Stockfish Testing Queue

Finished - 56600 tests

15-01-05 Roc KingAttackersVsDefender diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 9327 W: 1851 L: 1932 D: 5544
sprt @ 15+0.05 th 1 SImplifed version, Just one parameter for attackers which are attacked.
15-01-05 roh QuickMove diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 103449 W: 20945 L: 20765 D: 61739
sprt @ 15+0.05 th 1 V5: Tune code so Quick Move occurs more frequently. Do not search for 2nd best move if 35% of available time already spent. Also Changed aspiration window in search for 2nd best move. If a QuickMove is found, Use half the time saved in the next move.
15-01-05 mco no_passed_rule diff
LLR: -2.96 (-2.94,2.94) [-3.00,1.00]
Total: 31558 W: 6378 L: 6599 D: 18581
sprt @ 15+0.05 th 1 Try to remove a possibly useless rule in passed pawn evaluation
15-01-05 jos smp3 diff
ELO: 15.28 +-31.7 (95%) LOS: 82.8%
Total: 91 W: 11 L: 7 D: 73
10000 @ 240+0.05 th 12 Less LMR with higher number of available threads. This should already kick in with 12 cores. Priority 1 to grab the 3 or 4 machines capable to run this test. Test can always be stopped if it turns out to be bad.
15-01-05 Roc KingAttackersVsDefender diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 29636 W: 5982 L: 6007 D: 17647
sprt @ 15+0.05 th 1 Try # 3: reduce the kingAttackerWeight by one for each attacked KingAttacker
15-01-05 mco score_checks diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 35842 W: 7169 L: 7177 D: 21496
sprt @ 15+0.05 th 1 Score quiet checks as we do for evasions
15-01-05 jos smp4 diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 1876 W: 288 L: 388 D: 1200
sprt @ 15+0.05 th 3 Extend at split node in the first iterations. Test with 3 threads.
15-01-05 luc phase_based_moveimporta diff
ELO: -400.63 +-7.8 (95%) LOS: 0.0%
Total: 11461 W: 76 L: 9460 D: 1925
20000 @ 15+0.05 th 1 Get rid of ply-based move_importance. More aggressive TM in midgame, account more for PV instability in endgame. Quick crash test while tuning locally.
15-01-05 Roc KingAttackersVsDefender diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 33598 W: 6716 L: 6730 D: 20152
sprt @ 15+0.05 th 1 Remove three instead.
15-01-05 lbr spsa2 diff
LLR: -2.95 (-2.94,2.94) [-0.50,4.50]
Total: 7615 W: 1440 L: 1540 D: 4635
sprt @ 15+0.05 th 1 3 param. tuned.
15-01-06 roh QuickMove diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 13739 W: 2737 L: 2806 D: 8196
sprt @ 15+0.05 th 1 Tuning quick move parameters. 1. Spending less time on quick move search.
15-01-06 nab always_ks diff
ELO: -2.07 +-2.2 (95%) LOS: 3.0%
Total: 40000 W: 7916 L: 8154 D: 23930
40000 @ 15+0.05 th 1 (Reschedule wrong base bench, sorry) Don't compute main king safety eval only when kingRing is set to 0
15-01-06 roh QuickMove diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 16783 W: 3373 L: 3434 D: 9976
sprt @ 15+0.05 th 1 Further Tuning.
15-01-06 roh QuickMove diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 11707 W: 2318 L: 2393 D: 6996
sprt @ 15+0.05 th 1 QuickMove tuning. Allot more time to move after Quick move.
15-01-06 nab always_ks diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 27724 W: 4606 L: 4671 D: 18447
sprt @ 60+0.05 th 1 LTC Don't compute main king safety eval only when kingRing is set to 0
15-01-06 luc phase_based_moveimporta diff
ELO: -6.51 +-4.3 (95%) LOS: 0.2%
Total: 10088 W: 1954 L: 2143 D: 5991
20000 @ 15+0.05 th 1 (Fixed previous bugged commit, sorry!) Get rid of ply-based move_importance. More aggressive TM in midgame, account more for PV instability in endgame. Quick crash test while tuning locally.
15-01-06 Roc KingAttackersVsDefender diff
37552/40000 iterations
80000/80000 games played
80000 @ 15+0.05 th 1 Using SPSA to calibrate relationship between numattackers and attackersattacked, if any such relationship.
