Stockfish Testing Queue

Finished - 44210 tests

15-05-24 Mys K7 diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 21818 W: 4203 L: 4250 D: 13365
sprt @ 15+0.05 th 1 Heavier initial penalty for weak king pawn shelter and only small increments.
15-05-24 lbr after_null diff
LLR: -2.95 (-2.94,2.94) [-1.00,4.00]
Total: 7933 W: 1409 L: 1514 D: 5010
sprt @ 15+0.05 th 1 half history increment for null refutation
15-05-23 SC scale_regression_3 diff
ELO: -15.92 +-3.0 (95%) LOS: 0.0%
Total: 20000 W: 3558 L: 4474 D: 11968
20000 @ 10+0.05 th 1 scale_regression project, take 3: a linear model from a better set of features. StdErr down to 9.5.
15-05-23 Voy ShortfinMako diff
LLR: -3.37 (-2.94,2.94) [-1.50,4.50]
Total: 15167 W: 2814 L: 2894 D: 9459
sprt @ 15+0.05 th 1 Split History: Split at Depth 18..see if this makes a big difference like it did for AntarticCod.
15-05-23 Roc Passivity diff
LLR: -2.94 (-2.94,2.94) [-1.50,4.50]
Total: 9899 W: 1860 L: 1939 D: 6100
sprt @ 15+0.05 th 1 Penalty when not pressuring anything.
15-05-23 pec wider diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 8697 W: 1648 L: 1731 D: 5318
sprt @ 15+0.05 th 1 Widen closer to root, trim further away for non-improving nodes.
15-05-23 Roc OneWayRook diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 7496 W: 1386 L: 1472 D: 4638
sprt @ 15+0.05 th 1 Penalize Rooks which have one or zero mobility in one direction. Inspired by Mindbreaker idea.
15-05-23 jos unstoppable diff
LLR: -2.96 (-2.94,2.94) [0.00,6.00]
Total: 18671 W: 2248 L: 2277 D: 14146
sprt @ 60+0.05 th 1 LTC: Allow a knight for both sides. Final try.
15-05-23 Roc ConnectedV3b diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 23876 W: 4559 L: 4601 D: 14716
sprt @ 15+0.05 th 1 Fixed previous ConnectedV3b test (loop must go from 0 to 2)
15-05-23 Roc CanCastleNow diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 15830 W: 3038 L: 3102 D: 9690
sprt @ 15+0.05 th 1 Take 2: Lower bonus
15-05-23 Roc SkipThreats diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 2549 W: 433 L: 532 D: 1584
sprt @ 15+0.05 th 1 Fixed
15-05-22 jos unstoppable diff
LLR: 2.96 (-2.94,2.94) [-1.50,4.50]
Total: 64580 W: 9957 L: 9712 D: 44911
sprt @ 15+0.05 th 1 Allow a knight for both sides. Final try.
15-05-23 lbr doubled diff
LLR: -2.93 (-2.94,2.94) [-3.00,1.00]
Total: 14704 W: 2831 L: 3017 D: 8856
sprt @ 15+0.05 th 1 remove doubled pawns (test based on new master)
15-05-22 Fis smartTT2 diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 54458 W: 10198 L: 10160 D: 34100
sprt @ 15+0.05 th 1 A hopefully stronger version of smartTT. 2MB
15-05-23 Roc SkipThreats diff
LLR: -0.57 (-2.94,2.94) [-1.50,4.50]
Total: 214 W: 40 L: 61 D: 113
sprt @ 15+0.05 th 1 Skip main threat evaluation loops if less than 2 threats.
15-05-23 Mys K5.1 diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 23336 W: 4404 L: 4448 D: 14484
sprt @ 15+0.05 th 1 K5.1 - take 2 quick try to see with values aligned to midgame.
15-05-23 Fis easyMove2 diff
LLR: -3.18 (-2.94,2.94) [-1.50,4.50]
Total: 16141 W: 3001 L: 3072 D: 10068
sprt @ 15+0.05 th 1 If we don't have a valid PV decrease the stable counter but don't set it directly to 0.
