| 14-01-10 |
uri |
sometimes_faster |
diff
|
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 1019 W: 149 L: 252 D: 618
|
sprt
@ 15+0.05 th 1 |
this time I test playing twice faster if more than 95% of the time is used for the first move(meaning that stockfish refuted relatively fast the rest of the moves) |
| 14-01-10 |
pec |
tm_simple |
diff
|
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 11396 W: 2075 L: 2150 D: 7171
|
sprt
@ 15+0.05 th 1 |
use 20% more time for stable pv at the expense of unstable pv |
| 14-01-10 |
sg |
update_stats |
diff
|
LLR: 2.96 (-2.94,2.94) [-1.50,4.50]
Total: 62429 W: 11697 L: 11429 D: 39303
|
sprt
@ 15+0.05 th 1 |
update stats forr pv moves too. Use weighted counter move stats. |
| 14-01-10 |
bin |
pawn_psqt |
diff
|
LLR: 1.54 (-2.94,2.94) [-1.50,4.50]
Total: 41006 W: 7790 L: 7627 D: 25589
|
sprt
@ 15+0.05 th 1 |
pawn psqt take 3 |
| 14-01-10 |
uri |
fixing_a_bug |
diff
|
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 16732 W: 3107 L: 3168 D: 10457
|
sprt
@ 15+0.05 th 1 |
I fix a bug by replacing a misleading name and play slower if the program had in the past a fail low in the root(there are more ideas like stopping in the first move at sometime in this case but I decided to add one piece of code at a time and not to do it immediately |
| 14-01-10 |
jos |
kingAttWeight |
diff
|
LLR: 2.93 (-2.94,2.94) [-1.50,4.50]
Total: 36037 W: 6854 L: 6655 D: 22528
|
sprt
@ 15+0.05 th 1 |
Increase KnightAttackWeight. |
| 14-01-10 |
uri |
f14cd1bb89d080f36a11df3 |
diff
|
ELO: -1.02 +-1.9 (95%) LOS: 14.7%
Total: 40000 W: 6177 L: 6294 D: 27529
|
40000
@ 60+0.05 th 1 |
Test Marco simplification of TM (I am going to apply the patch, but I want to be sure there is no regression) |
| 14-01-10 |
bin |
pawn_psqt |
diff
|
LLR: 2.95 (-2.94,2.94) [-1.50,4.50]
Total: 15932 W: 3055 L: 2908 D: 9969
|
sprt
@ 15+0.05 th 1 |
pawn psqt take 4 |
| 14-01-10 |
rst |
se_at_root3 |
diff
|
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 10975 W: 2011 L: 2088 D: 6876
|
sprt
@ 15+0.05 th 1 |
New try for singular extensions at root nodes (with a view to using singular moves for TM decisions) |
| 14-01-10 |
pec |
mbonus |
diff
|
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 44717 W: 8370 L: 8358 D: 27989
|
sprt
@ 15+0.05 th 1 |
just random shot from looking at results of measurement test |
| 14-01-10 |
uri |
fixing_a_bug |
diff
|
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 79235 W: 14739 L: 14638 D: 49858
|
sprt
@ 15+0.05 th 1 |
second try of fixing a bug and this time play faster in part of the cases when there was no fail low in the root. |
| 14-01-10 |
gli |
shelter_impeded |
diff
|
LLR: -0.35 (-2.94,2.94) [-1.50,4.50]
Total: 8004 W: 1532 L: 1523 D: 4949
|
sprt
@ 15+0.05 th 1 |
No bonus for castling king safety if impeded |
| 14-01-10 |
joa |
supporting_pawns_1 |
diff
|
ELO: -0.80 +-2.9 (95%) LOS: 29.7%
Total: 20000 W: 3685 L: 3731 D: 12584
|
20000
@ 15+0.05 th 1 |
PP: Measuring supporting pawns 1 |
| 14-01-10 |
joa |
supporting_pawns_2 |
diff
|
ELO: -0.50 +-2.9 (95%) LOS: 36.8%
Total: 20000 W: 3700 L: 3729 D: 12571
|
20000
@ 15+0.05 th 1 |
PP: Measuring supporting pawns 2 |
| 14-01-10 |
gli |
storm_blocked |
diff
|
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 17142 W: 3173 L: 3233 D: 10736
|
sprt
@ 15+0.05 th 1 |
King blockading pawn storm |
| 14-01-10 |
gli |
storm_blocked |
diff
|
LLR: -2.92 (-2.94,2.94) [-1.50,4.50]
Total: 13655 W: 2596 L: 2664 D: 8395
|
sprt
@ 15+0.05 th 1 |
Even less penalty when king blockading storm pawn |
| 14-01-10 |
joa |
rook_pawns_1 |
diff
|
ELO: -0.36 +-2.9 (95%) LOS: 40.4%
Total: 20000 W: 3690 L: 3711 D: 12599
|
20000
@ 15+0.05 th 1 |
PP: Measure rook pawns 1 |
| 14-01-10 |
joa |
rook_pawns_2 |
diff
|
ELO: -1.70 +-2.9 (95%) LOS: 12.9%
Total: 20000 W: 3709 L: 3807 D: 12484
|
20000
@ 15+0.05 th 1 |
PP: Measure rook pawns 2 |
