Stockfish Testing Queue

Finished - 44284 tests

15-01-28 lbr skill diff
ELO: 25.72 +-6.2 (95%) LOS: 100.0%
Total: 10000 W: 4461 L: 3722 D: 1817
10000 @ 20+0.05 th 1 Double skill level resolution. Measure ELO gap: 18 vs 17
15-01-28 lbr skill diff
ELO: 21.29 +-6.1 (95%) LOS: 100.0%
Total: 10000 W: 4352 L: 3740 D: 1908
10000 @ 20+0.05 th 1 Double skill level resolution. Measure ELO gap: 19 vs 18
15-01-28 Roc SPSA_EvalSimplification diff
12016/12500 iterations
25000/25000 games played
25000 @ 15+0.05 th 1 SPSA: "simplification" idea was not harmful, but not exciting. Let's see if tweaking on some ThreatenedByPawn weights we can get something more interesting worth going further.
15-01-28 lbr skill diff
ELO: 24.95 +-6.1 (95%) LOS: 100.0%
Total: 10000 W: 4393 L: 3676 D: 1931
10000 @ 20+0.05 th 1 Double skill level resolution. 20 vs 19
15-01-28 lbr skill diff
ELO: 16.62 +-6.1 (95%) LOS: 100.0%
Total: 10000 W: 4265 L: 3787 D: 1948
10000 @ 20+0.05 th 1 Double skill level resolution. 21 vs 20
15-01-28 lbr skill diff
ELO: 18.78 +-6.1 (95%) LOS: 100.0%
Total: 10000 W: 4229 L: 3689 D: 2082
10000 @ 40+0.05 th 1 Double skill level resolution. 21 vs 20. redo this one with more tc to see if we hit the tc ceiling for that skill level
15-01-28 roh QuickMove diff
LLR: -2.96 (-2.94,2.94) [0.00,6.00]
Total: 21144 W: 3614 L: 3614 D: 13916
sprt @ 60+0.05 th 1 Testing with 1000 iterations at 60+0.05 passes sanity check. Check with 20,000 matches while FishTest is idle.
15-01-28 n_p SPSAKingSafety2 diff
58111/50000 iterations
94197/100000 games played
100000 @ 60+0.05 th 1 Another SPSA-session on king safety. The start values from KingSafetySPSA.
15-01-28 Roc FCC_20150127 diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 3513 W: 685 L: 783 D: 2045
sprt @ 15+0.05 th 1 A fix on take #3 of previous Fast contact checks attempt. A bit slower than RC3.
15-01-28 Roc FCC_20150127 diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 8367 W: 1691 L: 1775 D: 4901
sprt @ 15+0.05 th 1 A small variation.
15-01-29 Roc FCC_20150127 diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 8408 W: 1637 L: 1721 D: 5050
sprt @ 15+0.05 th 1 One more tweak.
15-01-29 Mys no_trapRK2 diff
ELO: -33.59 +-3.1 (95%) LOS: 0.0%
Total: 21000 W: 3667 L: 5691 D: 11642
20000 @ 15+0.05 th 1 Indicator of removing trappedrook #2
15-01-29 Mys no_trapRK1 diff
ELO: -10.17 +-3.1 (95%) LOS: 0.0%
Total: 20000 W: 3857 L: 4442 D: 11701
20000 @ 15+0.05 th 1 Indicator of removing trappedrook #1
15-01-29 Roc FCC_20150127 diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 10634 W: 2084 L: 2162 D: 6388
sprt @ 15+0.05 th 1 This time, use exactly the same bitboard for mobility andkingring calculation as master, and hoping that remaining code is still "improvement"
15-01-29 n_p KingSafety diff
ELO: 3.21 +-2.1 (95%) LOS: 99.8%
Total: 40000 W: 8181 L: 7812 D: 24007
40000 @ 15+0.05 th 1 Checking the values from the SPSA-session on king safety.
