Stockfish Testing Queue

Finished - 29280 tests

16-01-30 IIv SPSA_benchmark_tune diff
4984/5000 iterations
10000/10000 games played
10000 @ 10+0.1 th 1 Measuring a noise of the framework. A version without nodestime. Take 4.
16-01-30 Roc RookAlignedPieces diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 14237 W: 2582 L: 2646 D: 9009
sprt @ 10+0.1 th 1 Extend the RookOnPawn bonus to include alignment with opponent non-pawns,
16-01-29 Mys ph1 diff
ELO: -0.56 +-3.8 (95%) LOS: 38.7%
Total: 10000 W: 1550 L: 1566 D: 6884
10000 @ 30+0.3 th 1 Pawn hash quick measure #1.1 - TC scaling
16-01-30 lan loss_measurement diff
ELO: -42.78 +-4.5 (95%) LOS: 0.0%
Total: 10000 W: 1609 L: 2834 D: 5557
10000 @ 10+0.1 th 1 Measurement of the loss function with QueenValueMg=3500
16-01-30 Voy lmrt diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 3499 W: 608 L: 719 D: 2172
sprt @ 10+0.1 th 1 LMR tweak
16-01-30 lan loss_measurement diff
ELO: -701.75 +-29.6 (95%) LOS: 0.0%
Total: 5000 W: 24 L: 4851 D: 125
10000 @ 10+0.1 th 1 Measurement of the loss function with QueenValueMg = -500
16-01-30 Voy cm diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 19564 W: 3590 L: 3631 D: 12343
sprt @ 10+0.1 th 1 update cm for both fail high and singular extension
16-01-29 pb0 quasi_half diff
ELO: 0.21 +-3.6 (95%) LOS: 54.5%
Total: 10000 W: 1407 L: 1401 D: 7192
10000 @ 10+0.1 th 11 Testing a quasi-half-density approach on 11 threads. Low throughput
16-01-30 IIv rf_simplification diff
LLR: -2.96 (-2.94,2.94) [-3.00,1.00]
Total: 4598 W: 756 L: 921 D: 2921
sprt @ 10+0.1 th 1 Reduction formula simplification.
16-01-30 IIv SPSA_benchmark_tune diff
4959/5000 iterations
10000/10000 games played
10000 @ 10+0.1 th 1 Measuring a noise of the framework. A version with nodestime. Take4.
16-01-30 hxi noweights diff
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 15173 W: 2790 L: 2659 D: 9724
sprt @ 10+0.1 th 1 no weights
16-01-30 Voy se-cm_cmh diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 43930 W: 8198 L: 8133 D: 27599
sprt @ 10+0.1 th 1 Update both CM and CMH at singular extensions...
16-01-30 Voy fh-cm diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 34650 W: 6413 L: 6389 D: 21848
sprt @ 10+0.1 th 1 Update CM if a move caused a fail high.
16-01-30 Voy ft diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 28855 W: 5406 L: 5406 D: 18043
sprt @ 10+0.1 th 1 Futility Tweak
16-01-29 IIv SPSA_benchmark_tune diff
4968/5000 iterations
10000/10000 games played
10000 @ 10+0.1 th 1 Measuring a noise of the framework. A version without nodestime. Take 3.
16-01-29 IIv SPSA_benchmark_tune diff
4965/5000 iterations
10000/10000 games played
10000 @ 10+0.1 th 1 Measuring a noise of the framework. A version without nodestime. Take 2.
16-01-29 Mys ph2 diff
ELO: 0.17 +-4.1 (95%) LOS: 53.3%
Total: 10000 W: 1852 L: 1847 D: 6301
10000 @ 10+0.1 th 1 Pawn hash quick measure #2
16-01-29 Mys ph1 diff
ELO: 2.36 +-4.2 (95%) LOS: 86.5%
Total: 10000 W: 1934 L: 1866 D: 6200
10000 @ 10+0.1 th 1 Pawn hash quick measure #1
16-01-29 Voy SE-CMH diff
LLR: -2.98 (-2.94,2.94) [0.00,5.00]
Total: 37034 W: 6880 L: 6846 D: 23308
sprt @ 10+0.1 th 1 Update CMH bonus for singular extension moves.
