Stockfish Testing Queue

Finished - 3079 tests

18-06-21 Hai verification2 diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 14445 W: 2378 L: 2443 D: 9624
sprt @ 60+0.6 th 1 LTC: Take 2: With tighter fail-soft.
18-06-21 pb0 offset diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 36602 W: 6164 L: 6140 D: 24298
sprt @ 60+0.6 th 1 Take 2, LTC
18-06-20 MJZ Stop_D4 diff
ELO: 1.56 +-5.2 (95%) LOS: 72.4%
Total: 6000 W: 1048 L: 1021 D: 3931
6000 @ 60+0.6 th 1 Limited test on LTC to see if it works good
18-06-20 MJZ Stop_D3 diff
ELO: -11.59 +-76.4 (95%) LOS: 38.1%
Total: 30 W: 5 L: 6 D: 19
6000 @ 60+0.6 th 1 Limited test on LTC to see if it works good
18-06-19 Fis ContemptUsOnly diff
ELO: 191.96 +-2.5 (95%) LOS: 100.0%
Total: 40000 W: 21323 L: 1227 D: 17450
40000 @ 60+0.6 th 1 Only add contempt to positions in which it's our turn to move. To be compared to http://tests.stockfishchess.org/tests/view/5b1a174b0ebc5902ab9c3fe1
18-06-19 big kingpawn_tuning2 diff
972/500000 iterations
2201/1000000 games played
1000000 @ 60+0.6 th 1 big tuning of some king and pawn eval values version 2, merged with the latest passed values and the simplification patch. 30 sec games with nt=600. will test regularly if this is going somewhere. low throughput
18-06-19 pb0 TrappedRook diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 23326 W: 3952 L: 3981 D: 15393
sprt @ 60+0.6 th 1 LTC for robal: More rook trapped penalty when we cannot castle.
18-06-19 big kingpawn_t2 diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 21917 W: 3734 L: 3817 D: 14366
sprt @ 60+0.6 th 1 speculative ltc. was the asymetric values test just a lucky run or does the raw patch really scale well? low throughput since framework is empty.
18-06-18 pro ps_connected_tune diff
46256/50000 iterations
96035/100000 games played
100000 @ 60+.06 th 1 LTC tune of connected pawns values. Fixed because some values are going out of bounds.
18-06-18 pro ps_connected_tune diff
28522/50000 iterations
58893/100000 games played
100000 @ 60+.06 th 1 LTC tune of connected pawns values.
18-06-18 big kingpawn_t2m2 diff
LLR: 2.94 (-2.94,2.94) [0.00,4.00]
Total: 43459 W: 7836 L: 7535 D: 28088
sprt @ 60+0.6 th 1 manually adjusted passed file only kingpawn_t2, test directly at LTC
18-06-18 big kingpawn_t2m diff
LLR: 2.95 (-2.94,2.94) [0.00,4.00]
Total: 37906 W: 6953 L: 6668 D: 24285
sprt @ 60+0.6 th 1 some manually adjusted values of kingpawn_t2, test directly at LTC
18-06-18 big kingpawn_t2 diff
LLR: 2.95 (-2.94,2.94) [0.00,4.00]
Total: 14618 W: 2743 L: 2537 D: 9338
sprt @ 60+0.6 th 1 values after 505k games LTC
18-06-17 big kingpawn_tuning diff
244530/256000 iterations
509229/512000 games played
512000 @ 60+0.6 th 1 big tuning of some king and pawn eval values. 30 sec games with nt=600. will test regularly if this is going somewhere. low throughput
18-06-18 big kingpawn_t1 diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 22845 W: 3964 L: 4044 D: 14837
sprt @ 60+0.6 th 1 values after 276k games LTC
18-06-17 SC scaleFactorImprovement0 diff
LLR: -2.96 (-2.94,2.94) [-3.00,1.00]
Total: 16740 W: 2750 L: 2932 D: 11058
sprt @ 60+0.6 th 1 Unify two paths, factor 6. LTC just to be sure.
18-06-17 SC scaleFactorImprovement0 diff
LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 55524 W: 9535 L: 9471 D: 36518
sprt @ 60+0.6 th 1 Unify two paths, factor 5. LTC.