15-01-29 lbr skill diff
ELO: -0.98 +-16.0 (95%) LOS: 45.2%
Total: 1777 W: 875 L: 880 D: 22
10000 @ 10+0.05 th 1 measure simplified skill levels: 2 vs 0
15-01-29 lbr skill diff
ELO: 101.89 +-25.6 (95%) LOS: 100.0%
Total: 740 W: 462 L: 251 D: 27
10000 @ 10+0.05 th 1 measure simplified skill levels: 4 vs 2
15-01-30 gli less_mc_pruning diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 1339 W: 222 L: 326 D: 791
sprt @ 15+0.05 th 1 Do less movecount pruning if there are big threats on the board
15-01-30 gli less_mc_pruning diff
ELO: -19.74 +-7.0 (95%) LOS: 0.0%
Total: 3859 W: 674 L: 893 D: 2292
20000 @ 15+0.05 th 1 (Pre-emptive ELO loss measurement :) Do less movecount pruning if there are big threats on the board
15-01-30 gli measure_pv_pruning diff
ELO: -3.89 +-2.0 (95%) LOS: 0.0%
Total: 40000 W: 6535 L: 6983 D: 26482
40000 @ 60+0.05 th 1 Measure ELO of futility pruning in PV nodes
15-01-30 n_p KingSafety50 diff
ELO: 2.81 +-2.2 (95%) LOS: 99.4%
Total: 40000 W: 8283 L: 7960 D: 23757
40000 @ 15+0.05 th 1 The SPSA-values with 50% more change.
15-01-30 gli less_mc_pruning2 diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 1958 W: 314 L: 414 D: 1230
sprt @ 15+0.05 th 1 Don't movecount prune if bestValue is a lot lower than static eval. This implies we don't understand this position very well, so try more moves.
15-01-30 sg pawn_attack_threat diff
LLR: 2.95 (-2.94,2.94) [-1.50,4.50]
Total: 7925 W: 1666 L: 1537 D: 4722
sprt @ 15+0.05 th 1 Add bonus for possible safe pawn pushes which attack an enemy piece. Inspired by http://talkchess.com/forum/viewtopic.php?t=55142
15-01-30 sg pawn_attack_threat diff
LLR: 2.95 (-2.94,2.94) [0.00,6.00]
Total: 40109 W: 6841 L: 6546 D: 26722
sprt @ 60+0.05 th 1 LTC: Add bonus for possible safe pawn pushes which attack an enemy piece. Inspired by http://talkchess.com/forum/viewtopic.php?t=55142
15-01-31 Roc QueenPin diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 10171 W: 1994 L: 2073 D: 6104
sprt @ 15+0.05 th 1 Reducing mobility of pinned pieces against Queen. Simplified version, take 1
15-01-31 Roc QueenPin diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 10919 W: 2212 L: 2289 D: 6418
sprt @ 15+0.05 th 1 Take 2: If a piece is pinned against Queen, adjust King Safety too.
15-01-31 sg pawn_attack_threat2 diff
LLR: 2.95 (-2.94,2.94) [-1.50,4.50]
Total: 14079 W: 2910 L: 2764 D: 8405
sprt @ 15+0.05 th 1 Add bonus for possible safe pawn pushes which attack an enemy piece. Cover more cases by using a doubleAttackedBy array. (Take 2)
15-01-31 SC no_tempo diff
ELO: -8.65 +-2.1 (95%) LOS: 0.0%
Total: 42000 W: 8017 L: 9062 D: 24921
40000 @ 15+0.05 th 1 Measure how much elo is tempo evaluation worth.
15-01-31 vin passed_blockers diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 15827 W: 3178 L: 3241 D: 9408
sprt @ 15+0.05 th 1 Try a small bonus for a protected middle game passed pawn, since it is more likely to survive long-term into the ending. Local 10,000 game test was promising.
15-01-31 sg pawn_attack_threat2 diff
LLR: 2.95 (-2.94,2.94) [0.00,6.00]
Total: 6092 W: 1064 L: 933 D: 4095
sprt @ 60+0.05 th 1 LTC: Add bonus for possible safe pawn pushes which attack an enemy piece. Cover more cases by using a doubleAttackedBy array. (Take 2)
15-01-31 n_p KingSafety2 diff
ELO: 1.44 +-2.3 (95%) LOS: 88.8%
Total: 34359 W: 6910 L: 6768 D: 20681
40000 @ 15+0.05 th 1 When we accidentally got another set of king safety values after "the great purge". Test of these.
15-01-31 sg pawn_attack_threat2 diff
ELO: -1.07 +-2.0 (95%) LOS: 15.1%
Total: 37089 W: 6061 L: 6175 D: 24853
40000 @ 60+0.05 th 1 Both version of pawn attack threat passed (the second seems better at LTC counting the test run length, but this can misleading). So measure in a direct match which is the better one.
