Stockfish Testing Queue

Finished - 24198 tests

24-06-15 jo less_lmr diff
LLR: -3.05 (-2.94,2.94) [-1.50,4.50]
Total: 60021 W: 11372 L: 11322 D: 37327
sprt @ 15+0.05 th 1 Reduce LMR by one ply if previous iteration's score was a draw score. Idea from Andscacs, see http://talkchess.com/forum/viewtopic.php?t=56631
24-06-15 jo smp5 diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 10916 W: 1830 L: 1907 D: 7179
sprt @ 15+0.05 th 3 All attempts to intentionally widen the parallel search failed so far. Now let's try the opposite. Reduce search depth by one ply when splitting for non-pv nodes.
25-06-15 Vo SmartCaptures diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 16033 W: 2970 L: 3033 D: 10030
sprt @ 15+0.05 th 1 Use From-Rank to help order captures.
24-06-15 lb ttPrunePV diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 13353 W: 2478 L: 2546 D: 8329
sprt @ 15+0.05 th 1 is it really a gain to TT prune at PV nodes? enough to accept having truncated PVs from time to time?
24-06-15 Vo lmr diff
LLR: 2.97 (-2.94,2.94) [0.00,6.00]
Total: 21543 W: 3433 L: 3234 D: 14876
sprt @ 60+0.05 th 1 LMR Tweak. Increase depth if cmh>0 && history>0. LTC.
24-06-15 Vo lmr diff
LLR: 2.96 (-2.94,2.94) [-1.50,4.50]
Total: 9627 W: 1879 L: 1748 D: 6000
sprt @ 15+0.05 th 1 LMR Tweak. Increase depth if cmh>0 && history>0
24-06-15 lb captures2 diff
LLR: -3.18 (-2.94,2.94) [0.00,6.00]
Total: 25401 W: 3988 L: 3981 D: 17432
sprt @ 60+0.05 th 1 LTC for mb: Different capture order.
23-06-15 mb captures2 diff
LLR: 2.95 (-2.94,2.94) [-1.50,4.50]
Total: 20751 W: 3986 L: 3826 D: 12939
sprt @ 15+0.05 th 1 Different capture order.
23-06-15 Vo cmhWeight diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 17097 W: 3179 L: 3239 D: 10679
sprt @ 15+0.05 th 1 No cmh weight
23-06-15 jo seepruning diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 12585 W: 2278 L: 2350 D: 7957
sprt @ 15+0.05 th 1 Extend SEE-pruning by 2 plies, but more safe.
23-06-15 jo seepruning diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 4158 W: 750 L: 845 D: 2563
sprt @ 15+0.05 th 1 Adaptive SEE-pruning. Take 1
22-06-15 jo smp_seepruning diff
LLR: -3.13 (-2.94,2.94) [-1.50,4.50]
Total: 22629 W: 3812 L: 3866 D: 14951
sprt @ 15+0.05 th 3 Less SEE-pruning at split nodes. Hopefully helps at tactics.
22-06-15 Vo cmhFLBonus diff
LLR: -2.94 (-2.94,2.94) [-1.50,4.50]
Total: 14024 W: 2627 L: 2695 D: 8702
sprt @ 15+0.05 th 1 Decrease FH bonus
20-06-15 SC MVV_dynamic_OU_tuning diff
42003/30000 iterations
61000/60000 games played
60000 @ 60+0.05 th 1 MVV_dynamic_OU looked slightly better than MVV_dynamic but it looks like it is going to fail. Tune at 30s (framework is not really loaded now). If tuned version fails, I am moving to something else.
22-06-15 SC MVV_dynamic_OU_tuned diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 11520 W: 2133 L: 2208 D: 7179
sprt @ 15+0.05 th 1 Going to SPRT with almost final values.
21-06-15 My smp diff
ELO: -0.61 +-3.7 (95%) LOS: 37.3%
Total: 10866 W: 1717 L: 1736 D: 7413
20000 @ 15+0.05 th 7 Try to keep alpha as up to date as possible (smp). Test higher core scaling as 4 threads was not bad but not decisive to conclude.
21-06-15 Vo cmhWeight diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 138989 W: 26258 L: 25928 D: 86803
sprt @ 15+0.05 th 1 Decrease cmh weight to 2.
22-06-15 Ro Ranking diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 5526 W: 1008 L: 1099 D: 3419
sprt @ 15+0.05 th 1 A piece involved in many defense or attacks is assigned a higher value
21-06-15 My smp diff
ELO: 1.39 +-2.8 (95%) LOS: 83.6%
Total: 20000 W: 3378 L: 3298 D: 13324
20000 @ 15+0.05 th 4 Try to keep alpha as up to date as possible (smp).
21-06-15 Vo cmhWeight diff
LLR: -2.94 (-2.94,2.94) [0.00,4.00]
Total: 12079 W: 2180 L: 2295 D: 7604
sprt @ 15+0.05 th 1 Increase cmh weight to 4
20-06-15 jo tuned_checks diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 51033 W: 9526 L: 9505 D: 32002
sprt @ 15+0.05 th 1 Now let's try some SPSA values.
20-06-15 Vo cmhFHP2 diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 11374 W: 2089 L: 2165 D: 7120
sprt @ 15+0.05 th 1 Don't apply CMH FHP at depth 1.
20-06-15 Ro MobArea diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 6247 W: 1172 L: 1261 D: 3814
sprt @ 15+0.05 th 1 Reduce Rook mobility except on squares defending
20-06-15 Vo cmhFHP diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 2836 W: 500 L: 599 D: 1737
sprt @ 15+0.05 th 1 CMH Fail high penalty tweaks.