| 14-01-10 |
gli |
pawn_psqt |
diff
|
LLR: -0.75 (-2.94,2.94) [0.00,6.00]
Total: 25943 W: 4082 L: 3990 D: 17871
|
sprt
@ 60+0.05 th 1 |
Long TC for BI: pawn psqt take 4 |
| 14-01-10 |
joa |
less_pawns |
diff
|
ELO: -3.01 +-3.0 (95%) LOS: 2.3%
Total: 20000 W: 3681 L: 3854 D: 12465
|
20000
@ 15+0.05 th 1 |
PP: Measure less pawns bonus |
| 14-01-10 |
gli |
kingAttWeight |
diff
|
LLR: -0.75 (-2.94,2.94) [0.00,6.00]
Total: 50783 W: 7969 L: 7765 D: 35049
|
sprt
@ 60+0.05 th 1 |
Long TC for joster: Increase KnightAttackWeight |
| 14-01-10 |
bin |
pawn_psqt |
diff
|
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 20580 W: 3968 L: 4019 D: 12593
|
sprt
@ 15+0.05 th 1 |
pawn table 5 |
| 14-01-10 |
rst |
se_at_root3 |
diff
|
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 5033 W: 896 L: 988 D: 3149
|
sprt
@ 15+0.05 th 1 |
take 2 |
| 14-01-10 |
rst |
se_at_root3 |
diff
|
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 5347 W: 971 L: 1063 D: 3313
|
sprt
@ 15+0.05 th 1 |
take 3 (final) |
| 14-01-10 |
gli |
storm_blocked |
diff
|
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 9826 W: 1816 L: 1896 D: 6114
|
sprt
@ 15+0.05 th 1 |
Don't count blocked files in kingRing |
| 14-01-11 |
uri |
fixing_a_bug |
diff
|
LLR: 2.96 (-2.94,2.94) [-1.50,4.50]
Total: 19446 W: 3696 L: 3540 D: 12210
|
sprt
@ 15+0.05 th 1 |
another version of the same idea and now added the real intention of the code that stopped at the first move after significant time of no fail low because I believe that the intention was to failing low in the specific iteration and not to some history of failLow(maybe the history does not give much if it was many plies earlier but I prefered not to get rid of that code.because I still feel better to play faster if there was no fail low in all iterations |
| 14-01-11 |
rst |
blocked_storm_pawn_radi |
diff
|
LLR: 2.95 (-2.94,2.94) [-1.50,4.50]
Total: 6239 W: 1249 L: 1127 D: 3863
|
sprt
@ 15+0.05 th 1 |
Radical solution to get king safety eval of this position r2qk2r/ppp2p2/2npbn2/2b1p3/2P1P1P1/2NB1PPp/PPNP3K/R1BQ1R2 b kq - 0 13 about right |
| 14-01-11 |
gli |
storm_blocked |
diff
|
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 14089 W: 2578 L: 2646 D: 8865
|
sprt
@ 15+0.05 th 1 |
Don't remove penalty for shelter, but blocked files don't count against king ring |
| 14-01-11 |
bin |
pawn_psqt |
diff
|
LLR: -1.54 (-2.94,2.94) [-1.50,4.50]
Total: 34424 W: 6396 L: 6361 D: 21667
|
sprt
@ 15+0.05 th 1 |
pawn psqt take 6 |
| 14-01-11 |
sto |
ralphsidea_kingsafety |
diff
|
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 31228 W: 5769 L: 5793 D: 19666
|
sprt
@ 15+0.05 th 1 |
Idea taken out from Ralph's own idea. This time with a lower bonus for king safety. |
| 14-01-11 |
rst |
blocked_storm_pawn_radi |
diff
|
LLR: 2.96 (-2.94,2.94) [0.00,6.00]
Total: 38371 W: 6165 L: 5888 D: 26318
|
sprt
@ 60+0.05 th 1 |
LTC: Radical solution to get king safety eval of this position r2qk2r/ppp2p2/2npbn2/2b1p3/2P1P1P1/2NB1PPp/PPNP3K/R1BQ1R2 b kq - 0 13 about right |
| 14-01-11 |
rst |
pp_blocked |
diff
|
LLR: 2.96 (-2.94,2.94) [-1.50,4.50]
Total: 74017 W: 13857 L: 13559 D: 46601
|
sprt
@ 15+0.05 th 1 |
Blockage of rook passers is difficult eliminate |
| 14-01-11 |
uri |
fixing_a_bug |
diff
|
LLR: -2.95 (-2.94,2.94) [0.00,6.00]
Total: 20272 W: 3160 L: 3169 D: 13943
|
sprt
@ 60+0.05 th 1 |
another version of the same idea and now added the real intention of the code that stopped at the first move after significant time of no fail low because I believe that the intention was to failing low in the specific iteration and not to some history of failLow(maybe the history does not give much if it was many plies earlier but I prefered not to get rid of that code.because I still feel better to play faster if there was no fail low in all iterations |
| 14-01-11 |
jos |
KBBKN_fix |
diff
|