18-06-16 Voy lmrCapT diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 20921 W: 3510 L: 3549 D: 13862
sprt @ 60+0.6 th 1 ver 2 See how this scales. Low tp
18-06-16 SC scaleFactorImprovement0 diff
LLR: 2.94 (-2.94,2.94) [-3.00,1.00]
Total: 125143 W: 21213 L: 21262 D: 82668
sprt @ 60+0.6 th 1 Unify two paths. LTC
18-06-16 sg check_ext2 diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 16979 W: 2886 L: 2941 D: 11152
sprt @ 60+0.6 th 1 Speculative LTC of this yellow patch because framework is empty (low throughput). No check extension for tt move.
18-06-16 big lmrcapstat17b3 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 9659 W: 1563 L: 1648 D: 6448
sprt @ 60+0.6 th 1 lmr cap stat tweak b3 LTC
18-06-15 Viz tmSF3 diff
LLR: -2.94 (-2.94,2.94) [-3.00,1.00]
Total: 41321 W: 6898 L: 7119 D: 27304
sprt @ 60+0.6 th 1 LTC for vondele since framework is almost empty - simplify time managment a bit (tuned params, rebased) (alternative form)
18-06-14 vdv tmSF3 diff
LLR: -2.95 (-2.94,2.94) [-3.00,1.00]
Total: 22396 W: 3725 L: 3916 D: 14755
sprt @ 60+0.6 th 1 LTC: Also simplify timeReduction.
18-06-14 pro ps_endgames1 diff
LLR: 0.82 (-2.94,2.94) [-3.00,1.00]
Total: 18945 W: 3290 L: 3278 D: 12377
sprt @ 60+0.6 th 1 LTC: remove a bunch of suspect endgames. For information only.
18-06-11 sg remove_second_pp_kd2 diff
LLR: -2.95 (-2.94,2.94) [-3.00,1.00]
Total: 205036 W: 29913 L: 30359 D: 144764
sprt @ 60+0.6 th 1 LTC: Also compensate weight. Rebased to current master
18-06-11 vdv tmSF diff
LLR: -2.95 (-2.94,2.94) [-3.00,1.00]
Total: 285120 W: 41316 L: 41875 D: 201929
sprt @ 60+0.6 th 1 LTC: Take 7. Intermediate tuned.
18-06-13 sg singular_update diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 6291 W: 1000 L: 1098 D: 4193
sprt @ 60+0.6 th 1 Speculative LTC for this yellow test because singular search is more triggered at higher depth (Low throughput). If singular search failed do no negative stats update for best move of singular search.
18-06-13 big SingExt5 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 70200 W: 11894 L: 11735 D: 46571
sprt @ 60+0.6 th 1 Extension Tweak 5 LTC
18-06-11 sg master diff
ELO: 29.12 +-1.8 (95%) LOS: 100.0%
Total: 40000 W: 7364 L: 4019 D: 28617
40000 @ 60+0.6 th 1 Does for comparison also a regression test with old contempt=12 against SF9 after "Optimize an expression in endgame.cpp" of June, 11th (Low throughput)
18-06-13 Hai see_depth diff
LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 17352 W: 2968 L: 2842 D: 11542
sprt @ 60+0.6 th 1 LTC: Another try at removing a depth condition.
18-06-12 31m tweak_threatOnPawn^^ diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 104513 W: 17939 L: 17753 D: 68821
sprt @ 60+0.6 th 1 Speculative LTC for this 90K yellow, which will at least reveal whether this scales well. Low throughput (166).
18-06-12 sni lmrcapt5 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 30022 W: 5019 L: 5022 D: 19981
sprt @ 60+0.6 th 1 LTC for candirufish: lmr capt stat ttcapture tweak 5
18-06-12 vdv captSF diff
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 23390 W: 4020 L: 3903 D: 15467
sprt @ 60+0.6 th 1 LTC: Remove condition.
18-06-02 SC fixStatsOverflow diff
LLR: -0.14 (-2.94,2.94) [-3.00,1.00]
Total: 213823 W: 31179 L: 31494 D: 151150
sprt @ 60+0.6 th 1 Use a more robust implementation of overflow guard, see discussion here https://github.com/Stefano80/Stockfish/commit/013b5485dd80b847ea88dd9ff08a6c572f5213c5#commitcomment-29221098 Submitting as a non-regression to check whether there is a slow-down from using std::min. Bench should change at higher depths
18-06-11 pro ps_knightbishop diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 35835 W: 6076 L: 6114 D: 23645
sprt @ 60+0.6 th 1 LTC: split up MinorBehindPawn and used separately tuned values for knight/bishop (simplified version and updated master).
