Stockfish Testing Queue

Finished - 35494 tests

16-03-19 mbo simpler_pp diff
LLR: -2.96 (-2.94,2.94) [-3.00,1.00]
Total: 54372 W: 7072 L: 7292 D: 40008
sprt @ 60+0.6 th 1 LTC: Another PP simplification.
16-03-19 Elb tune_fmh_reductions diff
20412/40000 iterations
41374/80000 games played
80000 @ 20+0.2 th 1 Tune history, cmh and fmh separately for reductions.
16-03-20 Roc OutOfPlay diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 6067 W: 1082 L: 1182 D: 3803
sprt @ 10+0.1 th 1 Adjust attackUnits when pieces not defending our King and not attacking opponent.
16-03-20 sg lmr_weights diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 33336 W: 6063 L: 6105 D: 21168
sprt @ 10+0.1 th 1 Sorry i have entered the bench numbers in the wrong fields. In the reduction formula log(d)*log(mc) depth and move count have the same weight. That is probable not optimal. After some local testing i come up with this values.Their are selected to let average reduction unchanged.
16-03-20 fbv CagedRook053 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 8468 W: 1090 L: 1180 D: 6198
sprt @ 60+0.6 th 1 LTC. Reworked cutechess script for local tests.
16-03-20 jos space_tweak diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 5990 W: 1042 L: 1142 D: 3806
sprt @ 10+0.1 th 1 Also change condition when to evaluate space. Take 2.
16-03-20 Elb fmh_reductions diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 12765 W: 2253 L: 2324 D: 8188
sprt @ 10+0.1 th 1 Try tuned values
16-03-20 IIv piece_values diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 9285 W: 1749 L: 1875 D: 5661
sprt @ 10+0.1 th 1 Take 2: Values locally tuned by MSPSA, which is my first try of implementing Deep Learning into local/global tuner. The method tries to learn from previous iterations, and adjust ck/ak values accordingly.
16-03-20 lbr reduce_safety diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 5798 W: 985 L: 1086 D: 3727
sprt @ 10+0.1 th 1 reduce king safety with only 1 attacker
16-03-20 sni space_on_both_wings diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 3084 W: 560 L: 674 D: 1850
sprt @ 10+0.1 th 1 Space advantage on both wings by pawn pairs on 4th or 5th rank and distance 3 (like c4/f4 or a5/d5). Based on ideas of Lyudmil Tsvetkov and Stefan Geschwentner, see http://tests.stockfishchess.org/tests/view/555929170ebc5940ca5d6b79
16-03-20 fbv CagedRook053 diff
LLR: 2.95 (-2.94,2.94) [0.00,5.00]
Total: 29824 W: 5501 L: 5249 D: 19074
sprt @ 10+0.1 th 1 Reworked cutechess script for local tests.
16-03-20 sni space diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 25361 W: 4616 L: 4633 D: 16112
sprt @ 10+0.1 th 1 Use squares controlled by our pawns in the opponent side of the board in space calculation
16-03-20 n_p Lazy_tt diff
LLR: -0.01 (-2.94,2.94) [0.00,5.00]
Total: 157 W: 24 L: 24 D: 109
sprt @ 18+0.18 th 7 Run Lazy_tt with hash pressure.
16-03-19 n_p fmhRedMax diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 28769 W: 3812 L: 3835 D: 21122
sprt @ 60+0.6 th 1 LTC: Take2
16-03-19 sni space diff
LLR: -2.33 (-2.94,2.94) [-3.00,1.00]
Total: 32134 W: 5851 L: 6031 D: 20252
sprt @ 10+0.1 th 1 Try another simplification (remove weight)
16-03-19 Voy statRebound diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 6115 W: 730 L: 828 D: 4557
sprt @ 60+0.6 th 1 ltc
16-03-19 n_p fmhRedMax diff
LLR: 2.96 (-2.94,2.94) [0.00,5.00]
Total: 7218 W: 1375 L: 1219 D: 4624
sprt @ 10+0.1 th 1 Take 2:
16-03-17 Voy statRebound diff
LLR: 2.95 (-2.94,2.94) [0.00,5.00]
Total: 103093 W: 19047 L: 18480 D: 65566
sprt @ 10+0.1 th 1 stc
16-03-18 mbo simpler_pp diff
LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 67344 W: 12359 L: 12318 D: 42667
sprt @ 10+0.1 th 1 Another PP simplification.
16-03-19 Elb fmh_reductions diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 19898 W: 3603 L: 3643 D: 12652
sprt @ 10+0.1 th 1 Try with tuned values.
16-03-19 fbv CagedRook062 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 5073 W: 866 L: 970 D: 3237
sprt @ 10+0.1 th 1 Fixed formula. Tested with cutechess restart on
16-03-18 Elb piece_check_retuned3 diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 34911 W: 6326 L: 6363 D: 22222
sprt @ 10+0.1 th 1 Final try. Maybe the RookCheck parameter should be reduced. (now with correct bounds)
16-03-18 Elb fmh_reductions diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 14198 W: 2586 L: 2650 D: 8962
sprt @ 10+0.1 th 1 Combine the ideas of NicklasPersson and jcalovski for max of cmh and fmh.
16-03-18 n_p fmhRedMax diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 31419 W: 4189 L: 4203 D: 23027
sprt @ 60+0.6 th 1 LTC: Try max of cmh and fmh.
16-03-18 Mys red2 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 38753 W: 7106 L: 7065 D: 24582
sprt @ 10+0.1 th 1 Try the max idea with reductions.
16-03-18 Voy fmhTweak diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 21354 W: 3883 L: 3966 D: 13505
sprt @ 10+0.1 th 1 Make sure the 'middle' move is_ok.