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 11770 W: 2114 L: 2188 D: 7468
|
sprt
@ 15+0.05 th 1 |
Fixes a bug in KBBKN. Also return lower score, as many wins may need more than 50 moves. |
| 14-01-11 |
mco |
stsu |
diff
|
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 15176 W: 2765 L: 2831 D: 9580
|
sprt
@ 15+0.05 th 1 |
Verify a supposed +49 ELO (!) mega patch series |
| 14-01-11 |
hwi |
king_val |
diff
|
LLR: -0.36 (-2.94,2.94) [-1.50,4.50]
Total: 1486 W: 273 L: 282 D: 931
|
sprt
@ 15+0.05 th 1 |
Assign king big piece val |
| 14-01-11 |
gli |
storm_blocked |
diff
|
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 23724 W: 4436 L: 4479 D: 14809
|
sprt
@ 15+0.05 th 1 |
Same as before, except king safety parameters tuned with 25k games. |
| 14-01-12 |
uri |
try2_of_fixing_a_bug |
diff
|
LLR: -2.94 (-2.94,2.94) [-1.50,4.50]
Total: 19872 W: 3622 L: 3675 D: 12575
|
sprt
@ 15+0.05 th 1 |
changed the meaning of Signals.failedLowAtRoot to the logical meaning and I do not think that it is logical to consider fail low that happened at iteration 5 for decision about time to stop in iteration 20.
It may be logical to consider fail low in iteration 19 for iteration 20 so it is possible that this patch is going to fail. |
| 14-01-12 |
hwi |
see_king |
diff
|
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 16261 W: 3031 L: 3094 D: 10136
|
sprt
@ 15+0.05 th 1 |
Changing Position::see to return a low value for illegal captures |
| 14-01-12 |
bin |
stsu |
diff
|
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 51165 W: 9509 L: 9481 D: 32175
|
sprt
@ 15+0.05 th 1 |
check material.cpp only |
| 14-01-12 |
uri |
try2_of_fixing_a_bug |
diff
|
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 31181 W: 5756 L: 5780 D: 19645
|
sprt
@ 15+0.05 th 1 |
second try and this time I add a variable to count number of consecutive time that the program does not fail low.
Instead of never stopping in the first move when there was a fail low I simply do not stop in the first move only if there was a fail low in one of the last 6 iterations |
| 14-01-12 |
sg |
fix_countermoves |
diff
|
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 15633 W: 2844 L: 2908 D: 9881
|
sprt
@ 15+0.05 th 1 |
Fix missing check for countermoves |
| 14-01-12 |
lbr |
mate |
diff
|
LLR: 0.62 (-2.94,2.94) [-1.50,4.50]
Total: 128000 W: 24010 L: 23653 D: 80337
|
sprt
@ 15+0.05 th 1 |
Mate detection |
| 14-01-12 |
bin |
stsu |
diff
|
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 1448 W: 274 L: 380 D: 794
|
sprt
@ 15+0.05 th 1 |
check draw detection |
| 14-01-12 |
uri |
test_value_of_first_mov |
diff
|
ELO: -13.16 +-2.9 (95%) LOS: 0.0%
Total: 20000 W: 3374 L: 4131 D: 12495
|
20000
@ 15+0.05 th 1 |
test the value of the first move code(I guess that it should be significant with 20000 games but I prefer to see it to estimate how much I can earn by playing faster during first move if I can earn) |
| 14-01-12 |
jki |
npsqt |
diff
|
ELO: 0.99 +-2.2 (95%) LOS: 80.8%
Total: 40000 W: 8611 L: 8497 D: 22892
|
40000
@ 5+0.05 th 1 |
npsqt, 10k iter |
| 14-01-12 |
bin |
stsu |
diff
|
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 5864 W: 1028 L: 1118 D: 3718
|
sprt
@ 15+0.05 th 1 |
check wideness |
| 14-01-12 |
hwi |
see_king_2 |
diff
|
LLR: 2.95 (-2.94,2.94) [-1.50,4.50]
Total: 12359 W: 2411 L: 2273 D: 7675
|
sprt
@ 15+0.05 th 1 |
Make Position::see return low value for illegal captures. Optimized. Leave see_sign alone. |
| 14-01-12 |
bin |
stsu |
diff
|
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 1329 W: 234 L: 339 D: 756
|
sprt
@ 15+0.05 th 1 |
check rammed (last check, since 2 tests failed miserably and one neutral, this must be a gem) |
| 14-01-12 |
sg |
update_stats |
diff
|
LLR: -2.97 (-2.94,2.94) [0.00,6.00]
Total: 27706 W: 4327 L: 4303 D: 19076
|
sprt
@ 60+0.05 th 1 |
LTC: update stats forr pv moves too. Use weighted counter move stats. |