Stockfish Testing Queue

Pending - 0 tests 0.0 hrs

None

Active - 0 tests

Finished - 221 tests

18-04-10 xot dblpawn1 diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 116008 W: 23576 L: 23306 D: 69126
sprt @ 10+0.1 th 1 larger mg increase (24,38)
18-04-10 xot dblpawn1 diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 46377 W: 6943 L: 6958 D: 32476
sprt @ 60+0.6 th 1 LTC. Doesn't look like the other values will be any better, see what ltc does. Increase mg penalty for doubled pawns (22,38)
18-04-10 xot dblpawn1 diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 36918 W: 7543 L: 7566 D: 21809
sprt @ 10+0.1 th 1 smaller mg increase (20,38)
18-04-10 xot dblpawn1 diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 22429 W: 4567 L: 4644 D: 13218
sprt @ 10+0.1 th 1 middle mg increase seems best, add eg increase: (22,40)
18-04-10 xot dblpawn1 diff
LLR: 2.97 (-2.94,2.94) [0.00,5.00]
Total: 20019 W: 4156 L: 3934 D: 11929
sprt @ 10+0.1 th 1 increase mg penalty for doubled pawns
18-04-08 xot outpost_tune diff
19721/20000 iterations
40000/40000 games played
40000 @ 20+0.2 th 1 tune outpost values
18-04-08 xot outpost5 diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 18250 W: 3686 L: 3779 D: 10785
sprt @ 10+0.1 th 1 Test tuned outpost values.
18-04-07 xot outpost3 diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 6434 W: 1218 L: 1317 D: 3899
sprt @ 10+0.1 th 1 Try to speed up outpost eval. bench now matches master (thanks Rocky!). Local parallel bench suggests ~0.7% speedup, do STC to check if actual results are better or worse.
18-04-07 xot outpost3 diff
LLR: -0.12 (-2.94,2.94) [0.00,5.00]
Total: 27243 W: 5555 L: 5434 D: 16254
sprt @ 10+0.1 th 1 Try to speed up outpost eval. Local parallel bench suggests 0.66% to 1.5% speedup. But, bench has changed, not sure if this is expected. Try STC to see if/how actual results are affected.
18-04-06 xot queenatk1 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 10399 W: 2093 L: 2173 D: 6133
sprt @ 10+0.1 th 1 (0,12) bonus, rebased on latest master
18-04-06 xot queenatk1 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 13716 W: 2790 L: 2855 D: 8071
sprt @ 10+0.1 th 1 (0,24) bonus
18-04-06 xot queenatk1 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 31676 W: 6481 L: 6462 D: 18733
sprt @ 10+0.1 th 1 only endgame bonus
18-04-05 xot connect1 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 12713 W: 2535 L: 2605 D: 7573
sprt @ 10+0.1 th 1 modify connectivity to give bonus for attacking opponent's pieces as well as defending ours
18-04-04 xot queenatk1 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 15496 W: 3153 L: 3209 D: 9134
sprt @ 10+0.1 th 1 bonus for attacks by queen
18-04-04 xot farside1 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 6124 W: 1237 L: 1338 D: 3549
sprt @ 10+0.1 th 1 adjust psqt instead of using relative_rank < 4
18-04-04 xot farside1 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 9783 W: 1950 L: 2033 D: 5800
sprt @ 10+0.1 th 1 exclude our pawns
18-04-04 xot farside1 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 8243 W: 1588 L: 1678 D: 4977
sprt @ 10+0.1 th 1 only give bonus if relative_rank < 4
18-04-04 xot farside1 diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 9946 W: 1985 L: 2068 D: 5893
sprt @ 10+0.1 th 1 using existing attack list b
18-04-04 xot farside1 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 11190 W: 2259 L: 2336 D: 6595
sprt @ 10+0.1 th 1 bonus for bishops reaching opponent's half of board. use attacks_bb