15-01-07 sg spsa_pawns diff
51922/50000 iterations
99911/100000 games played
100000 @ 15+0.05 th 1 Tune pawn structure (except passed pawns and king shelter)
15-01-07 Roc AttackersCombo diff
34006/40000 iterations
80000/80000 games played
80000 @ 15+0.05 th 1 A different experiment. Look at all the different attacking pieces combinations (from 1 to 3 pieces), and see if we can improve on the current formula value (which is the starting point). The results on some extremely rare combo such as bbb, nnn or qqq are included anyway, and should not change much, This will give us some error bars on the other final values. For combos of 4 pieces and more, use the simple master formula (numpieces * sumofweights)
15-01-07 SC manual_KPP diff
LLR: -3.12 (-2.94,2.94) [-1.50,4.50]
Total: 29312 W: 5843 L: 5875 D: 17594
sprt @ 15+0.05 th 1 It seems that bench is dependent on whether I am relying on specialized endgame implementation. Try this one withouth resorting to the infrastructure.
15-01-07 SC manual_pawns diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 1416 W: 246 L: 350 D: 820
sprt @ 15+0.05 th 1 Only reward advanced pawns in pawn endgames with many pawns. Local speed-up ca. 2%.
15-01-07 sg pawns diff
LLR: 2.95 (-2.94,2.94) [-0.50,3.50]
Total: 164666 W: 33319 L: 32703 D: 98644
sprt @ 15+0.05 th 1 Test pawn structure tuned values
15-01-07 Roc AttackersCombo diff
ELO: -4.27 +-4.5 (95%) LOS: 3.0%
Total: 10000 W: 2092 L: 2215 D: 5693
10000 @ 10+0.05 th 1 Taking the SPSA run values "as is", before tuning some more
15-01-08 Mys MinMaj diff
LLR: -3.47 (-2.94,2.94) [-1.50,4.50]
Total: 10566 W: 1991 L: 2087 D: 6488
sprt @ 15+0.05 th 1 If a major is attacked by a minor, score it less as it's next move will generally be evasive.
15-01-08 luc phase_based_moveimporta diff
ELO: -2.68 +-3.1 (95%) LOS: 4.4%
Total: 20000 W: 4016 L: 4170 D: 11814
20000 @ 15+0.05 th 1 Check some hopefully better values. Maybe the idea just doesn't work this way and I'd better save resources.
15-01-08 jos less_imbalance_eg diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 7716 W: 1479 L: 1565 D: 4672
sprt @ 15+0.05 th 1 Halve material imbalance for endgame score. Many imbalance advantages vanish towards the endgame. Test this somewhat crazy idea.
15-01-08 mco official diff
ELO: 52.10 +-2.0 (95%) LOS: 100.0%
Total: 40549 W: 9948 L: 3913 D: 26688
40000 @ 60+0.05 th 1 Regression test until "Assorted formatting and comment tweaks in position.h". 2 functional patches since the last one (48.16 +-2.0).
15-01-08 SC ks_tempo diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 34973 W: 6923 L: 6934 D: 21116
sprt @ 15+0.05 th 1 Value tempo more if KS is low (and a tiny little bit less if not)
15-01-08 mco no_passed_rule diff
LLR: -2.96 (-2.94,2.94) [-3.00,1.00]
Total: 14342 W: 2816 L: 3004 D: 8522
sprt @ 15+0.05 th 1 Try to remove a possibly useless rule in passed pawn evaluation: take 2, compensate removed rule
15-01-08 Fis TTutilization diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 96467 W: 19293 L: 19134 D: 58040
sprt @ 15+0.05 th 1 Manage handing out empty TT entries in a more uniform way exploiting the fact we already know they will be saved to later. Requires proper initialization. 2MB STC
15-01-08 Fis PQ_values diff
LLR: 2.95 (-2.94,2.94) [-1.50,4.50]
Total: 9820 W: 2043 L: 1909 D: 5868
sprt @ 15+0.05 th 1 Test new pawn and queen values.
15-01-09 Fis PQ_values diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 44472 W: 7344 L: 7356 D: 29772
sprt @ 60+0.05 th 1 Test new pawn and queen values. LTC
15-01-09 sg pawns diff
LLR: -3.21 (-2.94,2.94) [0.00,4.00]
Total: 29850 W: 4948 L: 5020 D: 19882
sprt @ 60+0.05 th 1 LTC: Test pawn structure tuned values
15-01-09 SC ks_tempo_tuning diff
54772/50000 iterations
100000/100000 games played
100000 @ 15+0.05 th 1 Almost passed STC with more or less random pars, see how far we can go after tuning. Schedule 100k games, I will stop if it converges early.
15-01-09 jos imbalance_mg_eg diff
ELO: -22.71 +-2.7 (95%) LOS: 0.0%
Total: 30000 W: 6036 L: 7994 D: 15970
30000 @ 15+0.05 th 1 Do a first measure to see, if it's worth to follow this idea. Calculate material imbalance for midgame and endgame. Locally tuned values at ultrafast tc.
