Stockfish Testing Queue

Finished - 33100 tests

13-12-07 mco search diff
ELO: -3.64 +-1.9 (95%) LOS: 0.0%
Total: 40000 W: 5920 L: 6339 D: 27741
40000 @ 60+0.05 th 1 LTC fixed games: Simplification: skip updating killers on TT hit
13-12-07 ren alpha_value diff
LLR: -2.96 (-2.94,2.94) [0.00,6.00]
Total: 29050 W: 4418 L: 4390 D: 20242
sprt @ 60+0.05 th 1 LTC: Increase alpha value take 2.
13-12-07 rst combo_search diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 46783 W: 8717 L: 8700 D: 29366
sprt @ 15+0.05 th 1 combo of search tweaks
13-12-07 ren easy_move_dp diff
LLR: -0.85 (-2.94,2.94) [-1.50,4.50]
Total: 17370 W: 3224 L: 3209 D: 10937
sprt @ 15+0.05 th 1 Depth based easy move.
13-12-07 mco search diff
LLR: -2.96 (-2.94,2.94) [0.00,6.00]
Total: 21733 W: 3312 L: 3316 D: 15105
sprt @ 60+0.05 th 1 Increase child node futility pruning threshold (test directly at 60secs because it seems TC dependant)
13-12-07 ren easy_move_dp diff
ELO: -0.94 +-2.7 (95%) LOS: 24.9%
Total: 19579 W: 3034 L: 3087 D: 13458
20000 @ 60+0.05 th 1 STC failed. Evaluate at longer tc. Low priority. Depth based easy move.
13-12-07 rst lmr_search_step diff
LLR: 2.94 (-2.94,2.94) [-1.50,4.50]
Total: 14123 W: 2715 L: 2573 D: 8835
sprt @ 15+0.05 th 1 for big LMR, if a move doesn't fail low, then do another search with depth in between reduced depth and full depth
13-12-08 pec aspiration diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 19695 W: 3653 L: 3707 D: 12335
sprt @ 15+0.05 th 1 asymmetric aspiration window change
13-12-08 lbr lmr_search_step diff
LLR: 2.97 (-2.94,2.94) [0.00,6.00]
Total: 33724 W: 5303 L: 5050 D: 23371
sprt @ 60+0.05 th 1 for big LMR, if a move doesn't fail low, then do another search with depth in between reduced depth and full depth
13-12-08 cee qs_fut diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 17428 W: 3186 L: 3246 D: 10996
sprt @ 15+0.05 th 1 Lower futilityValue and data from real games
13-12-08 mco search diff
LLR: 2.95 (-2.94,2.94) [-1.50,4.50]
Total: 11363 W: 2204 L: 2069 D: 7090
sprt @ 15+0.05 th 1 A tweak of Ralph's research at intermediate depth based on the fact that research time is not linear with depth
13-12-08 rst lmr_search_step diff
LLR: 2.96 (-2.94,2.94) [-1.50,4.50]
Total: 16944 W: 3196 L: 3047 D: 10701
sprt @ 15+0.05 th 1 lower reduction threshold for intermeditate search
13-12-08 rst search diff
LLR: 2.95 (-2.94,2.94) [0.00,6.00]
Total: 7292 W: 1195 L: 1061 D: 5036
sprt @ 60+0.05 th 1 LTC: A tweak of Ralph's research at intermediate depth based on the fact that research time is not linear with depth
13-12-08 rst lmr_search_step diff
LLR: -2.96 (-2.94,2.94) [0.00,6.00]
Total: 3163 W: 436 L: 523 D: 2204
sprt @ 60+0.05 th 1 let's experiment a little with this new idea
13-12-08 rst lmr_search_step diff
LLR: -2.02 (-2.94,2.94) [0.00,6.00]
Total: 13675 W: 2072 L: 2080 D: 9523
sprt @ 60+0.05 th 1 increase incremental depth by one ply and start with reduction >= 4 * ONE_PLY
13-12-08 mco search^ diff
ELO: 0.31 +-2.4 (95%) LOS: 60.0%
Total: 24590 W: 3787 L: 3765 D: 17038
30000 @ 60+0.05 th 1 Direct LTC match between the 2 successful intermediate reduction patches
13-12-09 pec aspiration diff
LLR: 2.95 (-2.94,2.94) [-1.50,4.50]
Total: 22361 W: 4271 L: 4107 D: 13983
sprt @ 15+0.05 th 1 kind of unity patch of 2 aspiration window patches above 50% at 60+0.05
