| 15-02-09 |
SC |
tune_tempo |
diff
|
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 52492 W: 10509 L: 10476 D: 31507
|
sprt
@ 15+0.05 th 1 |
In all the tempo tuning SPSA, tempo was increased in average. Is tempo +1 enough? |
| 15-02-09 |
SC |
tune_tempo |
diff
|
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 55084 W: 10923 L: 10882 D: 33279
|
sprt
@ 15+0.05 th 1 |
And tempo +2? |
| 15-02-09 |
Roc |
LooseKnight |
diff
|
LLR: -2.97 (-2.94,2.94) [0.00,6.00]
Total: 23032 W: 3879 L: 3871 D: 15282
|
sprt
@ 60+0.05 th 1 |
With S(20,0) |
| 15-02-10 |
jos |
spsa_rook_imbal |
diff
|
52723/50000 iterations
103000/100000 games played
|
100000
@ 15+0.05 th 1 |
First tuning session, rook values. |
| 15-02-10 |
sni |
any_safe_push2 |
diff
|
LLR: -0.18 (-2.94,2.94) [0.00,6.00]
Total: 892 W: 151 L: 153 D: 588
|
sprt
@ 60+0.05 th 1 |
LTC: Add small bonus for all safe pawn pushes (with the KingDanger[] array initialization fix) |
| 15-02-10 |
SC |
search_tempo_game_phase |
diff
|
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 9934 W: 1968 L: 2048 D: 5918
|
sprt
@ 15+0.05 th 1 |
Game-phase based tempo evaluation. Probe material table instead of computing game phase every time. Same bench and speed-up 0.50% wrt the version which failed STC. |
| 15-02-10 |
vin |
passed_blockers2 |
diff
|
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 6149 W: 1188 L: 1278 D: 3683
|
sprt
@ 15+0.05 th 1 |
Test SPSA-tuned values at STC. |
| 15-02-10 |
n_p |
KingSafety |
diff
|
ELO: -0.46 +-2.1 (95%) LOS: 33.7%
Total: 40000 W: 7933 L: 7986 D: 24081
|
40000
@ 15+0.05 th 1 |
Test the new values on king safety from the SPSA-session. |
| 15-02-10 |
vin |
en_passant_bonus |
diff
|
LLR: 2.96 (-2.94,2.94) [-1.50,4.50]
Total: 17846 W: 3620 L: 3464 D: 10762
|
sprt
@ 15+0.05 th 1 |
Test idea from Lyudmil Tsvetkov, adding small bonus to 5th rank pawns versus opponent 2nd rank potential en passant pawns. |
| 15-02-11 |
lbr |
detempletize |
diff
|
LLR: -2.98 (-2.94,2.94) [-3.00,1.00]
Total: 22505 W: 4333 L: 4535 D: 13637
|
sprt
@ 15+0.05 th 1 |
see pull request. non functional change, but bench may not be representative (depends on how frequently generate_castling is called) |
| 15-02-11 |
n_p |
KingSafety |
diff
|
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 46720 W: 7768 L: 7773 D: 31179
|
sprt
@ 60+0.05 th 1 |
LTC: Test the new values on king safety from the SPSA-session. Does these values scale well do to tuned in LTC. |
| 15-02-11 |
Fis |
stats_update |
diff
|
LLR: -3.38 (-2.94,2.94) [-1.50,4.50]
Total: 12402 W: 2438 L: 2526 D: 7438
|
sprt
@ 15+0.05 th 1 |
Stats::update() now clamps out of bound values to Max instead of ignoring them. See issue #251. If this passes STC, LTC will be at [0,4]. Pri -3 |
| 15-02-11 |
jos |
matimb |
diff
|
ELO: -1.56 +-2.1 (95%) LOS: 7.7%
Total: 40000 W: 7880 L: 8060 D: 24060
|
40000
@ 15+0.05 th 1 |
Check the new rook values after 1st tuning session. |
| 15-02-11 |
SC |
search_tempo_game_phase |
diff
|
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 7308 W: 1387 L: 1474 D: 4447
|
sprt
@ 15+0.05 th 1 |
A further try on tempo evaluation. Restore feature of ignoring tempo for specialized evaluations. (I removed it to avoid increasing the code too muchl). |
| 15-02-11 |
vin |
en_passant_bonus_spsa |
diff
|
20899/20000 iterations
40000/40000 games played
|
40000
@ 15+0.05 th 1 |
STC passed - Try a quick tune of values at STC. Could Lyudmil be onto another winner? |
| 15-02-11 |
Roc |
LooseKnight |
diff
|
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 28843 W: 5732 L: 5760 D: 17351
|
sprt
@ 15+0.05 th 1 |
Adding roughly tuned penalty for each loose piece (yes... including the King !) Let's see if it more than just noise. |
