Stockfish Testing Queue

Finished - 30366 tests

13-09-25 rst passed_eval diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 10883 W: 2065 L: 2142 D: 6676
sprt @ 15+0.05 th 1 decrease passed pawn bonus on uncertanty margins
13-09-25 jos nooutpost diff
ELO: -4.59 +-3.0 (95%) LOS: 0.1%
Total: 20000 W: 3821 L: 4085 D: 12094
20000 @ 15+0.05 th 1 Let's check the impact of outposts.
13-09-25 rap static_null_B diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 22014 W: 4232 L: 4279 D: 13503
sprt @ 15+0.05 th 1 Use qsearch in static null move pruning
13-09-25 pec space diff
LLR: 0.74 (-2.94,2.94) [-1.50,4.50]
Total: 29184 W: 5329 L: 5228 D: 18627
sprt @ 30+0.05 th 1 Retune. Local results were much better than framewortk at 15+0.05. I suspect that cpu scaling might be responsible. With higher tc performance seems to increase. So I retest at 30+0.05.
13-09-25 jos undef_squares_1 diff
LLR: 0.08 (-2.94,2.94) [-1.50,4.50]
Total: 104 W: 22 L: 19 D: 63
sprt @ 15+0.05 th 1 Introducing undefended squares. First try. I have no clue of how big the penalty should be. And if more or less for endgame. I'm starting with (10,10). Suggestions welcome.
13-09-25 rst ext_remove diff
ELO: -9.63 +-3.0 (95%) LOS: 0.0%
Total: 20000 W: 3720 L: 4274 D: 12006
20000 @ 15+0.05 th 1 Let's measure the value of extensions (except singular extensions)
13-09-25 pec search_tweaks diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 60504 W: 11667 L: 11611 D: 37226
sprt @ 15+0.05 th 1 Several search tweaks
13-09-26 rst ext_remove diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 40142 W: 7743 L: 7741 D: 24658
sprt @ 15+0.05 th 1 remove castle moves from dangerous condition
13-09-26 pec space diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 19027 W: 3082 L: 3139 D: 12806
sprt @ 60+0.05 th 1 Retune. Instead of continuing stage I at 30+0.05 I am rescheduling it at 60+0.05, as doubling of TC showed increase of 3-4 elo and it it is of interest if performance will be even better at LTC
13-09-26 jos undef_squares_1 diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 9318 W: 1793 L: 1874 D: 5651
sprt @ 15+0.05 th 1 Introducing undefended squares. First try. I have no clue of how big the penalty should be. And if more or less for endgame. I'm starting with (10,10). Suggestions welcome. (Fixed a silly mistake.)
13-09-26 lbr reduction diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 6391 W: 1222 L: 1311 D: 3858
sprt @ 15+0.05 th 1 remove floor from LMR formula
13-09-26 uri more_lmr diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 12283 W: 2394 L: 2467 D: 7422
sprt @ 15+0.05 th 1 doing more lmr by reducing else from the code
13-09-26 lbr ortho_ext diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 35945 W: 6910 L: 6920 D: 22115
sprt @ 15+0.05 th 1 orthodox check extension
13-09-26 uri passed_pawn_try diff
LLR: -2.94 (-2.94,2.94) [-1.50,4.50]
Total: 11777 W: 2266 L: 2340 D: 7171
sprt @ 15+0.05 th 1 reduce the evaluation of passed pawns in case that it is very high
13-09-26 jos lowmob diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 19952 W: 3827 L: 3879 D: 12246
sprt @ 15+0.05 th 1 Penalty for pieces with extreme low mobility. Idea taken from Little Chess Evaluation Compendium by Lyudmil Tsvetkov.
13-09-26 jos undefended2 diff
LLR: -2.98 (-2.94,2.94) [-1.50,4.50]
Total: 11894 W: 2294 L: 2369 D: 7231
sprt @ 15+0.05 th 1 Undefended squares, Take 2. Though Lucas is right that this is somewhat redundant with mobility calculation, I want to give it another try.