16-01-29 Voy razor diff
LLR: -2.96 (-2.94,2.94) [-3.00,1.00]
Total: 39063 W: 7090 L: 7315 D: 24658
sprt @ 10+0.1 th 1 Razor Tweak
16-01-29 IIv SPSA_benchmark_tune diff
4977/5000 iterations
10000/10000 games played
10000 @ 10+0.1 th 1 Measuring a noise of the framework. A version with nodestime. Take3.
16-01-29 IIv SPSA_benchmark_tune diff
3938/5000 iterations
7918/10000 games played
10000 @ 10+0.1 th 1 Measuring a noise of the framework. A version with nodestime. Agressive ck and rk values (x6, x10).
16-01-29 IIv SPSA_benchmark_tune diff
4966/5000 iterations
10000/10000 games played
10000 @ 10+0.1 th 1 Measuring a noise of the framework. A version with nodestime. Tripling both ck and rk values.
16-01-29 IIv SPSA_benchmark_tune diff
4937/5000 iterations
10000/10000 games played
10000 @ 10+0.1 th 1 Measuring a noise of the framework. A version with nodestime. Take2.
16-01-26 n_p Lazy_tt diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 92285 W: 13617 L: 13501 D: 65167
sprt @ 20+0.2 th 3 LTC: Low hash pressure.
16-01-28 IIv SPSA_benchmark_tune diff
4987/5000 iterations
10000/10000 games played
10000 @ 10+0.1 th 1 Measuring a noise of the framework. A version with nodestime.
16-01-28 IIv SPSA_benchmark_tune diff
4977/5000 iterations
10000/10000 games played
10000 @ 10+0.1 th 1 Measuring a noise of the framework. I set priority to -1 so far, because I'm not 100% sure that the code is correct.
16-01-28 mbo calculate_mobility diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 13220 W: 2404 L: 2473 D: 8343
sprt @ 10+0.1 th 1 Reduces parameters for mobility. Take 2. Lower throughput.
16-01-23 Mys BT diff
LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 40639 W: 5555 L: 5461 D: 29623
sprt @ 20+0.2 th 7 LTC: In the past I've noticed thread depth to be less effective than one would think for selecting best moves. I'm hoping this tiny simplification might even add strength.
16-01-28 Voy slr diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 19773 W: 2573 L: 2669 D: 14531
sprt @ 60+0.6 th 1 LTC: Don't skip razoring if skipped early pruning.
16-01-28 jos lmr_tweak diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 10969 W: 1998 L: 2076 D: 6895
sprt @ 10+0.1 th 1 LMR tweak. Maybe this helps to reach higher depths in endgames.
16-01-28 sg update_stats1 diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 10649 W: 1920 L: 2000 D: 6729
sprt @ 10+0.1 th 1 update stats also for moves with alpha < value < beta. Use depth/2. Take 2
16-01-28 sg update_stats1 diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 8730 W: 1549 L: 1637 D: 5544
sprt @ 10+0.1 th 1 update stats also for moves with alpha < value < beta
16-01-28 sg MinorBehindPawn diff
LLR: -2.96 (-2.94,2.94) [0.00,5.00]
Total: 5811 W: 1021 L: 1122 D: 3668
sprt @ 10+0.1 th 1 Test with new tuned values
16-01-26 Voy se-cm diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 85201 W: 11746 L: 11575 D: 61880
sprt @ 60+0.6 th 1 LTC: FINALLY ended my Red/Yellow curse. Update CM at singular extension.
16-01-27 jhe prob_cut diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 4114 W: 727 L: 836 D: 2551
sprt @ 10+0.1 th 1 Apply improving criteria to probcut.