18-06-11 sni master diff
ELO: 29.72 +-1.9 (95%) LOS: 100.0%
Total: 40000 W: 7748 L: 4335 D: 27917
40000 @ 60+0.6 th 1 Regression/progression test against SF9 after "Optimize an expression in endgame.cpp" of June, 11th
18-06-11 big lmrcapstat13b diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 29197 W: 4178 L: 4195 D: 20824
sprt @ 60+0.6 th 1 lmr capt stat tweak 13b LTC
18-06-11 Viz remove_second_pp_kd2 diff
LLR: -0.46 (-2.94,2.94) [0.00,5.00]
Total: 3307 W: 474 L: 481 D: 2352
sprt @ 60+0.6 th 1 LTC for sg
18-06-09 sg e4f8a4fa7f5da8287579c0c diff
LLR: -2.95 (-2.94,2.94) [-3.00,1.00]
Total: 166641 W: 24121 L: 24510 D: 118010
sprt @ 60+0.6 th 1 LTC: Now check even contempt=24 is no regression against contempt=0. last Test if this passes contempt 24 seems the best else contempt 21 is the ways to go.
18-06-10 pro ps_kpkrEND2 diff
LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 94522 W: 13838 L: 13824 D: 66860
sprt @ 60+0.6 th 1 check the relevance of distance to rook in krkp ending.
18-06-10 big captprunemargin diff
LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 27050 W: 3975 L: 3864 D: 19211
sprt @ 60+0.6 th 1 prune margin tweak run as simplification LTC
18-06-09 xor unsafePawns diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 48145 W: 7068 L: 7015 D: 34062
sprt @ 60+0.6 th 1 LTC: Penalty for unsafe pawns.
18-06-09 sg e4f8a4fa7f5da8287579c0c diff
ELO: 209.07 +-2.7 (95%) LOS: 100.0%
Total: 40000 W: 22860 L: 1328 D: 15812
40000 @ 60+0.6 th 1 LTC: Test also master with contempt=24 against SF 7.
18-06-09 can delta diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 126176 W: 18648 L: 18434 D: 89094
sprt @ 60+0.6 th 1 LTC: second try
18-06-08 sg e4f8a4fa7f5da8287579c0c diff
ELO: 196.08 +-2.6 (95%) LOS: 100.0%
Total: 40000 W: 21639 L: 1191 D: 17170
40000 @ 60+0.6 th 1 LTC: For comparison test master with default contempt against SF 7 as a weaker opponent
18-06-08 sni TCEC_11_contempt diff
ELO: 199.74 +-2.6 (95%) LOS: 100.0%
Total: 40000 W: 22130 L: 1372 D: 16498
40000 @ 60+0.6 th 1 Revert to the contempt shape and value (18) used in TCEC 11 Premier division. To be compared with Stefan's tests with current shape and value 21, which gives +203.87 +/- 2.7 against SF7.
18-06-08 sg e4f8a4fa7f5da8287579c0c diff
ELO: 203.92 +-2.6 (95%) LOS: 100.0%
Total: 40000 W: 22447 L: 1340 D: 16213
40000 @ 60+0.6 th 1 LTC: Test also master with contempt=21 against SF 7.
18-06-08 vdv tmSF^ diff
LLR: -2.95 (-2.94,2.94) [-3.00,1.00]
Total: 116083 W: 16847 L: 17164 D: 82072
sprt @ 60+0.6 th 1 LTC: Take 4. Param 1.8/2.25.
18-06-08 sg e4f8a4fa7f5da8287579c0c diff
LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 47666 W: 6914 L: 6832 D: 33920
sprt @ 60+0.6 th 1 Non-regression test with contempt=24 struggles and delivers not more elo against SF7 as contempt=21. So test contempt=21 on LTC. Now check in between contempt=21 is no regression against contempt=0
18-06-03 sg update_history diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 73412 W: 10644 L: 10585 D: 52183
sprt @ 60+0.6 th 1 LTC: For best promotion moves update quiet histories.