15-01-10 Mys KSP diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 19995 W: 3961 L: 4013 D: 12021
sprt @ 15+0.05 th 1 Make Pawn attacks on King's undefended squares also attackunits. Low pri.
15-01-10 SC ks_tempo diff
LLR: 2.96 (-2.94,2.94) [-1.50,4.50]
Total: 84573 W: 17028 L: 16688 D: 50857
sprt @ 15+0.05 th 1 Using tuned values from spsa session
15-01-10 SC same_tempo diff
LLR: 3.27 (-2.94,2.94) [-1.50,4.50]
Total: 10497 W: 2163 L: 2016 D: 6318
sprt @ 15+0.05 th 1 Parameter from ks_tempo_tuning converged to (in average) exactly 2*Eval::tempo, suggesting that the factor 2 in search should actually be 1. Test whether doubling Eval::tempo and removing factor 2 from search.cpp does help. Mantainers should feel free to reschedule this as (-3, 1)
15-01-10 sg pruning diff
ELO: 1.17 +-2.5 (95%) LOS: 82.2%
Total: 30000 W: 6029 L: 5928 D: 18043
30000 @ 15+0.05 th 1 Measure the effect of allowing move pruning at PV nodes. Inspired by following talkchess discussion: http://www.talkchess.com/forum/viewtopic.php?t=54761&start=80
15-01-10 n_p Stefan_TunePawn diff
LLR: -3.30 (-2.94,2.94) [0.00,4.00]
Total: 60764 W: 12215 L: 12170 D: 36379
sprt @ 15+0.05 th 1 Using Stefan's SPSA-pawn tuning but preserving all symmetries in Double, Backward and Isolated by taking averages of some of the parameters.
15-01-10 sg pruning diff
ELO: 1.71 +-2.6 (95%) LOS: 90.2%
Total: 23404 W: 4008 L: 3893 D: 15503
30000 @ 60+0.05 th 1 LTC: Measure the effect of allowing move pruning at PV nodes. Little gain for STC. Because i'am interrested how this scales i prefer a fixed games test instead of a no-regression-sprt.
15-01-10 SC same_tempo diff
LLR: -2.96 (-2.94,2.94) [0.00,6.00]
Total: 6909 W: 1066 L: 1136 D: 4707
sprt @ 60+0.05 th 1 Parameter from ks_tempo_tuning converged to (in average) exactly 2*Eval::tempo, suggesting that the factor 2 in search should actually be 1. Test whether doubling Eval::tempo and removing factor 2 from search.cpp does help. Mantainers should feel free to reschedule this as (-3, 1)
15-01-11 lbr pruneqspv diff
LLR: 4.20 (-2.94,2.94) [-3.00,1.00]
Total: 85573 W: 17195 L: 17125 D: 51253
sprt @ 15+0.05 th 1 prune pv nodes in qsearch (stefan's test does it in search only and not in step 7)
15-01-11 tko SPSA_Tuning2 diff
67053/50000 iterations
113112/100000 games played
100000 @ 15+0.05 th 1 Try #2 of LMR SPSA. Fixed wrong base issue.
15-01-11 lbr pruning diff
LLR: -2.95 (-2.94,2.94) [-0.50,4.50]
Total: 6910 W: 1345 L: 1448 D: 4117
sprt @ 15+0.05 th 1 don't mark bad checks as dangerous, so they can be pruned
15-01-11 roh QuickMove diff
LLR: -3.20 (-2.94,2.94) [-1.50,4.50]
Total: 13999 W: 2797 L: 2874 D: 8328
sprt @ 15+0.05 th 1 With locally tuned parameters.
15-01-11 Roc AttackersCombo diff
LLR: -2.98 (-2.94,2.94) [-1.50,4.50]
Total: 4204 W: 786 L: 882 D: 2536
sprt @ 15+0.05 th 1 Testing the 2nd SPSA 80M session values "as is". See more notes in git.
15-01-11 sg pruning diff
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 21553 W: 4342 L: 4221 D: 12990
sprt @ 15+0.05 th 1 No regression test: allow move pruning on pv nodes. Futility pruning in step 7 is added (pointed out by Lucas)
15-01-11 n_p SPSA_KingSafety diff
37641/50000 iterations
94982/100000 games played
100000 @ 15+0.05 th 1 Another SPSA tuning attempt at king safety. Very similar to Stefan's last session but STC.
15-01-11 sg pruning diff
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 7675 W: 1351 L: 1209 D: 5115
sprt @ 60+0.05 th 1 LTC: No regression test: allow move pruning on pv nodes. Futility pruning in step 7 is added (pointed out by Lucas)