Stockfish Testing Queue

Finished - 24198 tests

19-07-15 Vo cmhStats diff
LLR: -2.94 (-2.94,2.94) [-1.50,4.50]
Total: 1397 W: 200 L: 300 D: 897
sprt @ 15+0.05 th 1 Try a replacement strategy
19-07-15 My RB diff
LLR: -2.95 (-2.94,2.94) [-3.00,1.00]
Total: 7292 W: 1327 L: 1499 D: 4466
sprt @ 15+0.05 th 1 See if we can remove space squares behind pawns
19-07-15 lb decay diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 7559 W: 1378 L: 1471 D: 4710
sprt @ 15+0.05 th 1 Glaurung style decay: take 2
19-07-15 Vo StatsTweak diff
LLR: 0.52 (-2.94,2.94) [-1.50,4.50]
Total: 9488 W: 1829 L: 1785 D: 5874
sprt @ 15+0.05 th 1 The idea is this: If the stats table is 0, we won't be able to update the stats if depth>15, since 16*16>250 (max). This tweak will make it so we can update a value if depth>15.
19-07-15 lb decay diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 5693 W: 1063 L: 1165 D: 3465
sprt @ 15+0.05 th 1 take 3
19-07-15 lb decay^ diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 2417 W: 397 L: 513 D: 1507
sprt @ 15+0.05 th 1 Glaurung style decay (never tried with CMH)
18-07-15 SC QKfork2 diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 22025 W: 4298 L: 4345 D: 13382
sprt @ 15+0.05 th 1 If QK are diagonally adjacent to each other, the risk of been forked or X-rayed is higher. Give a little higher penalty, but only if we have a knight or a bishop. Take 2.
19-07-15 lb master diff
ELO: 58.63 +-24.1 (95%) LOS: 100.0%
Total: 335 W: 99 L: 43 D: 193
40000 @ 60+0.05 th 1 Regression test until 4095ff0e
18-07-15 Vo StatsTweak2 diff
LLR: -2.94 (-2.94,2.94) [-1.50,4.50]
Total: 5988 W: 1062 L: 1151 D: 3775
sprt @ 15+0.05 th 1 Stats Tweak v.2
17-07-15 tv SPHistory^ diff
ELO: -1.26 +-3.1 (95%) LOS: 21.3%
Total: 14327 W: 2099 L: 2151 D: 10077
20000 @ 30+0.05 th 7 Final test, Depth == 6 / thr 7 (low prio)
18-07-15 Vo StatsTweak diff
LLR: 0.11 (-2.94,2.94) [-1.50,4.50]
Total: 1220 W: 230 L: 223 D: 767
sprt @ 15+0.05 th 1 The idea is this: If the stats table is 0, we won't be able to update the stats if depth>15, since 16*16>250 (max). This tweak will make it so we can update a value if depth>15.
19-07-15 lb decay diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 18445 W: 2825 L: 2877 D: 12743
sprt @ 60+0.05 th 1 LTC: decay really gently.
18-07-15 lb decay diff
LLR: 2.95 (-2.94,2.94) [0.00,5.00]
Total: 73626 W: 13980 L: 13532 D: 46114
sprt @ 15+0.05 th 1 decay really gently. Reschedule, because test was stopped by (ab)user icewulf
18-07-15 SC QKfork diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 13406 W: 2466 L: 2536 D: 8404
sprt @ 15+0.05 th 1 If QK are diagonally adjacent to each other, the risk of been forked or X-rayed is higher. Give a little penalty. Rescheduled: wrong bench from icewulf.
18-07-15 mc improving diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 11715 W: 2151 L: 2226 D: 7338
sprt @ 15+0.05 th 1 Stricter 'improving' condition
18-07-15 mc se_lmr diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 7647 W: 1425 L: 1511 D: 4711
sprt @ 15+0.05 th 1 Increase LMR when singular extended
18-07-15 mc sp_history diff
ELO: -69.34 +-13.8 (95%) LOS: 0.0%
Total: 792 W: 57 L: 213 D: 522
20000 @ 30+0.05 th 3 Test per-splitpoint CounterMovesHistory instead of History
18-07-15 mb threats1 diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 11451 W: 2108 L: 2183 D: 7160
sprt @ 15+0.05 th 1 Piece shuffling is caused by overevaluation of something. Try a more accurate threat evaluation for major pieces, and also reduce the threat evaluation overall. Rescheduled.