18-04-03 xot farside1 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 14553 W: 2952 L: 3013 D: 8588
sprt @ 10+0.1 th 1 bonus for bishops that can see opponents half of board
18-04-02 xot complex3 diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 8888 W: 1780 L: 1867 D: 5241
sprt @ 10+0.1 th 1 use array for pawn count, probably last test in this series
18-04-02 xot complex2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 19459 W: 3922 L: 3960 D: 11577
sprt @ 10+0.1 th 1 estimate constants manually
18-04-02 xot complex2 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 9011 W: 1758 L: 1845 D: 5408
sprt @ 10+0.1 th 1 Use array instead of formula for pawnasymmetry bonus. Values from second tune.
18-04-01 xot complex2tune diff
29358/30000 iterations
59427/60000 games played
60000 @ 20+0.2 th 1 tune array values for pawn asymmetry bonus (stop at 50k) [continue previous tune]
18-04-01 xot complex2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 13277 W: 2660 L: 2727 D: 7890
sprt @ 10+0.1 th 1 Use array for pawnAsymmetry bonus values
18-04-01 xot complex2tune diff
24640/30000 iterations
50073/60000 games played
60000 @ 20+0.2 th 1 tune array values for pawn asymmetry bonus (stop at 50k)
18-03-30 xot repeat1 diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 12756 W: 2543 L: 2613 D: 7600
sprt @ 10+0.1 th 1 If a few moves result in VALUE_DRAW from single repetitions, we lose the information about which move is best. Return eval/4 instead of 0.
18-03-12 xot bish1 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 12845 W: 2566 L: 2635 D: 7644
sprt @ 10+0.1 th 1 Penalty for bishop on 2nd rank with pawn in front towards centre of board.
18-03-09 xot draw2_maxply26 diff
ELO: 10.13 +-3.1 (95%) LOS: 100.0%
Total: 20000 W: 4405 L: 3822 D: 11773
20000 @ 10+0.1 th 1 Compare returning beta from qsearch (max_ply 25) with returning alpha+1 (max_ply one higher, i.e. 26)
18-03-07 xot draw2_beta_maxply24 diff
ELO: 4.50 +-2.6 (95%) LOS: 100.0%
Total: 20000 W: 3147 L: 2888 D: 13965
20000 @ 60+0.6 th 1 LTC: return beta in qsearch. MAX_PLY set to 24 in test and base. Run fixed 20,000 games to compare elo against returning alpha and alpha+1 tests. The MAX_PLY limit should be hit more in LTC exaggerating the Elo gain if there is one.
18-03-07 xot draw2_maxply24 diff
ELO: 7.40 +-2.6 (95%) LOS: 100.0%
Total: 20000 W: 3163 L: 2737 D: 14100
20000 @ 60+0.6 th 1 LTC: return alpha+1 in qsearch. MAX_PLY set to 24 in test and base. Run fixed 20,000 games to compare elo against returning alpha. The MAX_PLY limit should be hit more in LTC exaggerating the Elo gain if there is one.
18-03-08 xot draw2_maxply20_move2 diff
ELO: 5.26 +-3.2 (95%) LOS: 99.9%
Total: 20000 W: 4564 L: 4261 D: 11175
20000 @ 10+0.1 th 1 incorporate sn's returning beta on odd-numbered depth. return alpha+1 at max_ply.
18-03-08 xot draw2_maxply20_move diff
ELO: 4.40 +-3.2 (95%) LOS: 99.6%
Total: 20000 W: 4573 L: 4320 D: 11107
20000 @ 10+0.1 th 1 Meant to submit this as NumGames, not SPRT. Return alpha+1 or beta according to side_to_move. (May never return beta ... run to check.)
18-03-08 xot draw2_maxply20 diff
ELO: 8.62 +-3.2 (95%) LOS: 100.0%
Total: 20000 W: 4785 L: 4289 D: 10926
20000 @ 10+0.1 th 1 Reschedule as 20,000 games to compare with draw2_maxply20_move. returning value_draw seems wrong to me, but I doubt this is any better. try it anyway
18-03-07 xot draw2_alpha_maxply24 diff
ELO: -0.19 +-2.6 (95%) LOS: 44.3%
Total: 20000 W: 2947 L: 2958 D: 14095