13-09-26 pec opp_bishops diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 12293 W: 2364 L: 2437 D: 7492
sprt @ 15+0.05 th 1 tune scaling for multi-piece opposite bishops ending 50 -> 46
13-09-26 mco global_move_array diff
ELO: -0.09 +-2.1 (95%) LOS: 46.8%
Total: 40000 W: 7539 L: 7549 D: 24912
40000 @ 15+0.05 th 1 Use a global array instead of stack to store moves: test single thread
13-09-26 mco global_move_array diff
ELO: 0.10 +-2.0 (95%) LOS: 54.0%
Total: 40000 W: 7006 L: 6994 D: 26000
40000 @ 15+0.05 th 3 Use a global array instead of stack to store moves: test SMP
13-09-26 rap razor_A diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 439 W: 37 L: 139 D: 263
sprt @ 15+0.05 th 1 Increase razor margins if there is queens on the board
13-09-26 rap razor_B diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 648 W: 82 L: 187 D: 379
sprt @ 15+0.05 th 1 Increase razor margins if there is pawns on 7th rank
13-09-26 pec space diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 9815 W: 1702 L: 1781 D: 6332
sprt @ 30+0.05 th 1 Tune space. Stage I at 30+0.05 as tc sensitive
13-09-26 rap razor_A diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 417 W: 46 L: 153 D: 218
sprt @ 15+0.05 th 1 Increase razor margins if there is queens on the board, less aggressive
13-09-26 uri change_big_eval diff
ELO: -6.62 +-3.0 (95%) LOS: 0.0%
Total: 20000 W: 3777 L: 4158 D: 12065
20000 @ 15+0.05 th 1 tuning reduce positional score by 10 percent.
13-09-26 uri change_big_eval diff
ELO: -3.09 +-3.0 (95%) LOS: 2.3%
Total: 20000 W: 3888 L: 4066 D: 12046
20000 @ 15+0.05 th 1 second try this time increase positional evaluation
13-09-27 lbr ortho_ext diff
ELO: 1.13 +-1.7 (95%) LOS: 89.7%
Total: 50000 W: 8283 L: 8121 D: 33596
50000 @ 60+0.05 th 1 orthodox check extension. framework almost empty. see how it behaves at long TC, with low priority.
13-09-27 rap lmr_check diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 3318 W: 625 L: 723 D: 1970
sprt @ 15+0.05 th 1 Don't LMR discorvered checks.
13-09-27 jos lowmob diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 31392 W: 6224 L: 6245 D: 18923
sprt @ 15+0.05 th 1 Refined version of lowmob.
13-09-27 rap q_discovered diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 1261 W: 202 L: 305 D: 754
sprt @ 15+0.05 th 1 Dont prune negative SEE discovered checks. Take 2: in qsearch.
13-09-27 uri do_not_print_pv_at_smal diff
ELO: -1.86 +-3.0 (95%) LOS: 11.1%
Total: 20000 W: 3801 L: 3908 D: 12291
20000 @ 15+0.05 th 1 check if stockfish can earn significant elo at fast time control from not printing the pv at small depths
13-09-27 rap discoverd_ext diff
LLR: -2.94 (-2.94,2.94) [-1.50,4.50]
Total: 23563 W: 4580 L: 4622 D: 14361
sprt @ 15+0.05 th 1 Extend discovered checks with negative SEE
13-09-27 rap passed_pawns_A diff
LLR: -2.94 (-2.94,2.94) [-1.50,4.50]
Total: 9253 W: 1809 L: 1890 D: 5554
sprt @ 15+0.05 th 1 Tweak passed pawns bonus. Take 1
13-09-27 rap passed_pawns_B diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 16588 W: 3174 L: 3236 D: 10178
sprt @ 15+0.05 th 1 Tweak passed pawns bonus. Take 2
13-09-27 rap passed_pawns_C diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 7853 W: 1493 L: 1578 D: 4782
sprt @ 15+0.05 th 1 Tweak passed pawns bonus. Take 3
13-09-27 rap passed_pawns_D diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 4869 W: 916 L: 1009 D: 2944
sprt @ 15+0.05 th 1 Tweak passed pawns bonus. Take 4
13-09-27 rap threat_margins diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 5659 W: 1068 L: 1159 D: 3432
sprt @ 15+0.05 th 1 Add to margins in threat evaluation
13-09-27 rst ortho_ext_try diff
LLR: -2.94 (-2.94,2.94) [-1.50,4.50]
Total: 52590 W: 10069 L: 10034 D: 32487
sprt @ 15+0.05 th 1 orthodox check extension experiment including removing castle moves from dangerous moves definition and a new try for reducing killers
13-09-27 pec space diff
LLR: 2.96 (-2.94,2.94) [-1.50,4.50]
Total: 18719 W: 3368 L: 3217 D: 12134
sprt @ 30+0.05 th 1 Try increasing.