18-07-15 tv HistHash diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 1771 W: 297 L: 399 D: 1075
sprt @ 15+0.05 th 1 Try Hashing History data
18-07-15 SC QKfork diff
LLR: -1.11 (-2.94,2.94) [-1.50,4.50]
Total: 2511 W: 463 L: 496 D: 1552
sprt @ 15+0.05 th 1 If QK are diagonally adjacent to each other, the risk of been forked or X-rayed is higher. Give a little penalty. Rescheduled: wrong bench from icewulf. Please stop it.
18-07-15 lb decay diff
LLR: -2.48 (-2.94,2.94) [0.00,5.00]
Total: 15746 W: 2925 L: 2962 D: 9859
sprt @ 15+0.05 th 1 decay really gently
18-07-15 Fi TTreplace diff
LLR: 2.96 (-2.94,2.94) [-4.00,0.00]
Total: 22549 W: 4257 L: 4178 D: 14114
sprt @ 15+0.05 th 1 No regression w/ low hash pressure. 16MB STC
17-07-15 Vo CM-Tweak2 diff
LLR: -3.83 (-2.94,2.94) [-1.50,4.50]
Total: 19247 W: 3573 L: 3659 D: 12015
sprt @ 15+0.05 th 1 CM-Tweak2: Fixed logic.
17-07-15 Vo RookMob diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 9142 W: 1661 L: 1787 D: 5694
sprt @ 15+0.05 th 1 Rook mobility tweak.
16-07-15 tv SPHistory diff
ELO: -0.03 +-3.7 (95%) LOS: 49.3%
Total: 10000 W: 1486 L: 1487 D: 7027
10000 @ 30+0.05 th 7 Do we gain with more threads, depth == 8
17-07-15 mb threats1 diff
LLR: -0.04 (-2.94,2.94) [-1.50,4.50]
Total: 190 W: 38 L: 39 D: 113
sprt @ 15+0.05 th 1 Piece shuffling is caused by overevaluation of something. Try a more accurate threat evaluation for major pieces, and also reduce the threat evaluation overall.
17-07-15 SC QKfork diff
LLR: -0.14 (-2.94,2.94) [-1.50,4.50]
Total: 105 W: 21 L: 26 D: 58
sprt @ 15+0.05 th 1 If QK are diagonally adjacent to each other, the risk of been forked or X-rayed is higher. Give a little penalty.
16-07-15 Fi TTreplace diff
LLR: 2.95 (-2.94,2.94) [0.00,4.00]
Total: 134618 W: 21276 L: 20716 D: 92626
sprt @ 60+0.05 th 1 (Rescheduling due to a worker stop by fatmurphy) I just realized this is probably just a tuning patch. If you disagree let me know and I'll reschedule w/ [0,6] instead. I will also run a no regression test w/ low hash pressure if we pass LTC 8MB.
17-07-15 Vo Boxfish-2 diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 10388 W: 1933 L: 2011 D: 6444
sprt @ 15+0.05 th 1 CMH Weigh
16-07-15 Vo CM-Tweak diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 8206 W: 1508 L: 1592 D: 5106
sprt @ 15+0.05 th 1 Clear CM entry if killer move.
14-07-15 tv SPHistory^ diff
ELO: 1.36 +-2.7 (95%) LOS: 83.9%
Total: 20000 W: 3139 L: 3061 D: 13800
20000 @ 30+0.05 th 3 Depth == 6 (low prio)
14-07-15 tv SPHistory^^ diff
ELO: 0.78 +-2.7 (95%) LOS: 71.8%
Total: 20000 W: 3080 L: 3035 D: 13885
20000 @ 30+0.05 th 3 Depth == 7 (low prio)
16-07-15 Vo RetireCM diff
LLR: -2.95 (-2.94,2.94) [-3.00,1.00]
Total: 20899 W: 3908 L: 4104 D: 12887
sprt @ 15+0.05 th 1 See if we can now retire countermoves. Since it is no longer being used in LMR.
