQuiets is set, don't keep doing quiets. . even if they are positive.
18-02-06 pro ps_pickbest6 diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 9532 W: 2044 L: 2129 D: 5359
sprt @ 10+0.1 th 1 testing a pick_best alternative.
18-02-03 pro ps_1bishoppawns diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 3349 W: 732 L: 849 D: 1768
sprt @ 10+0.1 th 1 stage 1: only apply pawn-on-bishop-color penalty if we have only one bishop.
18-02-02 pro ps_1bishoppawns diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 3248 W: 679 L: 795 D: 1774
sprt @ 10+0.1 th 1 1 bishop pawn rules. Penalty for pawns on our own square only when score is above X, and bonus for enemy pawns on that same color.
18-02-02 pro ps_recaptures diff
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 20231 W: 3533 L: 3411 D: 13287
sprt @ 60+0.6 th 1 simplification. updated master. LTC. if we're only looking at captures that recapture the recapture square, scoring and pick_besting may not be necessary.
18-01-31 pro ps_qsearch diff
LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 99755 W: 22158 L: 22192 D: 55405
sprt @ 10+0.1 th 1 simplification: unwind some logic in qsearch stages.
18-01-31 pro ps_recaptures diff
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 85583 W: 18978 L: 18983 D: 47622
sprt @ 10+0.1 th 1 simplification. Since we only look at captures that recapture the recapture square, scoring and pick_besting may not be necessary.
18-01-31 pro ps_killers diff
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 42802 W: 9615 L: 9535 D: 23652
sprt @ 10+0.1 th 1 simplification. combine stages for code reuse. Ugly code, testing concept.
18-01-30 pro ps_capkillers diff
LLR: -0.93 (-2.94,2.94) [0.00,5.00]
Total: 1169 W: 250 L: 286 D: 633
sprt @ 10+0.1 th 1 if a "good capture" is a killer move, just skip it (we do killers right afterwards).
18-01-30 pro ps_killers diff
LLR: -2.96 (-2.94,2.94) [-3.00,1.00]
Total: 66500 W: 14555 L: 14852 D: 37093
sprt @ 10+0.1 th 1 simplification. combine killers.
18-01-28 pro ps_goto3 diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 11567 W: 2508 L: 2627 D: 6432
sprt @ 10+0.1 th 1 Remove goto's and labels. non-functional simplification. Testing [0,4] as requested.
18-01-27 pro ps_goto3 diff
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 18137 W: 4101 L: 3972 D: 10064
sprt @ 10+0.1 th 1 simplification. try#2. remove goto's and labels.
18-01-27 pro ps_goto diff
LLR: -2.95 (-2.94,2.94) [-3.00,1.00]
Total: 24250 W: 5284 L: 5497 D: 13469
sprt @ 10+0.1 th 1 simplification. Remove goto. Even a tiny bit faster on my machine.
18-01-26 pro ps_pawnislands diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 10080 W: 2231 L: 2313 D: 5536
sprt @ 10+0.1 th 1 Try #2: Only apply island penalty if we have the same # of pawns as our opponent.
18-01-26 pro ps_pawnislands diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 21039 W: 4704 L: 4731 D: 11604
sprt @ 10+0.1 th 1 penalty for more pawn islands. Try #1.
18-01-19 pro ps_curmoves diff
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 150110 W: 26772 L: 26869 D: 96469
sprt @ 10+0.1 th 1 simplification. Initialize cur (moves) in constructor, not in switch statement.
18-01-16 pro ps_captures diff
LLR: -2.95 (-2.94,2.94) [-3.00,1.00]
Total: 11028 W: 1890 L: 2065 D: 7073
sprt @ 10+0.1 th 1 simplification. Piece value seems to have little effect here.
18-01-12 pro ps_dualsort7 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 22964 W: 4108 L: 4136 D: 14720
sprt @ 10+0.1 th 1 last dual sort. Dual sort those moves over the limit from ends working inward(less copying).
18-01-05 pro ps_dualsort4 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 19256 W: 3446 L: 3489 D: 12321
sprt @ 10+0.1 th 1 partial dual sort. Dual sort, but only on the moves that are over the limit.
17-12-31 pro ps_pickbest5 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 13775 W: 2537 L: 2603 D: 8635
sprt @ 10+0.1 th 1 Try #2. Only do pick_best at depth 1. Prio. -1
17-12-27 pro ps_pickworst diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 8860 W: 1524 L: 1612 D: 5724
sprt @ 10+0.1 th 1 after an aggressive search limit (-8K*depth), begin picking the worst moves to force early cutoff.
17-12-15 pro ps_pickbest3 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 11521 W: 2050 L: 2126 D: 7345
sprt @ 10+0.1 th 1 use pick_best for moves past the sort limit.