20000 @ 60+0.6 th 1 LTC: return alpha instead of alpha+1. MAX_PLY set to 24 in test and base. Run fixed 20,000 games to compare elo against returning alpha+1 (will be submitted shortly). The MAX_PLY limit should be hit more in LTC exaggerating the Elo gain if there is one.
18-03-08 xot draw2_maxply20_move diff
LLR: 2.96 (-2.94,2.94) [0.00,5.00]
Total: 13570 W: 3083 L: 2884 D: 7603
sprt @ 10+0.1 th 1 fishtest is struggling because there is no work! create some! Return alpha+1 or beta according to side_to_move. (May never return beta ... run to check. And to speed fishtest up!)
18-03-06 xot draw2_maxply24 diff
ELO: 1.95 +-3.1 (95%) LOS: 89.2%
Total: 20000 W: 4142 L: 4030 D: 11828
20000 @ 10+0.1 th 1 return alpha+1 in qsearch. MAX_PLY set to 24 in test and base. Run fixed 20,000 games to compare elo against returning alpha.
18-03-06 xot draw2_alpha_maxply24 diff
ELO: 1.36 +-3.1 (95%) LOS: 80.6%
Total: 20000 W: 4129 L: 4051 D: 11820
20000 @ 10+0.1 th 1 return alpha instead of alpha+1. MAX_PLY set to 24 in test and base. Run fixed 20,000 games to compare elo against returning alpha+1 (will be submitted shortly).
18-03-06 xot draw2_beta_maxply24 diff
ELO: 0.59 +-3.1 (95%) LOS: 64.7%
Total: 20000 W: 4099 L: 4065 D: 11836
20000 @ 10+0.1 th 1 return beta in qsearch. MAX_PLY set to 24 in test and base. Run fixed 20,000 games to compare elo against returning alpha and alpha+1 tests.
18-03-05 xot draw3b diff
LLR: -2.96 (-2.94,2.94) [-3.00,1.00]
Total: 40322 W: 8622 L: 8865 D: 22835
sprt @ 10+0.1 th 1 non-regression test for change to return value from qsearch() and search() when ply>=MAX_PLY. now returns beta from search()
18-03-05 xot draw2 diff
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 43017 W: 7032 L: 6946 D: 29039
sprt @ 60+0.6 th 1 LTC. non-regression test for change to return value from qsearch when ply>=MAX_PLY. A similar change to search() may also be appropriate.
18-03-05 xot draw3 diff
LLR: -2.95 (-2.94,2.94) [-3.00,1.00]
Total: 49603 W: 10562 L: 10822 D: 28219
sprt @ 10+0.1 th 1 non-regression test for change to return value from qsearch() and search() when ply>=MAX_PLY
18-03-05 xot draw2 diff
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 23236 W: 5166 L: 5047 D: 13023
sprt @ 10+0.1 th 1 non-regression test for change to return value from qsearch when ply>=MAX_PLY. A similar change to search() may also be appropriate.
18-03-05 xot draw2_maxply20 diff
LLR: 3.02 (-2.94,2.94) [0.00,5.00]
Total: 6224 W: 1468 L: 1302 D: 3454
sprt @ 10+0.1 th 1 returning value_draw seems wrong to me, but I doubt this is any better. try it anyway
18-03-02 xot delta1 diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 40527 W: 8809 L: 8814 D: 22904
sprt @ 10+0.1 th 1 use constants from second tune for delta (now using [0,4] bounds)
18-03-02 xot tune_delta1 diff
10002/12500 iterations
20384/25000 games played
25000 @ 10+0.1 th 1 tune initial and increment values for delta (aspiration window) - continue previous tune
18-03-02 xot delta1 diff
LLR: 0.09 (-2.94,2.94) [0.00,5.00]
Total: 2587 W: 582 L: 565 D: 1440
sprt @ 10+0.1 th 1 try tuned values for delta
18-03-02 xot tune_delta1 diff
19763/22500 iterations
40936/45000 games played
45000 @ 10+0.1 th 1 tune initial and increment values for delta (aspiration window)
18-02-28 xot psqt_tune1 diff
19049/20000 iterations
40000/40000 games played
40000 @ 10+0.1 th 1 tune pawn psqt values in centre of board (bugfix)
18-03-01 xot psqt1 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 15597 W: 3302 L: 3357 D: 8938
sprt @ 10+0.1 th 1 pawn psqt adjustments from tuning