Stockfish Testing Queue

Finished - 33316 tests

15-12-20 lbr aspiration diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 17524 W: 3278 L: 3327 D: 10919
sprt @ 10+0.1 th 1 widen faster
15-12-19 pb0 distinct_iter_11 diff
LLR: 2.95 (-2.94,2.94) [0.00,5.00]
Total: 7068 W: 1121 L: 975 D: 4972
sprt @ 5+0.1 th 21 rotating symmetric halfdensity patterns with gradually increasing skipsizes till size 4 (now commit correctly pushed)
15-12-19 Roc mobtune2 diff
44449/50000 iterations
97431/100000 games played
100000 @ 20+0.1 th 1 Fixed overflow problem. KISS.
15-12-20 np tuned diff
ELO: 1.15 +-4.9 (95%) LOS: 67.5%
Total: 10000 W: 2648 L: 2615 D: 4737
10000 @ 2+0.02 th 1 Test some tuned values.
15-12-19 pec tm_cleaner diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 12523 W: 1638 L: 1715 D: 9170
sprt @ 60+0.6 th 1 Let see if scales at LTC, as result with almost random coefficients was unexpected at STC. More general form.
15-12-19 pec tm_cleaner diff
ELO: 2.88 +-2.9 (95%) LOS: 97.3%
Total: 20000 W: 3792 L: 3626 D: 12582
20000 @ 10+0.1 th 1 More general form, measure elo
15-12-19 Voy lmrs diff
LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 8459 W: 1619 L: 1477 D: 5363
sprt @ 10+0.1 th 1 LMR simplification...
15-12-19 Voy lmrt3b diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 47386 W: 8756 L: 8677 D: 29953
sprt @ 10+0.1 th 1 Do the same with the secondary killer as well.
15-12-19 pb0 distinct_iter_12 diff
Pending...
sprt @ 5+0.1 th 10 Testing a high-density-skip map approach
15-12-19 jos tuningXP diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 13354 W: 2416 L: 2527 D: 8411
sprt @ 10+0.1 th 1 Try some fine-tuned values by playing 8 games instead of 2 per iteration.
15-12-18 Voy lmrt3 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 55716 W: 10416 L: 10299 D: 35001
sprt @ 10+0.1 th 1 take 2
15-12-18 pb0 distinct_iter_11 diff
LLR: 0.13 (-2.94,2.94) [0.00,5.00]
Total: 1034 W: 152 L: 143 D: 739
sprt @ 5+0.1 th 21 STC rotating symmetric halfdensity patterns with gradually increasing skipsizes till size 4, now based on master.
15-12-19 Roc mobtune2 diff
4446/50000 iterations
9000/100000 games played
100000 @ 20+0.1 th 1 Recalibrate the global mobility weight per piece and tune also small delta on mg, eg per piece
15-12-18 Voy lmrt3 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 4409 W: 781 L: 888 D: 2740
sprt @ 10+0.1 th 1 Don't LMR if move is main killer from prior ply.
15-12-18 jki master diff
LLR: 0.58 (-2.94,2.94) [0.00,5.00]
Total: 88 W: 26 L: 3 D: 59
sprt @ 60+0.05 th 1 Regression test, 1 thread (corrected book!)
15-12-18 jki master diff
ELO: 80.37 +-12.0 (95%) LOS: 100.0%
Total: 1320 W: 428 L: 128 D: 764
40000 @ 60+0.05 th 1 Regression test, 1 thread
15-12-18 pb0 distinct_iter_paths10 diff
ELO: 5.30 +-4.1 (95%) LOS: 99.4%
Total: 8788 W: 1482 L: 1348 D: 5958
10000 @ 5+0.1 th 13 Quick test if we benefit from homogeneous skip-length patterns.
15-12-18 pb0 distinct_iter_paths10 diff
LLR: -0.08 (-2.94,2.94) [0.00,5.00]
Total: 72 W: 8 L: 11 D: 53
sprt @ 5+0.1 th 21 STC rotating symmetric halfdensity patterns with gradually increasing skipsizes till size 4.
15-12-16 Fis modSearchSMP diff
LLR: -1.37 (-2.94,2.94) [0.00,5.00]
Total: 40326 W: 6700 L: 6597 D: 27029
sprt @ 5+0.1 th 7 See github comments.
15-12-18 Voy lmrt2 diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 24650 W: 4599 L: 4617 D: 15434
sprt @ 10+0.1 th 1 Ver 2.
15-12-18 Voy lmrt diff
LLR: -2.09 (-2.94,2.94) [0.00,5.00]
Total: 5767 W: 1059 L: 1123 D: 3585
sprt @ 10+0.1 th 1 LMR tweak...
15-12-18 Voy mhs2a diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 22924 W: 4219 L: 4246 D: 14459
sprt @ 10+0.1 th 1 Version 2a: Increase history bonus a bit.
15-12-17 Fis modSearchSMP diff
LLR: 2.95 (-2.94,2.94) [0.00,5.00]
Total: 34225 W: 5763 L: 5505 D: 22957
sprt @ 5+0.1 th 7 See github comments. Bug fixed version.
15-12-18 Voy mhs4 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 17544 W: 3130 L: 3181 D: 11233
sprt @ 10+0.1 th 1 Take 4
15-12-18 Voy mhs3 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 6696 W: 1154 L: 1251 D: 4291
sprt @ 10+0.1 th 1 take 3.
15-12-18 n_p ComboPatch2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 11006 W: 2000 L: 2078 D: 6928
sprt @ 10+0.1 th 1 ComboPatch of STC test which failed yellow. Subset of the last patch.