17-12-06 pro ps_activeKing diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 4952 W: 858 L: 962 D: 3132
sprt @ 10+0.1 th 1 first try. when enemy queen is dead, pressure king to center or ave of pawn positions.
17-11-30 pro ps_moreLazy diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 79308 W: 14386 L: 14174 D: 50748
sprt @ 10+0.1 th 1 scale LazyThreshold. 1450 + 10 * depth when depth > 0.
17-11-23 pro ps_time diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 11094 W: 1997 L: 2075 D: 7022
sprt @ 10+0.1 th 1 gradually increase optimumTime if we are getting behind in score.
17-10-28 pro ps_quitsort diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 12952 W: 2281 L: 2351 D: 8320
sprt @ 10+0.1 th 1 don't sort (except top move) if there are no decent moves. -6k.
17-09-30 pro ps_custom diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 5971 W: 1047 L: 1147 D: 3777
sprt @ 10+0.1 th 1 testing some custom search limits. -2500, -9k, -12k, . . . -24k. Prio: -1
17-09-14 pro ps_sigmoids diff
LLR: -1.60 (-2.94,2.94) [0.00,5.00]
Total: 12132 W: 2204 L: 2220 D: 7708
sprt @ 10+0.1 th 1 sigmoids. Very strong at 10+.1, and 5+.05 games. range -22k, scale 1, depth-5. See spreadsheet in forum.
17-09-01 pro ps_sort_custom diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 16078 W: 2925 L: 2981 D: 10172
sprt @ 10+0.1 th 1 retry last (updated master) Best sorting: ~(1000<<depth) when depth < 5, otherwise sort all roughly (10-bit). Priority -1.
17-08-23 pro ps_sort_custom diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 23254 W: 4114 L: 4141 D: 14999
sprt @ 10+0.1 th 1 local machines converting to: (575+time*2)<<depth
17-08-19 pro ps_sort_custom diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 46709 W: 8374 L: 8302 D: 30033
sprt @ 10+0.1 th 1 balancing between 7+.07 and 10+.1. (700+time)<<depth.
17-08-17 pro ps_sort_custom diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 39063 W: 7011 L: 6971 D: 25081
sprt @ 10+0.1 th 1 stage 2: converging to ~(128<<((time>>7)+depth)
17-08-17 pro ps_sort_custom diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 24656 W: 4430 L: 4450 D: 15776
sprt @ 10+0.1 th 1 stage 2: dynamically calculate sort limit: 1024 - 15*depth*optimumTime
17-08-16 pro ps_sort_zeros diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 16192 W: 2895 L: 2951 D: 10346
sprt @ 10+0.1 th 1 1024<<depth when depth < 5, otherwise sort all at 8-bit resolution.
17-08-16 pro ps_sort_zeros diff
LLR: -1.38 (-2.94,2.94) [0.00,5.00]
Total: 33280 W: 5849 L: 5769 D: 21662
sprt @ 10+0.1 th 1 limit = 1024<<depth, for depth < 5. otherwise, sort everything roughly.
17-08-15 pro ps_sort_optimize diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 9258 W: 1609 L: 1695 D: 5954
sprt @ 10+0.1 th 1 testing convergence to +1k, -4K*depth.
17-08-14 pro ps_sort_half diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 5250 W: 898 L: 1001 D: 3351
sprt @ 10+0.1 th 1 optimizing number of moves per depth, instead of sort limit (depth).
17-08-11 pro ps_sort_hybrid diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 16422 W: 2896 L: 2952 D: 10574
sprt @ 10+0.1 th 1 simplified sorting when depth > 5 (at that depth we usually sort everything anyways).
17-08-10 pro ps_optimize diff
LLR: -2.02 (-2.94,2.94) [0.00,5.00]
Total: 15882 W: 2863 L: 2881 D: 10138
sprt @ 10+0.1 th 1 Slight improvement locally. VERY IMPORTANT: 10+0.1 games only.
17-08-09 pro ps_optimize diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 7304 W: 1283 L: 1377 D: 4644
sprt @ 10+0.1 th 1 test value convergence for 10+.1 games.
17-08-09 pro ps_sort_custom diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 6840 W: 1202 L: 1298 D: 4340
sprt @ 10+0.1 th 1 stage 1: optimizing sort limits for 10+0.1 games.
17-08-07 pro ps_limits diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 2590 W: 422 L: 537 D: 1631
sprt @ 10+0.1 th 1 shift sort limit calculations down to depth 3+.
17-08-07 pro ps_sort_half diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 4061 W: 685 L: 793 D: 2583
sprt @ 10+0.1 th 1 instead of limits, always sort 1/2 of moves.
17-08-05 pro ps_sort_hybrid diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 10714 W: 1867 L: 1947 D: 6900
sprt @ 10+0.1 th 1 hybrid sorting at 8 bit resolution.