| 15-02-11 |
sni |
any_safe_push2 |
diff
|
LLR: 2.97 (-2.94,2.94) [0.00,6.00]
Total: 17684 W: 3042 L: 2854 D: 11788
|
sprt
@ 60+0.05 th 1 |
LTC: Add small bonus for all safe pawn pushes (with Stefan's fix for the bug in my code: now using popcount<full> instead of popcount<max15>) |
| 15-02-11 |
sg |
spsa_pawn_attack_threat |
diff
|
46633/50000 iterations
99018/100000 games played
|
100000
@ 15+0.05 th 1 |
Use different pawn attack threat bonus by piece type. Now tune this parameters, starting at value (20,20) from the current version. |
| 15-02-12 |
sg |
pawn_attack_threat4 |
diff
|
ELO: 1.21 +-2.8 (95%) LOS: 79.8%
Total: 23278 W: 4746 L: 4665 D: 13867
|
30000
@ 15+0.05 th 1 |
Quick measure of the First tuned parameters. Successful any safe pawn push patch is now merged in. |
| 15-02-12 |
Roc |
MoreDblAttacks |
diff
|
ELO: 0.80 +-3.3 (95%) LOS: 68.3%
Total: 16856 W: 3393 L: 3354 D: 10109
|
20000
@ 15+0.05 th 1 |
See if any value to consider different weights and also e4, d4 and file c and f for double pawn attack idea. |
| 15-02-12 |
sg |
spsa_pawn_attack_threat |
diff
|
46974/50000 iterations
97604/100000 games played
|
100000
@ 15+0.05 th 1 |
The first tuning is done without the any_safe_pawn2 patch. The measurement (now including any_safe_pawn2 patch) gives no significant gain and this two ideas seems strongly interacting as expected. Tuning now is done based on this passed patch. Only my new parameters tuned, not the 2 from the other patch, because we add code so it have to prove first by itself. |
| 15-02-12 |
Roc |
MoreDblAttacks |
diff
|
37570/40000 iterations
80000/80000 games played
|
80000
@ 15+0.05 th 1 |
SPSA test on all possible double pawn binds. Once we get the result, will look for a simplification. |
| 15-02-12 |
SC |
tune_tempo |
diff
|
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 38725 W: 6532 L: 6561 D: 25632
|
sprt
@ 60+0.05 th 1 |
Both tempo += 1 and tempo += 2 were "yellow" at STC. Test tempo += 2 at LTC and then call it a day. |
| 15-02-12 |
vin |
en_passant_bonus |
diff
|
LLR: 2.96 (-2.94,2.94) [-1.50,4.50]
Total: 11769 W: 2443 L: 2303 D: 7023
|
sprt
@ 15+0.05 th 1 |
Test the SPSA-tuned values at STC. They look implausibly high, but also spookily similar to Lyudmil's original estimate. |
| 15-02-12 |
vin |
en_passant_bonus |
diff
|
LLR: 2.95 (-2.94,2.94) [0.00,6.00]
Total: 43724 W: 7469 L: 7156 D: 29099
|
sprt
@ 60+0.05 th 1 |
Test of tuned values at LTC after STC SPRT test passed. |
| 15-02-13 |
Roc |
CenterLever |
diff
|
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 34876 W: 6927 L: 6938 D: 21011
|
sprt
@ 15+0.05 th 1 |
S(10,10) bonus for lever c4 against d5 and f4 against e5 (and symmetrical for Black). |
| 15-02-13 |
sg |
pawn_attack_threat4 |
diff
|
LLR: 2.95 (-2.94,2.94) [-1.50,4.50]
Total: 9307 W: 1916 L: 1784 D: 5607
|
sprt
@ 15+0.05 th 1 |
Now hopefully the correct test of the tuned parameters.
It's just not my day. |
| 15-02-13 |
Roc |
MoreDblAttacks |
diff
|
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 17932 W: 3530 L: 3624 D: 10778
|
sprt
@ 15+0.05 th 1 |
Testing the SPSA values for the double binds on center files only. There is more weight for the endgame. |
| 15-02-13 |
jos |
rook_filebonus |
diff
|
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 50560 W: 10077 L: 10045 D: 30438
|
sprt
@ 15+0.05 th 1 |
Give file-dependant bonus for rook on open file. Another idea by L. Tsvetkov. |
| 15-02-13 |
sg |
pawn_attack_threat4 |
diff
|
LLR: -2.95 (-2.94,2.94) [0.00,6.00]
Total: 10955 W: 1726 L: 1777 D: 7452
|
sprt
@ 60+0.05 th 1 |
LTC: Now hopefully the correct test of the tuned parameters.