15-05-23 Fis capTime diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 6998 W: 1298 L: 1385 D: 4315
sprt @ 15+0.05 th 1 Cap max time to 2x available.
15-05-23 Voy AntarticCod-3 diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 5554 W: 1012 L: 1103 D: 3439
sprt @ 15+0.05 th 1 Try this condition: if (ply > 18 && depth<=3)
15-05-23 lbr backward diff
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 14143 W: 2334 L: 2203 D: 9606
sprt @ 60+0.05 th 1 LTC: simplify backward pawns
15-05-22 Voy AntarticCod-2 diff
LLR: 2.96 (-2.94,2.94) [-1.50,4.50]
Total: 24732 W: 4824 L: 4652 D: 15256
sprt @ 15+0.05 th 1 Count based method to clear history stats.
15-05-22 pec wider diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 22194 W: 4178 L: 4225 D: 13791
sprt @ 15+0.05 th 1 widen nonimproving near root by redefing improving
15-05-22 Voy AntarticCod diff
LLR: 2.95 (-2.94,2.94) [-1.50,4.50]
Total: 11368 W: 2192 L: 2057 D: 7119
sprt @ 15+0.05 th 1 I believe I set the boundary too low. Lets try ply>18..
15-05-22 jos verification diff
LLR: -2.97 (-2.94,2.94) [0.00,6.00]
Total: 12926 W: 2001 L: 2044 D: 8881
sprt @ 60+0.05 th 1 LTC: No null-move during verification search. Since it passed at STC I would really like to see how it's doing at LTC. Lower throughput (300)
15-05-22 lbr backward diff
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 52322 W: 10011 L: 9945 D: 32366
sprt @ 15+0.05 th 1 simplify backward pawns
15-05-22 Voy AntarticCod diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 5280 W: 974 L: 1066 D: 3240
sprt @ 15+0.05 th 1 Clear History Stats near the leaves (ply>10) on new search.
15-05-22 lbr doubled diff
LLR: -2.96 (-2.94,2.94) [-3.00,1.00]
Total: 30046 W: 4609 L: 4807 D: 20630
sprt @ 60+0.05 th 1 LTC: simplify doubled pawns
15-05-22 Voy HistorySync diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 1488 W: 245 L: 348 D: 895
sprt @ 15+0.05 th 1 Sync cmh with history...
15-05-22 Voy Albacore diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 23647 W: 4421 L: 4464 D: 14762
sprt @ 15+0.05 th 1 Try my luck of using cmh to help sort evasion moves...
15-05-19 jki ee0371f86e319aa24bc1 diff
LLR: 2.96 (-2.94,2.94) [-4.00,0.00]
Total: 51564 W: 8106 L: 8111 D: 35347
sprt @ 60+0.05 th 1 Cleanup work in misc.cpp (Regression test), Hash = 8MB
15-05-21 Voy ShortfinMako3 diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 2743 W: 448 L: 546 D: 1749
sprt @ 15+0.05 th 1 History Split: Prevent discontinuity by overlapping the two tables by two plies.
15-05-21 SC scale_regression_2 diff
ELO: -17.11 +-2.9 (95%) LOS: 0.0%
Total: 20000 W: 3195 L: 4179 D: 12626
20000 @ 15+0.05 th 1 Analogously, check if new weigths after fix in regression code give some improvements.
15-05-21 jos movecount_pruning diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 34840 W: 6578 L: 6591 D: 21671
sprt @ 15+0.05 th 1 Less movecount-pruning close to the root. Take 2.
15-05-21 lbr doubled diff
LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 40254 W: 7675 L: 7588 D: 24991
sprt @ 15+0.05 th 1 simplify doubled pawns
15-05-21 Voy Whiptail diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 36105 W: 6779 L: 6789 D: 22537
sprt @ 15+0.05 th 1 Increase cmh weight for moves near the root.