16-03-18 fbv CagedRook001 diff
LLR: -1.01 (-2.94,2.94) [0.00,5.00]
Total: 19575 W: 3571 L: 3530 D: 12474
sprt @ 10+0.1 th 1 New formula. And trying to find a relation system between my local results vs framework results
16-03-15 Elb piece_check_retuned2 diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 112704 W: 15120 L: 14968 D: 82616
sprt @ 60+0.6 th 1 LTC Take 2: Test retuned QueenCheck, RookCheck, KnightCheck and BishopCheck (fixed bench)
16-03-18 fbv AlignedPawns023 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 3828 W: 652 L: 761 D: 2415
sprt @ 10+0.1 th 1 New approach
16-03-18 Elb piece_check_retuned3 diff
LLR: -0.02 (-2.94,2.94) [0.00,5.00]
Total: 178 W: 31 L: 31 D: 116
sprt @ 10+0.1 th 1 Final try. Maybe the RookCheck parameter should be reduced.
16-03-17 mbo endgame_pp2 diff
LLR: -3.53 (-2.94,2.94) [0.00,5.00]
Total: 37753 W: 6962 L: 6949 D: 23842
sprt @ 10+0.1 th 1 Raise endgame PP and material values. Take 2
16-03-17 Elb piece_check_retuned3 diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 20860 W: 3717 L: 3803 D: 13340
sprt @ 10+0.1 th 1 Take 3: Test retuned QueenCheck, RookCheck, KnightCheck and BishopCheck
16-03-17 n_p fmhRedMax diff
LLR: 2.95 (-2.94,2.94) [0.00,5.00]
Total: 23340 W: 4391 L: 4165 D: 14784
sprt @ 10+0.1 th 1 Try max of cmh and fmh.
16-03-15 Elb tune_fmh_reductions diff
39879/40000 iterations
80000/80000 games played
80000 @ 20+0.2 th 1 Tune fmh reductions parameters
16-03-17 Mys mov diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 12700 W: 2248 L: 2319 D: 8133
sprt @ 10+0.1 th 1 Maybe capture ordering can use this reversed file instead of evasion
16-03-17 sni bishop8 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 6411 W: 1109 L: 1207 D: 4095
sprt @ 10+0.1 th 1 Non-linear BishopPawns penalty
16-03-17 Mys mo diff
LLR: -1.10 (-2.94,2.94) [0.00,5.00]
Total: 2951 W: 494 L: 528 D: 1929
sprt @ 10+0.1 th 1 File based evasion scoring
16-03-15 Mys chk diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 81614 W: 14937 L: 14814 D: 51863
sprt @ 10+0.1 th 1 Think it makes sense to simply increase safe check penalty now following the refinement of the simplification patch.
16-03-17 Voy lmr-enigma diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 13131 W: 2298 L: 2368 D: 8465
sprt @ 10+0.1 th 1 Try my luck with using fmh with reduction. Divisor = HighestPossibleScore / 3. Also, limit the decrease in reduction.
16-03-15 Voy lmr diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 61527 W: 11330 L: 11191 D: 39006
sprt @ 10+0.1 th 1 Don't reduce reduction too much.
16-03-16 Elb fmh_reductions diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 36846 W: 6743 L: 6710 D: 23393
sprt @ 10+0.1 th 1 Try fmh reductions with tuned values
16-03-15 Roc BishopPawnCenter diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 33864 W: 6181 L: 6161 D: 21522
sprt @ 10+0.1 th 1 2nd try with the bug fixed version, and a more reasonable central pawn penalty (which adds up to the other penalty...)
16-03-15 jos values_spsa2 diff
LLR: 2.94 (-2.94,2.94) [0.00,4.00]
Total: 42091 W: 7779 L: 7479 D: 26833
sprt @ 10+0.1 th 1 Another SPSA2 session on outpost values. Also on RookOnFile bonus.
16-03-15 Voy statT diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 38704 W: 7034 L: 7058 D: 24612
sprt @ 10+0.1 th 1 Stat update tweak.
16-03-15 pb0 king_eval diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 28134 W: 5140 L: 5145 D: 17849
sprt @ 10+0.1 th 1 KingDanger array has MG scores only, so we can skip attackUnits-calculation in EG-positions
16-03-15 Roc BishopPawnCenter diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 7097 W: 1298 L: 1393 D: 4406
sprt @ 10+0.1 th 1 Respin of the first test which passed STC, but with a bug fix in pawns.cpp
16-03-15 mbo endgame_pp diff
LLR: -0.25 (-2.94,2.94) [0.00,5.00]
Total: 422 W: 85 L: 94 D: 243
sprt @ 10+0.1 th 1 Raise endgame PP and material values.
16-03-15 Mys fh diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 3454 W: 564 L: 674 D: 2216
sprt @ 10+0.1 th 1 History pruning fmh based tweak take 2
16-03-15 Mys union diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 22312 W: 4012 L: 4042 D: 14258
sprt @ 10+0.1 th 1 Union patch: http://tests.stockfishchess.org/tests/view/56e3a4130ebc59301a353b70 http://tests.stockfishchess.org/tests/view/56de46510ebc59301a353a63
16-03-15 sg remove_countermoves diff
LLR: -2.96 (-2.94,2.94) [-3.00,1.00]
Total: 11875 W: 2102 L: 2280 D: 7493
sprt @ 10+0.1 th 1 Test if counter moves still needed after adding fmh for move ordering. Fixed the bench issue. I forget to remove the use of killers[2] which now is non-existing and so has an undefined value.