It's just not my day. |
| 15-02-13 |
sni |
any_safe_push3 |
diff
|
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 47204 W: 9405 L: 9382 D: 28417
|
sprt
@ 15+0.05 th 1 |
Double the bonus for safe pawn pushes |
| 15-02-13 |
jki |
ep2 |
diff
|
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 4776 W: 895 L: 989 D: 2892
|
sprt
@ 15+0.05 th 1 |
safe pawn push + ep handling |
| 15-02-13 |
sni |
square_control |
diff
|
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 7182 W: 1413 L: 1500 D: 4269
|
sprt
@ 15+0.05 th 1 |
Add bonus for threatening piece domination |
| 15-02-14 |
Roc |
MoreDblAttacks |
diff
|
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 14688 W: 2833 L: 2939 D: 8916
|
sprt
@ 15+0.05 th 1 |
The SPSA values looked quite stable for dbl-attacked central squares on rank 5 and above, so keep only them, and see if the endgame bonus has any value. |
| 15-02-14 |
sg |
spsa_pawn_attack_threat |
diff
|
48900/50000 iterations
99385/100000 games played
|
100000
@ 60+0.05 th 1 |
The tunings on STC gives good results on STC but bad at LTC. So a strong TC dependency seems to exist. So do a last tuning try on LTC. |
| 15-02-14 |
vin |
en_passant_bonus_spsa |
diff
|
24973/20000 iterations
45000/40000 games played
|
40000
@ 15+0.05 th 1 |
Since there is possible overlap between this patch and safe pawn pushes, re-tune now the latter has been merged. |
| 15-02-14 |
sni |
square_control |
diff
|
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 11000 W: 2218 L: 2295 D: 6487
|
sprt
@ 15+0.05 th 1 |
Piece domination: take 2 bis (simpler than take 2, same bench) |
| 15-02-14 |
sni |
square_control |
diff
|
LLR: -3.20 (-2.94,2.94) [-1.50,4.50]
Total: 12257 W: 2403 L: 2485 D: 7369
|
sprt
@ 15+0.05 th 1 |
Piece domination: take 3 |
| 15-02-14 |
jos |
knight_outpost |
diff
|
LLR: -2.97 (-2.94,2.94) [-1.50,4.50]
Total: 6603 W: 1257 L: 1346 D: 4000
|
sprt
@ 15+0.05 th 1 |
Give bonus for knight outposts close to the enemy king. |
| 15-02-14 |
n_p |
MedKingSafety |
diff
|
ELO: -0.71 +-2.6 (95%) LOS: 29.5%
Total: 27864 W: 5568 L: 5625 D: 16671
|
40000
@ 15+0.05 th 1 |
King safety tuning using the last four SPSA-session and trying to use these results to extrapolate. |
| 15-02-14 |
jki |
pawnmob |
diff
|
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 6843 W: 1387 L: 1476 D: 3980
|
sprt
@ 15+0.05 th 1 |
Safe pawn push tweak try |
| 15-02-14 |
sni |
any_safe_push3 |
diff
|
LLR: -2.96 (-2.94,2.94) [-1.50,4.50]
Total: 7697 W: 1532 L: 1618 D: 4547
|
sprt
@ 15+0.05 th 1 |
Increase the bonus for safe pawn pushes in endgame |
| 15-02-14 |
mco |
SpaceThreshold |
diff
|
5171/50000 iterations
10509/100000 games played
|
100000
@ 15+0.05 th 1 |
Tune space evaluation threshold. I think it has never been properly tuned before. |
| 15-02-14 |
jki |
smp |
diff
|
ELO: 50.50 +-8.1 (95%) LOS: 100.0%
Total: 2591 W: 667 L: 293 D: 1631
|
5000
@ 15+0.05 th 16 |
smp improvement attempt (16 threads) |
| 15-02-14 |
jki |
pmob |
diff
|
LLR: -0.36 (-2.94,2.94) [-3.00,3.00]
Total: 13242 W: 2602 L: 2615 D: 8025
|
sprt
@ 15+0.05 th 1 |
Remove piece checks for safe pawn pushes. sprt [-3, 3] |
| 15-02-14 |
mco |
SpaceThreshold |
diff
|
21048/50000 iterations
43140/100000 games played
|
100000
@ 15+0.05 th 1 |
Tune space evaluation threshold. I think it has never been properly tuned before. Take 2 (wider changes) |
| 15-02-14 |
jki |
pmob |
diff
|
LLR: -3.85 (-2.94,2.94) [-3.00,1.00]
Total: 33801 W: 6682 L: 6954 D: 20165
|
sprt
@ 15+0.05 th 1 |
Remove piece checks for safe pawn pushes. No regression test. |
| 15-02-15 |
Roc |
Battery |
diff
|
LLR: -2.95 (-2.94,2.94) [-1.50,4.50]
Total: 7526 W: 1473 L: 1559 D: 4494
|
sprt
@ 15+0.05 th 1 |
Another attempt at the Q-> R battery idea. |
| 15-02-15 |
jki |
space |
diff
|
LLR: -3.92 (-2.94,2.94) [-1.50,4.50]
Total: 18337 W: 3718 L: 3809 D: 10810
|
sprt
@ 15+0.05 th 1 |
space try inspired by Lyudmil |
| 15-02-15 |
Roc |
MoreDblAttacks |
diff
|
LLR: -3.20 (-2.94,2.94) [0.00,4.00]
Total: 38342 W: 7572 L: 7605 D: 23165
|
sprt
@ 15+0.05 th 1 |
Back to original S(16, 0), Verifying additional binds on c6, f6 and b7 g7 since SPSA tuning showed singularities on those squares, |