15-05-21 jos unstoppable diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 33367 W: 5007 L: 5034 D: 23326
sprt @ 15+0.05 th 1 Calculate unstoppable also if one side has a knight. A knight often has difficulties to stop a passer. (Throughput 1100) Test with 8moves book to give more weight to the endgame.
15-05-21 SC score_evasions_8 diff
ELO: -1.09 +-3.1 (95%) LOS: 24.2%
Total: 20000 W: 4030 L: 4093 D: 11877
20000 @ 10+0.05 th 1 After a major improvement in my regression code, I think I have go much better weights than before. Short check that this is indeed the case.
15-05-21 SC score_captures_1 diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 5585 W: 1024 L: 1115 D: 3446
sprt @ 15+0.05 th 1 Last byproduct of regression code improvement.
15-05-21 sni attack_info diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 9132 W: 1695 L: 1776 D: 5661
sprt @ 15+0.05 th 1 Use attack info for capture ordering
15-05-21 Voy ShortfinMako2 diff
LLR: -3.00 (-2.94,2.94) [-1.50,4.50]
Total: 24701 W: 4600 L: 4642 D: 15459
sprt @ 15+0.05 th 1 SplitHistory; Clear history at leaves on new search.
15-05-12 SC score_evasions_7 diff
LLR: -0.85 (-2.94,2.94) [-3.00,1.00]
Total: 137944 W: 21553 L: 21807 D: 94584
sprt @ 60+0.05 th 1 Try different formulas for evasions scoring. Take 7: use regression formula based on current m.value(). LTC.
15-05-21 jos movecount_pruning diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 5345 W: 954 L: 1045 D: 3346
sprt @ 15+0.05 th 1 No movecount pruning for the first 2 plies from the root.
15-05-20 Voy ShortfinMako diff
LLR: -2.94 (-2.94,2.94) [-1.50,4.50]
Total: 12150 W: 2171 L: 2244 D: 7735
sprt @ 15+0.05 th 1 Split History: The poor results of FlyingFish was due to a bug. Hopefully this shark will prove much stronger.
15-05-21 aji threats_tune_1 diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 5110 W: 906 L: 998 D: 3206
sprt @ 15+0.05 th 1 Use the same values in threats tuning session, but give a hanging piece bonus only once per piece : STC
15-05-16 Mys K5 diff
LLR: -2.96 (-2.94,2.94) [0.00,6.00]
Total: 140431 W: 22347 L: 21806 D: 96278
sprt @ 60+0.05 th 1 Place emphasis on the quality of our king's pawn shelter in the endgame also. LTC.
15-05-18 lbr master diff
ELO: 275.66 +-3.4 (95%) LOS: 100.0%
Total: 40370 W: 28796 L: 2138 D: 9436
40000 @ 10+0.1 th 1 master vs. SF 3 (reference point)
15-05-20 aji threats_tuned diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 3943 W: 696 L: 791 D: 2456
sprt @ 15+0.05 th 1 SPSA tuned values for all threats by each attacking piece and attacked piec and differentiated as defended/weak/hanging : STC
15-05-19 SC scale_regression diff
ELO: -58.91 +-3.2 (95%) LOS: 0.0%
Total: 20000 W: 2871 L: 6230 D: 10899
20000 @ 10+0.05 th 1 What would happens if we would throw to the dogs all the endgame knowledge for scaling factors and replace by a simple regression formula?
15-05-17 lbr verif diff
ELO: 276.95 +-3.5 (95%) LOS: 100.0%
Total: 40000 W: 28599 L: 2102 D: 9299
40000 @ 10+0.1 th 1 verif vs. SF 3
15-05-20 sg space diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 10220 W: 1874 L: 1953 D: 6393
sprt @ 15+0.05 th 1 Space advantage on both wings by pawn pairs on 5th rank and distance 3 (like a5/d5). Use 50% more bonus (take 3).