16-01-27 sg tune_MinorBehindPawn diff
19696/20000 iterations
40000/40000 games played
40000 @ 10+0.1 th 1 The tuned values give small elo but not enough. Furthermore some values seems to be not converged, So try a second tuning from the last values. ck values seems ok, but i double up rk values for faster movement.
16-01-27 pb0 depth_dependend_margin diff
ELO: -0.83 +-4.2 (95%) LOS: 34.8%
Total: 10000 W: 1875 L: 1899 D: 6226
10000 @ 10+0.1 th 1 Trying to prove that depth margin in TT-save has dependence to rootDepth
16-01-27 Voy slr diff
LLR: 2.96 (-2.94,2.94) [0.00,4.00]
Total: 20795 W: 3927 L: 3698 D: 13170
sprt @ 10+0.1 th 1 Don't skip razoring if skipped early pruning.
16-01-27 mbo calculate_mobility diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 9797 W: 1758 L: 1842 D: 6197
sprt @ 10+0.1 th 1 Reduces parameters for mobility.
16-01-27 sg MinorBehindPawn diff
ELO: 0.69 +-2.7 (95%) LOS: 69.4%
Total: 24251 W: 4514 L: 4466 D: 15271
30000 @ 10+0.1 th 1 Quick measure of tuned values. Probably a further tuning is needed.
16-01-26 sg tune_MinorBehindPawn diff
19640/20000 iterations
39813/40000 games played
40000 @ 10+0.1 th 1 From my experiments so fare it seems minor behind own pawn is more important than behind opponent pawn. Additionally logical this cases are different situations. Par example 'behind' a opponent pawn means blocking them or its seems illogical to give a bonus for a minor on the base rank behind a own pawn. So i have generalized the bonus to an array indexed by own/opponent pawn and rank. Now try tuning the values.
16-01-26 sg MinorBehindPawn diff
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 39673 W: 7461 L: 7479 D: 24733
sprt @ 10+0.1 th 1 Try the opposite: Use MinorBehindPawn bonus only for own pawns
16-01-26 Voy se-cm diff
LLR: 2.95 (-2.94,2.94) [0.00,5.00]
Total: 7593 W: 1496 L: 1337 D: 4760
sprt @ 10+0.1 th 1 Take 2. Don't update cm if capture or promotions. (Noted by sg)
16-01-26 sg MinorBehindPawn diff
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 7867 W: 1431 L: 1562 D: 4874
sprt @ 10+0.1 th 1 Use MinorBehindPawn bonus only for opponent pawns
16-01-26 n_p Lazy_tt diff
LLR: 2.94 (-2.94,2.94) [-3.00,1.00]
Total: 38481 W: 7213 L: 7123 D: 24145
sprt @ 10+0.1 th 1 Try the most critical test, 1-thread no-regression and high hash pressure before the LTC test.
16-01-25 pec easy diff
ELO: -5.61 +-2.9 (95%) LOS: 0.0%
Total: 20000 W: 3504 L: 3827 D: 12669
20000 @ 10+0.1 th 1 Another naive definition of easy move
16-01-26 Voy se-cm diff
LLR: -2.97 (-2.94,2.94) [0.00,5.00]
Total: 30934 W: 5667 L: 5660 D: 19607
sprt @ 10+0.1 th 1 Update countermove at singular extension.
16-01-25 pb0 tt_quality_classes2 diff
ELO: -5.39 +-4.3 (95%) LOS: 0.7%
Total: 10000 W: 1929 L: 2084 D: 5987
10000 @ 2+0.02 th 7 Quality classes in TT for Lazy SMP. Attempt 4 (attempts 2 and 3 were bugged)
16-01-25 pec tm diff
ELO: 1.63 +-3.3 (95%) LOS: 83.1%
Total: 20000 W: 4873 L: 4779 D: 10348
20000 @ 5 th 1 Wondering what is effect of that very low depth condition which is accessible only in short sudden death games