13-12-09 sg pawns diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 21194 W: 3939 L: 3989 D: 13266
sprt @ 15+0.05 th 1 Additional penalty for doubled backward pawns
13-12-09 lbr update_on_tt_hit diff
LLR: 2.95 (-2.94,2.94) [-1.50,4.50]
Total: 52690 W: 9955 L: 9712 D: 33023
sprt @ 15+0.05 th 1 Update History and Counter move on TT hit
13-12-09 ren aspiration diff
LLR: -2.95 (-2.94,2.94) [0.00,6.00]
Total: 9758 W: 1479 L: 1536 D: 6743
sprt @ 60+0.05 th 1 LTC for pe : kind of unity patch of 2 aspiration window patches above 50% at 60+0.05
13-12-09 jos wrong_bishop_fix diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 26873 W: 4909 L: 4944 D: 17020
sprt @ 15+0.05 th 1 For most cases, this should fix the problem of not recognizing draws with wrong-coloured bishop endgame. I simply skip futility-pruning in qsearch, when we are in very late endgame. Not very elegant, but effective. And it should hardly have a noticeable drawback at this late stage of the game.
13-12-09 mco master diff
ELO: 7.94 +-1.9 (95%) LOS: 100.0%
Total: 40000 W: 6576 L: 5662 D: 27762
40000 @ 60+0.05 th 1 First regression test after SF DD release
13-12-09 lbr update_on_tt_hit diff
LLR: 2.96 (-2.94,2.94) [0.00,6.00]
Total: 5555 W: 935 L: 808 D: 3812
sprt @ 60+0.05 th 1 LTC: Update History and Counter move on TT hit
13-12-10 mco master diff
ELO: 0.03 +-2.1 (95%) LOS: 51.3%
Total: 40000 W: 7525 L: 7521 D: 24954
40000 @ 15+0.05 th 1 Verify a small fix to Lucas patch does not regress
13-12-10 lbr master diff
ELO: -1.02 +-1.9 (95%) LOS: 14.6%
Total: 40000 W: 6089 L: 6206 D: 27705
40000 @ 60+0.05 th 1 LTC: Verify a small fix to Lucas patch does not regress
13-12-10 rst lmr_inter diff
ELO: -0.03 +-2.1 (95%) LOS: 49.0%
Total: 40000 W: 7431 L: 7434 D: 25135
40000 @ 15+0.05 th 1 Simpler condition as proposed by ceebo. d==d2 to be true is a rare case (it happens in the second bench position), so it is more of a formal fix than a performance relevant one. Bench unchanged. Verify at short TC it is no regression.
13-12-10 jos wrong_bishop_fix^ diff
LLR: 0.11 (-2.94,2.94) [-1.50,4.50]
Total: 146000 W: 27183 L: 26799 D: 92018
sprt @ 15+0.05 th 1 Take 2. Increase limit for LateEndgame.
13-12-10 jos wrong_bishop_fix diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 15076 W: 2791 L: 2857 D: 9428
sprt @ 15+0.05 th 1 Take 3. Further increase limit for LateEndgame.
13-12-10 gli update_on_tt_hit diff
ELO: 1.42 +-2.7 (95%) LOS: 85.0%
Total: 19785 W: 3090 L: 3009 D: 13686
40000 @ 60+0.05 th 1 Low pri: measure elo gain of "Update History and Counter move on TT hit"
13-12-10 rst iid_tune diff
LLR: 2.95 (-2.94,2.94) [-1.50,4.50]
Total: 17792 W: 3407 L: 3255 D: 11130
sprt @ 15+0.05 th 1 try a different condition for internal iterative deepening
13-12-10 rst iid_tune diff
LLR: -2.97 (-2.94,2.94) [0.00,6.00]
Total: 11774 W: 1839 L: 1887 D: 8048
sprt @ 60+0.05 th 1 LTC: try a different condition for internal iterative deepening
13-12-11 rst threatmove_lmr diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 3273 W: 575 L: 672 D: 2026
sprt @ 15+0.05 th 1 try a new condition for threat move fail low after lmr intermediate patch
13-12-11 ren cut_node diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 9327 W: 1750 L: 1831 D: 5746
sprt @ 15+0.05 th 1 Always cut node in prob cut.