15-12-18 Voy mhs2 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 63908 W: 12015 L: 11862 D: 40031
sprt @ 10+0.1 th 1 Material History Split ver. 2
15-12-18 pec tm_cleaner diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 17383 W: 3242 L: 3292 D: 10849
sprt @ 10+0.1 th 1 Use previous score also.
15-12-17 Voy mhp diff
ELO: 0.77 +-4.4 (95%) LOS: 63.4%
Total: 6316 W: 849 L: 835 D: 4632
8000 @ 60+0.6 th 1 Quick measurement to see how it scales, to aid in my research. Set at 8k games, will stop early if necessary.
15-12-17 wf imb-tune diff
ELO: -6.39 +-4.9 (95%) LOS: 0.5%
Total: 10000 W: 2461 L: 2645 D: 4894
10000 @ 2+0.02 th 1 smoke test some values tuned at an even faster TC
15-12-16 Roc TNT2 diff
LLR: 0.02 (-2.94,2.94) [0.00,4.00]
Total: 44533 W: 6265 L: 6141 D: 32127
sprt @ 60+0.6 th 1 Testing ttune values at LTC to see how it scales. Low throughput 250
15-12-17 pec tm_cleaner diff
LLR: -2.96 (-2.94,2.94) [-1.00,3.00]
Total: 90576 W: 16807 L: 16807 D: 56962
sprt @ 10+0.1 th 1 Remove some coefficients. Not sure how to test as simplification or parameter change, so I will test with internediate sprt[-1,3]
15-12-17 sni asymmetry_weight diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 46462 W: 8853 L: 8774 D: 28835
sprt @ 10+0.1 th 1 Take 2 (weight = 13)
15-12-17 sni asymmetry_weight diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 22392 W: 4271 L: 4298 D: 13823
sprt @ 10+0.1 th 1 take 3 (weight = 16)
15-12-16 Voy mhp diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 16230 W: 3008 L: 3063 D: 10159
sprt @ 10+0.1 th 1 Split up history table based on remaining material.
15-12-16 sni asymmetry_weight diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 29124 W: 5435 L: 5434 D: 18255
sprt @ 10+0.1 th 1 Raise asymmetry weight
15-12-16 IIv rft diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 24156 W: 4373 L: 4447 D: 15336
sprt @ 10+0.1 th 1 One and only one try on this idea based on recent tuning.
15-12-16 sto pawn_eval_weights diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 19483 W: 3686 L: 3775 D: 12022
sprt @ 10+0.1 th 1 Higher penalties for both middlegame in general and also for the g- and h- files
15-12-13 jki distinct_iter_paths7 diff
ELO: 8.86 +-3.7 (95%) LOS: 100.0%
Total: 8000 W: 1076 L: 872 D: 6052
8000 @ 30+0.3 th 20 test with many threads: 20 threads, 30+0.3, 8000 games
15-12-16 Voy sbt diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 41143 W: 7658 L: 7672 D: 25813
sprt @ 10+0.1 th 1 Stats bonus tweak. (fixed bench)
15-12-16 sto doubled_pawn_penalty diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 28175 W: 5265 L: 5268 D: 17642
sprt @ 10+0.1 th 1 A central doubled pawn is a *serious* disadvantage. The reason I changed the h-file value for a doubled pawn is this: kings are more likely to castle kingside than queenside which means that if we are trying to attack the opponent's king, we will be thwarted by our own doubled h-pawns.
15-12-16 roh Multi-Thread-Search diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 14045 W: 1927 L: 1998 D: 10120
sprt @ 30+0.3 th 3 Try alternate method of helper thread depth that is independent of this->idx. Now, trying at LTC. STC Results [sprt @ 5+0.1 th 3 Total: 11147 W: 2071 L: 1899 D: 7177]
15-12-15 n_p ComboPatch diff
LLR: -2.94 (-2.94,2.94) [0.00,5.00]
Total: 91758 W: 17193 L: 16919 D: 57646
sprt @ 10+0.1 th 1 Combo patch of patches [0,4] which has failed yellow at STC recently. Tested with [0,5].
15-12-15 Roc TNT2 diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 79284 W: 14888 L: 14768 D: 49628
sprt @ 10+0.1 th 1 Testing ttune values
15-12-16 adp tunepassedpawn diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 23561 W: 4404 L: 4479 D: 14678
sprt @ 10+0.1 th 1 Tuned Passed Pawn Values after SPSA
15-12-15 Voy spt diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 26322 W: 3586 L: 3663 D: 19073
sprt @ 60+0.6 th 1 LTC: Shallow Pruning tweak
15-12-15 adp passedpawn diff
29008/30000 iterations
59783/60000 games played
60000 @ 20+0.1 th 1 Tuning of passed pawn values.
15-12-15 roh Multi-Thread-Search diff
LLR: 2.96 (-2.94,2.94) [0.00,5.00]
Total: 11147 W: 2071 L: 1899 D: 7177
sprt @ 5+0.1 th 3 Try alternate method of helper thread depth that is independent of this->idx
15-12-15 Kal new_space diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 21725 W: 4005 L: 4037 D: 13683
sprt @ 10+0.1 th 1 Moved pawn starting square popcount to Pawn::Entry::probe() to minimize popcounts; lowered critical number of pawns on starting squares from 6 to 5
15-12-15 Voy spt diff
LLR: 2.95 (-2.94,2.94) [0.00,4.00]
Total: 15202 W: 2924 L: 2714 D: 9564
sprt @ 10+0.1 th 1 Shallow Pruning tweak