15-07-15 Fi tarrasch diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 17926 W: 3333 L: 3391 D: 11202
sprt @ 15+0.05 th 1 Take 2: Values suggested by ChessQuake based on his local testing.
15-07-15 Fi TTreplace diff
LLR: 0.04 (-2.94,2.94) [0.00,4.00]
Total: 73 W: 9 L: 7 D: 57
sprt @ 60+0.05 th 1 I just realized this is probably just a tuning patch. If you disagree let me know and I'll reschedule w/ [0,6] instead. I will also run a no regression test w/ low hash pressure if we pass LTC 8MB.
15-07-15 Fi TTreplace diff
LLR: 2.96 (-2.94,2.94) [-1.50,4.50]
Total: 16353 W: 3216 L: 3066 D: 10071
sprt @ 15+0.05 th 1 A different way of combining generation and depth to calculate a TT entries replace value. 2MB
14-07-15 tv SPHistory diff
ELO: -0.07 +-3.8 (95%) LOS: 48.6%
Total: 10000 W: 1550 L: 1552 D: 6898
10000 @ 30+0.05 th 3 Marco's version of SplitPoint history with depth == 8
15-07-15 Vo SEE-RF2-Clean diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 11029 W: 2004 L: 2080 D: 6945
sprt @ 15+0.05 th 1 SEE-Risk Factor
15-07-15 jo dynamic_futility diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 5191 W: 964 L: 1056 D: 3171
sprt @ 15+0.05 th 1 Less futility-pruning during the first iterations.
14-07-15 tv SPHistory diff
ELO: -3.97 +-4.6 (95%) LOS: 4.5%
Total: 7000 W: 1079 L: 1159 D: 4762
10000 @ 30+0.05 th 3 Try depth == Threads.minimumSplitDepth
14-07-15 lb latejoin_tweak diff
ELO: 1.34 +-2.7 (95%) LOS: 83.7%
Total: 20000 W: 3118 L: 3041 D: 13841
20000 @ 30+0.05 th 3 Joerg's patch with 3 threads. All machines with < 7 threads are idle at the moment.
15-07-15 Vo Enpassant-2 diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 5760 W: 1046 L: 1136 D: 3578
sprt @ 15+0.05 th 1 Add enpassant in move picker. This time don't apply relative rank penalty.
14-07-15 Vo LMR-S2 diff
LLR: 3.01 (-2.94,2.94) [-3.00,1.00]
Total: 32410 W: 5092 L: 4986 D: 22332
sprt @ 60+0.05 th 1 LTC: LMR Simplification
14-07-15 mb ks_tweak diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 7499 W: 1377 L: 1463 D: 4659
sprt @ 15+0.05 th 1 Give fewer attackUnits from safe checks if only one piece is attacking.
13-07-15 jo latejoin_tweak diff
ELO: 3.40 +-3.6 (95%) LOS: 96.9%
Total: 10000 W: 1430 L: 1332 D: 7238
10000 @ 30+0.05 th 7 Resubmit with 7 cores @30+0.05. Compute a score for available split points when trying to late join. Parameters are: number of parent splits (most important), remaining search depth (very important) and moveCount (least important).
14-07-15 Fi consistentTT diff
LLR: 2.96 (-2.94,2.94) [-4.00,0.00]
Total: 17401 W: 3324 L: 3227 D: 10850
sprt @ 15+0.05 th 1 STC 16MB [-4,0] as requested by lucasart
13-07-15 Vo LMR-S2 diff
LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 24450 W: 4632 L: 4517 D: 15301
sprt @ 15+0.05 th 1 LMR Simplification
13-07-15 lb combo diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 4509 W: 803 L: 910 D: 2796
sprt @ 15+0.05 th 1 combo a couple of small tuning patches that failed yellow at SPRT(0,4). don't want to spend too much resources for that, so use SPRT(0,5).
13-07-15 tv CounterMoves diff
ELO: -0.34 +-2.8 (95%) LOS: 40.6%
Total: 18543 W: 2854 L: 2872 D: 12817
20000 @ 30+0.05 th 3 Fixed depth tweak
13-07-15 Fi tarrasch diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 3022 W: 507 L: 604 D: 1911
sprt @ 15+0.05 th 1 Rook behind passed pawn bonus. Note: We already do something in evaluate_passed_pawns() but let's see how this performs anyway.