13-12-11 lbr no_singular diff
ELO: -24.37 +-7.2 (95%) LOS: 0.0%
Total: 3342 W: 509 L: 743 D: 2090
40000 @ 15+0.05 th 1 measure the value of singular extension
13-12-11 mco search diff
LLR: -2.95 (-2.94,2.94) [0.00,6.00]
Total: 7066 W: 1058 L: 1127 D: 4881
sprt @ 60+0.05 th 1 Update history and counter moves only when updating killers. Directly at LTC because it is TC sensitive.
13-12-11 Fis aspiration_db_sqrt diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 10644 W: 1973 L: 2050 D: 6621
sprt @ 15+0.05 th 1 Nonlinear modification of renouves original depth based aspiration window patch.
13-12-11 jos wrong_bishop_fix^ diff
LLR: -2.96 (-2.94,2.94) [0.00,6.00]
Total: 19108 W: 2949 L: 2964 D: 13195
sprt @ 60+0.05 th 1 LTC: Take 2. Increase limit for LateEndgame.
13-12-11 rst null_tt_store diff
LLR: -2.96 (-2.94,2.94) [0.00,6.00]
Total: 6688 W: 1018 L: 1089 D: 4581
sprt @ 60+0.05 th 1 store null move verification fail high in TT before return. LTC because works only for high depths.
13-12-11 rst lmr_inter_compare diff
ELO: -2.33 +-3.1 (95%) LOS: 7.3%
Total: 14613 W: 2220 L: 2318 D: 10075
40000 @ 60+0.05 th 1 Measure ELO: LMR intermediate search removed against master. Low prio.
13-12-12 jhe lmr_inter diff
LLR: 2.96 (-2.94,2.94) [-1.50,4.50]
Total: 6628 W: 1275 L: 1153 D: 4200
sprt @ 15+0.05 th 1 Try to save a little more time on intermediate search.
13-12-12 cee lmr_inter diff
LLR: -2.95 (-2.94,2.94) [0.00,6.00]
Total: 9457 W: 1473 L: 1531 D: 6453
sprt @ 60+0.05 th 1 LTC: Try to save a little more time on intermediate search.
13-12-13 rst ks_t diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 3433 W: 661 L: 759 D: 2013
sprt @ 15+0.05 th 1 try a different calc of attackUnits base value
13-12-13 gli master diff
ELO: 6.85 +-1.9 (95%) LOS: 100.0%
Total: 40000 W: 6572 L: 5783 D: 27645
40000 @ 60+0.05 th 1 Low pri. regression test after latest changes
13-12-13 rst lmr_adv_pp diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 5559 W: 1029 L: 1120 D: 3410
sprt @ 15+0.05 th 1 reduce advanced pawn pushes by one ply less
13-12-13 rst lmr_adv_pp diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 34468 W: 6510 L: 6524 D: 21434
sprt @ 15+0.05 th 1 do not reduce advanced pawn pushes
13-12-13 rst lmr_adv_pp diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 11086 W: 2144 L: 2220 D: 6722
sprt @ 15+0.05 th 1 reduce advanced pawn pushes by a half ply less
13-12-14 lbr update_in_check diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 3757 W: 635 L: 730 D: 2392
sprt @ 15+0.05 th 1 Update quiet move ordering when in check
13-12-14 rst losing_captures diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 3396 W: 612 L: 709 D: 2075
sprt @ 15+0.05 th 1 move ordering: treat some of the bad captures as good captures
13-12-14 mco joachim diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 1626 W: 315 L: 421 D: 890
sprt @ 15+0.05 th 1 King psqt tweak by Joachim
13-12-14 mco joachim diff
ELO: -32.56 +-3.4 (95%) LOS: 0.0%
Total: 17807 W: 3073 L: 4737 D: 9997
40000 @ 15+0.05 th 1 King psqt tweak by Joachim (fixed games)