| 19-08-15 |
Voy |
anmpT |
diff
|
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 37084 W: 8191 L: 8172 D: 20721
|
sprt
@ 10+0.1 th 1 |
stc |
| 19-08-15 |
MJZ |
TT-Save2 |
diff
|
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 10712 W: 2329 L: 2441 D: 5942
|
sprt
@ 10+0.1 th 1 |
Set ttPv=true if we have a close upper and lower bounds. depth > 6, abs(ttValue-beta) < 50 verification |
| 19-08-15 |
sg |
tuned_king_danger |
diff
|
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 31319 W: 7043 L: 7051 D: 17225
|
sprt
@ 10+0.1 th 1 |
As expected the tuned values seems neutral. Try now to double the effects. |
| 19-08-15 |
Elb |
nwp |
diff
|
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 21078 W: 4631 L: 4691 D: 11756
|
sprt
@ 10+0.1 th 1 |
Give bonus for knight attacking weak pawns |
| 19-08-15 |
31m |
PawnlessFlank_dist |
diff
|
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 30570 W: 6796 L: 6809 D: 16965
|
sprt
@ 10+0.1 th 1 |
Broader effect (3 ranks, rather than 4). |
| 19-08-15 |
sg |
tuned_king_danger |
diff
|
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 24660 W: 5539 L: 5581 D: 13540
|
sprt
@ 10+0.1 th 1 |
Test the fine tuned king danger values. As expected the move not much but 2 parameters are clearly and 3 parameters slightly outside of the variance given by the RandomWalk parameter so not everything seems random noise. |
| 19-08-15 |
MJZ |
TT-Save2 |
diff
|
LLR: -2.94 (-2.94,2.94) [0.50,4.50]
Total: 15167 W: 3315 L: 3404 D: 8448
|
sprt
@ 10+0.1 th 1 |
Set ttPv=true if we have a close upper and lower bounds. depth > 6 |
| 19-08-15 |
xor |
qstatT |
diff
|
LLR: -2.95 (-2.94,2.94) [0.00,3.50]
Total: 13144 W: 2108 L: 2247 D: 8789
|
sprt
@ 60+0.6 th 1 |
LTC: take 1. |
| 19-08-15 |
MJZ |
TT-Save2 |
diff
|
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 6393 W: 1369 L: 1502 D: 3522
|
sprt
@ 10+0.1 th 1 |
Set ttPv=true if we have a close upper and lower bounds. |
| 19-08-15 |
MJZ |
TT-Save |
diff
|
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 2147 W: 394 L: 547 D: 1206
|
sprt
@ 10+0.1 th 1 |
Don't save into TT if TTdepth > depth. |
| 19-08-15 |
Voy |
nmpFt |
diff
|
LLR: -2.95 (-2.94,2.94) [0.00,3.50]
Total: 18371 W: 3005 L: 3129 D: 12237
|
sprt
@ 60+0.6 th 1 |
Curious how this yellow will fare at LTC, since nmp is quite finicky. Low TP |
| 19-08-14 |
sg |
tune_king_danger |
diff
|
47690/50000 iterations
99908/100000 games played
|
100000
@ 20+0.2 th 1 |
Try a fine tuning of king danger terms hence low c(=5) but higher R(0.0025) for larger movement. I introduce an additional parameter RandomWalk which has no effect on the engine. Its purpose is to get a visual tool for observing the inherent variance of of the tuning process. Further i udes nodestime=800 instead of nodestime=600 because this values is outdated. Its based on the half of the fishtest base NPS. At the time nodestimes were i ntroduced this was 1200KN/s (=1200 N/ms => half is nodestime=600). Now the framework use 1600KN/S so nodestime=800. Else the tuning would run with a effectivly TC of 7.5+0.075.