13-09-28 pec search_tweaks diff
LLR: 2.95 (-2.94,2.94) [-1.50,4.50]
Total: 3413 W: 739 L: 623 D: 2051
sprt @ 15+0.05 th 1 Several search tweaks
13-09-28 lbr ortho_ext diff
ELO: 0.90 +-1.9 (95%) LOS: 82.0%
Total: 50000 W: 9995 L: 9866 D: 30139
50000 @ 15+0.05 th 1 orthodox check extension. framework empty again. measure now at short TC too, to get an idea of the scaling property of this patch (low priority).
13-09-28 lbr ttvalue diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 4426 W: 828 L: 923 D: 2675
sprt @ 15+0.05 th 1 get rid of TT refined eval
13-09-28 gli ortho_ext_try diff
LLR: -2.96 (-2.94,2.94) [0.00,6.00]
Total: 6645 W: 1046 L: 1117 D: 4482
sprt @ 60+0.05 th 1 Long TC speculative: orthodox check extension experiment including removing castle moves from dangerous moves definition and a new try for reducing killers
13-09-28 pec search_tweaks diff
LLR: -2.97 (-2.94,2.94) [0.00,6.00]
Total: 5856 W: 951 L: 1026 D: 3879
sprt @ 60+0.05 th 1 LTC. Several search tweaks
13-09-28 uri use_more_time diff
ELO: 1.42 +-1.9 (95%) LOS: 92.2%
Total: 40000 W: 6660 L: 6497 D: 26843
40000 @ 60+0.25 th 1 same test at longer time control (also proportinal to 120+30)
13-09-28 uri use_more_time diff
ELO: 1.54 +-2.9 (95%) LOS: 85.0%
Total: 15985 W: 2380 L: 2309 D: 11296
20000 @ 120+0.5 th 1 I may delete this patch later and I do not want people to complain about spending machine time so I started with a very small number of games when I can always increase it later.
13-09-28 mco space diff
LLR: -2.97 (-2.94,2.94) [0.00,6.00]
Total: 36702 W: 6153 L: 6080 D: 24469
sprt @ 60+0.05 th 1 LTC: Try increasing.
13-09-28 rap razor_A diff
LLR: -2.62 (-2.94,2.94) [-1.50,4.50]
Total: 34802 W: 6783 L: 6783 D: 21236
sprt @ 15+0.05 th 1 bugfix: position-depended razor margins
13-09-28 ren my_change diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 11337 W: 2256 L: 2332 D: 6749
sprt @ 15+0.05 th 1 Try a bonus for adjacent pawn on the same rank.
13-09-28 gli pawn_king diff
LLR: -2.94 (-2.94,2.94) [-1.50,4.50]
Total: 3669 W: 670 L: 766 D: 2233
sprt @ 15+0.05 th 1 Some pawn/knight related king safety tweaks
13-09-29 lbr fd5921eb6f94ffdb989cb57 diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 6117 W: 1151 L: 1241 D: 3725
sprt @ 15+0.05 th 1 simplify LMR formula: take 1