20-06-15 Vo Capture2 diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 1075 W: 155 L: 258 D: 662
sprt @ 15+0.05 th 1 Take-2
20-06-15 My Draws diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 902 W: 171 L: 282 D: 449
sprt @ 15+0.05 th 1 Try to address material drawish games with fewer pawns
20-06-15 SC MVV_dynamic_OU diff
LLR: -3.06 (-2.94,2.94) [-1.50,4.50]
Total: 18176 W: 3408 L: 3469 D: 11299
sprt @ 15+0.05 th 1 First shot was not that bad. Small variation: return to default values rather than to 0.
20-06-15 Ro MobArea diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 5042 W: 958 L: 1051 D: 3033
sprt @ 15+0.05 th 1 Reduce the rook mobility on its side of the board
20-06-15 My RK1 diff
LLR: -2.17 (-2.94,2.94) [-1.50,4.50]
Total: 4489 W: 856 L: 922 D: 2711
sprt @ 15+0.05 th 1 Improve rook's mobility squares
20-06-15 SC MVV_dynamic diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 9533 W: 1792 L: 1872 D: 5869
sprt @ 15+0.05 th 1 A first, very primitive attempt at collecting capture stats during play and use it for probcut stage in move picker.
19-06-15 hw nocmhfl diff
LLR: -2.95 (-2.94,2.94) [-3.00,1.00]
Total: 73795 W: 13890 L: 14179 D: 45726
sprt @ 15+0.05 th 1 Maybe we don't need the tt refutation penalty at all.
19-06-15 Vo Extend diff
LLR: -3.21 (-2.94,2.94) [-1.50,4.50]
Total: 74741 W: 14135 L: 14052 D: 46554
sprt @ 15+0.05 th 1 Don't apply bonuses at search extensions.
19-06-15 SC adjudication_1 diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 54821 W: 10740 L: 10698 D: 33383
sprt @ 10+0.05 th 1 Testing whether fishtest is sensitive to non-functional manipulations of score at short TC, see https://groups.google.com/forum/?fromgroups=#!topic/fishcooking/ClIF6fyNeow. Take 2, UCI::value /= 10.
19-06-15 Vo Capture1 diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 8928 W: 1662 L: 1744 D: 5522
sprt @ 15+0.05 th 1 Capture heuristic based on mined data [relRank][attacker][victim] Let see how this goes...
19-06-15 My KS11 diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 26438 W: 4981 L: 5017 D: 16440
sprt @ 15+0.05 th 1 King file safety.
19-06-15 Vo cmhFLBonus diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 18144 W: 2797 L: 2894 D: 12453
sprt @ 60+0.05 th 1 LTC: Give opponent's TT move a bonus, if we can't refute it.
19-06-15 jo tuned_checks diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 12691 W: 2346 L: 2459 D: 7886
sprt @ 15+0.05 th 1 Test some values found with txt framework. See also git notes.
19-06-15 SC adjudication_1 diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 422 W: 72 L: 191 D: 159
sprt @ 10+0.05 th 1 Testing whether fishtest is sensitive to non-functional manipulations of score at short TC, see https://groups.google.com/forum/?fromgroups=#!topic/fishcooking/ClIF6fyNeow. Take 1, UCI::value *= 10.
19-06-15 Vo cmhFLBonus diff
LLR: 2.95 (-2.94,2.94) [-1.50,4.50]
Total: 13210 W: 2555 L: 2415 D: 8240
sprt @ 15+0.05 th 1 Give opponent's TT move a bonus, if we can't refute it.
17-06-15 Vo cmhPP diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 81568 W: 12710 L: 12616 D: 56242
sprt @ 60+0.05 th 1 LTC
17-06-15 My OP3 diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 67532 W: 10555 L: 10502 D: 46475
sprt @ 60+0.05 th 1 Refine the outpost scope to more effective squares, hence increase the bonus slightly also. Tuning (LTC)
18-06-15 jo rooks diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 6232 W: 1145 L: 1282 D: 3805
sprt @ 15+0.05 th 1 Bonus values for rooks tuned with Texel tuning method.
18-06-15 My RKA diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 11252 W: 2073 L: 2149 D: 7030
sprt @ 15+0.05 th 1 An attempt to make better strategic use of rooks
18-06-15 Ro MVV_Z diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 7900 W: 1511 L: 1596 D: 4793
sprt @ 15+0.05 th 1 Final try. Using P<=N=B=R<=Q and cut #.
18-06-15 Vo HistoryStats2 diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 13767 W: 2522 L: 2591 D: 8654
sprt @ 15+0.05 th 1 Extra bonus if first quiet move is a killer move.
17-06-15 Ro MVV_Z diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 9623 W: 1784 L: 1864 D: 5975
sprt @ 15+0.05 th 1 True MVV and cut # Bench 7884918
17-06-15 Vo HistoryStats diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 848 W: 114 L: 217 D: 517
sprt @ 15+0.05 th 1 Only update stats if we have at least one "bad" quiet move.
17-06-15 mb captures1 diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 3854 W: 675 L: 770 D: 2409
sprt @ 15+0.05 th 1 New capture move order for mainsearch.
17-06-15 Ro MVV_Z diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 14776 W: 2737 L: 2804 D: 9235
sprt @ 15+0.05 th 1 Using Mikael data, but sorting by MVV first then by number of cuts
17-06-15 Ro MVV_Z diff
LLR: -3.13 (-2.94,2.94) [-1.50,4.50]
Total: 6714 W: 1214 L: 1308 D: 4192
sprt @ 15+0.05 th 1 My last try with that data set, but sorted by MVV then by cut% Note I considered N and B of equal value when I sorted by MVV (same in previous test)