Stockfish Testing Queue

Finished - 33316 tests

15-12-09 pec tuned_nfl diff
LLR: 1.09 (-2.94,2.94) [-3.00,1.00]
Total: 37253 W: 4862 L: 4856 D: 27535
sprt @ 100+1 th 1 LTC. Simplification(removal of hard stops ) + ponder off bug fix
15-12-09 lbr tune diff
10102/40000 iterations
20542/80000 games played
80000 @ 20+0.2 th 1 last try. this time tune at longer tc, because king safety tuning at stc gave unscalable results.
15-12-10 Roc TestNewThreat diff
LLR: -0.09 (-2.94,2.94) [0.00,5.00]
Total: 209 W: 31 L: 34 D: 144
sprt @ 60+0.6 th 1 LTC (60+0.6) for tuned parameters after 117000 games Most notable change compared to last master is mThreatenedByPawn_4 which raised from 180 to 217
15-12-10 Roc TestNewThreat diff
LLR: 0.08 (-2.94,2.94) [0.00,4.00]
Total: 1263 W: 159 L: 152 D: 952
sprt @ 100+1 th 1 LTC for tuned parameters after 117000 games Most notable change compared to last master is mThreatenedByPawn_4 which raised from 180 to 217
15-12-09 IIv low_depths diff
94450/100000 iterations
200000/200000 games played
200000 @ 2+0.02 th 1 Correcting time controls. Sorry for stopping the previous run, but this will be much faster and more correct.
15-12-10 jos tune_trapped_rook diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 15691 W: 3008 L: 3111 D: 9572
sprt @ 10+0.1 th 1 Re-tuning of penalty for trapped rook.
15-12-09 pec tuned_nfl diff
ELO: 1.39 +-2.7 (95%) LOS: 84.7%
Total: 20000 W: 3104 L: 3024 D: 13872
20000 @ 40+0.4 th 1 Best version for longer tc. It seems huge jump form 10+0.1 to 100+1 so I measure elo first at 40+0.4 to see if there is deteoration and fine tuning at longer tc is needed. Simplification(removal of hard stops ) + ponder off bug fix
15-12-09 sni bishops4 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 26740 W: 5176 L: 5183 D: 16381
sprt @ 10+0.1 th 1 Opposite bishop scaling by number of pawns (Take 1, for BXs vs BXs)
15-12-09 Roc TestNewThreat diff
LLR: 2.96 (-2.94,2.94) [0.00,4.00]
Total: 13189 W: 2595 L: 2390 D: 8204
sprt @ 10+0.1 th 1 Tuned parameters after 117000 games Most notable change compared to last master is mThreatenedByPawn_4 which raised from 180 to 217
15-12-09 sni bishops4 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 16309 W: 3102 L: 3156 D: 10051
sprt @ 10+0.1 th 1 Opposite bishop scaling by number of pawns (Take 2, for B vs B)
15-12-09 jos bad_bishop diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 10310 W: 1998 L: 2079 D: 6233
sprt @ 10+0.1 th 1 Reduced penalties and slightly reduced material bonus in the imbalance table. Take 3.
15-12-06 Roc newthreattune diff
56058/60000 iterations
117000/120000 games played
120000 @ 10+0.1 th 1 Tuning the individual threat scores which I multiplied by S(404, 241) / 256
15-12-09 IIv low_depths diff
8663/100000 iterations
17970/200000 games played
200000 @ 2+0.2 th 1 Running corrected test. Once again sorry for my mistakes.
15-12-08 IIv low_depths diff
56190/200000 iterations
159271/400000 games played
400000 @ 2+0.02 th 1 Tuning reductions for the first 5 depths. Using very short time controls. The purposes of this tuning are: 1) improve reduction formula; 2) investigate performance of SF on low depths as explained on the forum.
15-12-08 pec tune_nfl diff
ELO: 0.52 +-3.0 (95%) LOS: 63.3%
Total: 20000 W: 3894 L: 3864 D: 12242
20000 @ 10+0.1 th 1 Progress after last 39.2K games
15-12-09 pb0 distinct_iter_paths_4 diff
ELO: 2.30 +-13.7 (95%) LOS: 62.9%
Total: 908 W: 169 L: 163 D: 576
20000 @ 5+0.1 th 7 Last check before proceeding to LLR. Testing resume feature.
15-12-08 lbr safety diff
LLR: -2.95 (-2.94,2.94) [-3.00,1.00]
Total: 6847 W: 898 L: 1057 D: 4892
sprt @ 100+1 th 1 LTC: simplify King Safety
15-12-08 pec tune_nfl diff
19867/20000 iterations
39219/40000 games played
40000 @ 10+0.1 th 1 ok, continue tuning where I stopped it. Seems it is going somewhere as check showing +5 elo after 9K games
15-12-07 pb0 distinct_iter_paths3 diff
ELO: 5.47 +-2.8 (95%) LOS: 100.0%
Total: 20000 W: 3605 L: 3290 D: 13105
20000 @ 5+0.1 th 7 Quick test with half density. (Now referring correct git-branch)
15-12-08 lbr safety diff
LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 35232 W: 7251 L: 7156 D: 20825
sprt @ 10+0.1 th 1 simplify King Safety
15-12-08 Voy hpf2 diff
LLR: 2.95 (-2.94,2.94) [0.00,5.00]
Total: 8813 W: 1769 L: 1603 D: 5441
sprt @ 10+0.1 th 1 Take 2. Less aggressive pruning at higher rootDepth.
15-12-08 Voy hpf diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 19402 W: 2786 L: 2837 D: 13779
sprt @ 60+0.4 th 1 LTC: History Pruning Formula.