15-01-31 Roc QueenPin diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 8660 W: 1701 L: 1784 D: 5175
sprt @ 15+0.05 th 1 See git notes.
15-01-31 n_p KingSafety diff
LLR: 2.95 (-2.94,2.94) [0.00,4.00]
Total: 47765 W: 8091 L: 7785 D: 31889
sprt @ 60+0.05 th 1 Try the best values of king safety in LTC.
15-01-31 Roc AttackBuildUp diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 5008 W: 958 L: 1051 D: 2999
sprt @ 15+0.05 th 1 Compute all safe squares to which an attacker can actually move to join the attack on the king ring, or retreat and keep the attack. Take 1.
15-01-31 Roc QueenAttacksThreat diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 779 W: 132 L: 241 D: 406
sprt @ 15+0.05 th 1 Another follow-up on Stefan tests. Covering the latent attacks on Queen by Rooks or Minors. Take 1.
15-02-01 pec mc_pv diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 7147 W: 1367 L: 1454 D: 4326
sprt @ 15+0.05 th 1 Don't allow move count pruning at minimal depth + minimal distance to pv similar to old Stephan patch. At least don't drop queen in Gary's position.
15-02-01 Roc QueenAttacksThreat diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 516 W: 69 L: 178 D: 269
sprt @ 15+0.05 th 1 Improved version.
15-02-01 vin passed_blockers2 diff
LLR: 2.96 (-2.94,2.94) [-1.50,4.50]
Total: 2884 W: 655 L: 539 D: 1690
sprt @ 15+0.05 th 1 Minor pieces make better blockaders - so try small bonus for passer blocked by heavy piece. Smoke test ok at 7+0.05.
15-02-01 n_p SPSAKingSafety diff
45754/50000 iterations
87003/100000 games played
100000 @ 60+0.05 th 1 Another SPSA-session on king safety. The start values from KingSafety.
15-02-01 mco probcut diff
LLR: -2.96 (-2.94,2.94) [-3.00,1.00]
Total: 1574 W: 238 L: 401 D: 935
sprt @ 15+0.05 th 1 Attempt a simplification in probcut
15-02-01 sg spsa_pawn_attack_threat diff
48483/50000 iterations
100000/100000 games played
100000 @ 15+0.05 th 1 Tune parameters of my passed pawn attack threat patch. Use 100000 games because the parameters are completly untuned.
15-02-01 vin passed_blockers2 diff
LLR: -2.97 (-2.94,2.94) [0.00,6.00]
Total: 13043 W: 2204 L: 2244 D: 8595
sprt @ 60+0.05 th 1 Minor pieces make better blockaders - so try small bonus for passer blocked by heavy piece. LTC test after STC passed (alarmingly quickly!)
15-02-01 Roc LatentQAttack diff
LLR: -2.94 (-2.94,2.94) [-1.50,4.50]
Total: 6734 W: 1307 L: 1395 D: 4032
sprt @ 15+0.05 th 1 Fixed error. This test with rook and Minors latent threats on non-attacked enemy Queen, score at most twice.
15-02-01 Roc LatentQAttack diff
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 4722 W: 891 L: 985 D: 2846
sprt @ 15+0.05 th 1 Same as previous, but don't compute Rook latent threats on Queen.
15-02-01 mbo queen_tweak diff
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 8856 W: 1746 L: 1829 D: 5281
sprt @ 15+0.05 th 1 Sometimes the queen ends up passive on the first rank behind own pawns. Adjusts PST, and removes mobility for the first rank.
15-02-01 jos spsa_queen_imbal diff
50339/50000 iterations
102000/100000 games played
100000 @ 15+0.05 th 1 Try to tune material imbalance values of the queen. They seem to have the most impact elowise.
15-02-01 Roc AttackBuildUp diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 3916 W: 745 L: 841 D: 2330
sprt @ 15+0.05 th 1 Half the weight of the first test submitted.
15-02-01 SC search_tempo diff
ELO: -0.14 +-4.4 (95%) LOS: 47.6%
Total: 10000 W: 2128 L: 2132 D: 5740
10000 @ 9+0.03 th 1 Do not add tempo during evaluation and move it completely to search. Now tempo is being added for specialized endgames. It should not change anything (apart of specialized endgames).
15-02-02 Roc LatentQAttack diff
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 17700 W: 3503 L: 3561 D: 10636
sprt @ 15+0.05 th 1 Fix # 2. Only scoring Minor attacks. If it does not work, next step will be to lower the weight.