15-12-08 pec tune_nfl diff
ELO: 3.51 +-3.0 (95%) LOS: 99.0%
Total: 20000 W: 3898 L: 3696 D: 12406
20000 @ 10+0.1 th 1 It looks like previous tuning run went completely wrong. I made small mistake in specifying initial parameters, and it seems parameteres started to move out of optimum. Check.
15-12-07 pec tune_nfl diff
ELO: 1.36 +-3.0 (95%) LOS: 81.5%
Total: 20000 W: 3833 L: 3755 D: 12412
20000 @ 10+0.1 th 1 Check after additional 39K games. FL still continues to drop
15-12-07 lbr tune diff
31120/50000 iterations
62922/100000 games played
100000 @ 10+0.1 th 1 tune safety: take 2
15-12-07 Voy hpf diff
LLR: 2.95 (-2.94,2.94) [0.00,5.00]
Total: 22910 W: 4515 L: 4286 D: 14109
sprt @ 10+0.1 th 1 History Pruning Formula.
15-12-07 pec tune_nfl diff
12904/20000 iterations
24313/40000 games played
40000 @ 10+0.1 th 1 next 40k games with FL initiaized even lower as it still did not converge
15-12-07 adp TunedRazor diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 22153 W: 4296 L: 4375 D: 13482
sprt @ 10+0.1 th 1 Tuned Razor Margin Values
15-12-07 pec tune_nfl diff
19734/20000 iterations
39970/40000 games played
40000 @ 10+0.1 th 1 next 40k games with FL initiaized lower as it did not converge
15-12-07 Voy PSQTt diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 32313 W: 6272 L: 6254 D: 19787
sprt @ 10+0.1 th 1 PSQT tweaks (fixed bench)
15-12-05 adp RazorTuning diff
39489/40000 iterations
80000/80000 games played
80000 @ 20+0.2 th 1 Tuning of razor margin values, earlier attempt introduced some noise probably, a good time to tune it further and better, hopefully it will add some ELO.
15-12-07 pb0 distinct_iter_paths2 diff
ELO: -0.00 +-11.4 (95%) LOS: 50.0%
Total: 1255 W: 221 L: 221 D: 813
20000 @ 5+0.1 th 7 Quick test with half density (in avg each 2nd ply get skipped)
15-12-07 mbo ponder_off_tm diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 8433 W: 1577 L: 1666 D: 5190
sprt @ 10+0.1 th 1 Tuned time management values.
15-12-07 pec tune_nfl diff
ELO: -1.34 +-2.9 (95%) LOS: 18.7%
Total: 20000 W: 3719 L: 3796 D: 12485
20000 @ 10+0.1 th 1 Check after 36K games
15-12-07 pb0 distinct_iter_paths2 diff
ELO: -0.55 +-4.0 (95%) LOS: 39.5%
Total: 9522 W: 1588 L: 1603 D: 6331
20000 @ 5+0.1 th 7 Another quick test with smoothed formula (no expection for thread with maximum idx)
15-12-06 mbo ponder_off_tune diff
36601/40000 iterations
74147/80000 games played
80000 @ 10+0.1 th 1 Tune time management with ponder set to false. 2nd try because XShift was off by factor 10. I also changed Slow Mover to start at 90 (the value from previous single parameter tuning).
15-12-07 pec tune_nfl diff
19114/20000 iterations
39449/40000 games played
40000 @ 10+0.1 th 1 tune time management without first root moves
15-12-06 Voy fhbt2 diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 42406 W: 8162 L: 8169 D: 26075
sprt @ 10+0.1 th 1 Take 2
15-12-07 Voy fht2 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 11850 W: 2232 L: 2306 D: 7312
sprt @ 10+0.1 th 1 Take 2.
15-12-06 tzi RookalignedPawnsKing diff
LLR: -0.16 (-2.94,2.94) [0.00,5.00]
Total: 1313 W: 257 L: 258 D: 798
sprt @ 10+0.1 th 1 Bonus for rook aligning with enemy king
15-12-06 Voy fhbt diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 17521 W: 3359 L: 3455 D: 10707
sprt @ 10+0.1 th 1 Increase fail low bonus.
15-12-06 sni threats9 diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 13963 W: 2683 L: 2748 D: 8532
sprt @ 10+0.1 th 1 Try different threats accumulator: take 3 (prefer threats on higher ranks)
15-12-06 lbr safety diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 38170 W: 7280 L: 7303 D: 23587
sprt @ 10+0.1 th 1 tuned safety
15-12-06 sni threats9 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 12856 W: 2393 L: 2463 D: 8000
sprt @ 10+0.1 th 1 Try different threats accumulator: take 2 (prefer threats near opponent king)
15-12-06 sni threats9 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 11800 W: 2287 L: 2361 D: 7152
sprt @ 10+0.1 th 1 Try different threats accumulator: take 1 (threat on most valuable piece)
15-12-04 n_p ThreatWeight diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 65965 W: 9541 L: 9503 D: 46921
sprt @ 60+0.4 th 1 LTC: Is pushing the MG value of threat weight higher, even better?
15-12-06 Voy fht diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 37623 W: 7306 L: 7264 D: 23053
sprt @ 10+0.1 th 1 Higher bonus if the move that failed high is a non-pv move.
15-12-05 mbo ponder_off_tune diff
27982/40000 iterations
56308/80000 games played
80000 @ 10+0.1 th 1 Tune time management with ponder set to false.
15-12-05 lbr safety diff
32440/50000 iterations
64873/100000 games played
100000 @ 10+0.1 th 1 tune king safety
15-12-05 jos no_altering diff
ELO: 3.09 +-4.0 (95%) LOS: 93.6%
Total: 10000 W: 1757 L: 1668 D: 6575
10000 @ 5+0.1 th 7 Combination of pb00067's 'distinct_iter_paths2' and my 'no_altering', that we would like to see tested.