Based on my passed unsafe checks patch. |
| 19-08-15 |
31m |
PawnlessFlank_dist |
diff
|
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 27058 W: 6103 L: 6132 D: 14823
|
sprt
@ 10+0.1 th 1 |
Apply the PawnlessFlank penalty even if there is only one pawn, if it is very far (more than 4 ranks) away. |
| 19-08-15 |
xor |
qstatT |
diff
|
LLR: 2.96 (-2.94,2.94) [0.50,4.50]
Total: 22054 W: 5050 L: 4775 D: 12229
|
sprt
@ 10+0.1 th 1 |
take 1. |
| 19-08-15 |
31m |
PawnlessFlank_dist |
diff
|
LLR: -2.94 (-2.94,2.94) [0.50,4.50]
Total: 12949 W: 2823 L: 2923 D: 7203
|
sprt
@ 10+0.1 th 1 |
A more general version, covering cases where there are multiple pawns but the furthest ones are far away. |
| 19-08-14 |
sni |
nmp_bound |
diff
|
ELO: -13.86 +-4.0 (95%) LOS: 0.0%
Total: 9959 W: 1508 L: 1905 D: 6546
|
10000
@ 60+0.6 th 1 |
Estimate the cost of bounding null move pruning to depth 12 |
| 19-08-15 |
31m |
TOP_blocked |
diff
|
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 52972 W: 11804 L: 11706 D: 29462
|
sprt
@ 10+0.1 th 1 |
S(0, 20) for both Minor and Rook cases. |
| 19-08-15 |
Viz |
goodCaptCountLMR3 |
diff
|
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 14660 W: 3213 L: 3305 D: 8142
|
sprt
@ 10+0.1 th 1 |
Take 3 |
| 19-08-15 |
Viz |
goodCaptCountLMR2 |
diff
|
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 18662 W: 4114 L: 4186 D: 10362
|
sprt
@ 10+0.1 th 1 |
Take 2 |
| 19-08-14 |
Viz |
goodCaptCountLMR1 |
diff
|
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 88684 W: 19941 L: 19663 D: 49080
|
sprt
@ 10+0.1 th 1 |
Take 1 |
| 19-08-15 |
31m |
TOP_blocked |
diff
|
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 65125 W: 14448 L: 14290 D: 36387
|
sprt
@ 10+0.1 th 1 |
Double effect |
| 19-08-15 |
31m |
TOP_blocked |
diff
|
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 15387 W: 3378 L: 3466 D: 8543
|
sprt
@ 10+0.1 th 1 |
For the sake of curiosity, try a massive effect. S(0, 40) |
| 19-08-14 |
Oua |
EndgameSpeedUp |
diff
|
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 71860 W: 16048 L: 15927 D: 39885
|
sprt
@ 10+0.1 th 1 |
Slight speep up fetching the endgame table.
Replace calls to count(key) + operator[key] with a single call to find(key).
Replace the std::map with std::unordered_map which provide O(1) access, although the map has a really small number of objects. Maybe a std::vector or std::array using the endgame code as the index would improve this further a bit. |
| 19-08-14 |
Viz |
minorSafeCheckTweak3 |
diff
|
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 46601 W: 10329 L: 10263 D: 26009
|
sprt
@ 10+0.1 th 1 |
Take 2 |
| 19-08-14 |
31m |
TOP_blocked |
diff
|
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 37663 W: 8403 L: 8380 D: 20880
|
sprt
@ 10+0.1 th 1 |
Only pawns blocked by pawns, not blocked by pieces in general. |
| 19-08-14 |
Viz |
minorSafeCheckTweak1 |
diff
|
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 43502 W: 9628 L: 9577 D: 24297
|
sprt
@ 10+0.1 th 1 |
Take 1 |
| 19-08-14 |
sg |
unsafe_checks |
diff
|
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 36931 W: 8307 L: 8287 D: 20337
|
sprt
@ 10+0.1 th 1 |
Exclude unsafe checks at a blocked pawn. |
| 19-08-14 |
pro |
ps_knights8 |
diff
|
LLR: -2.95 (-2.94,2.94) [0.00,4.00]
Total: 47071 W: 10533 L: 10510 D: 26028
|
sprt
@ 10+0.1 th 1 |
updated knights mobility values. v2 |
| 19-08-14 |
Voy |
nmpFt |
diff
|
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 50812 W: 11259 L: 11172 D: 28381
|
sprt
@ 10+0.1 th 1 |
stc |
| 19-08-14 |
Viz |
separateBlockers2 |
diff
|
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 34624 W: 7653 L: 7646 D: 19325
|
sprt
@ 10+0.1 th 1 |
Take 1 |
| 19-08-14 |
sg |
passed_pawn2 |
diff
|
LLR: -2.94 (-2.94,2.94) [0.50,4.50]
Total: 28740 W: 6428 L: 6449 D: 15863
|
sprt
@ 10+0.1 th 1 |
Based on w_mg=6*r-16 with a compensation factor of 12/13. Use directly in score formula for finer resolution. |
| 19-08-14 |
sg |
unsafe_checks |
diff
|
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 21019 W: 4679 L: 4739 D: 11601
|
sprt
@ 10+0.1 th 1 |
Consider pawn blocked only if not attacked by opponent pawn. |
| 19-08-14 |
pro |
ps_knights8 |
diff
|
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 17603 W: 3868 L: 3963 D: 9772
|
sprt
@ 10+0.1 th 1 |
test the value of eg knight mobility. (eg = 0). |
| 19-08-14 |
sg |
unsafe_checks |
diff
|
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 15732 W: 3456 L: 3543 D: 8733
|
sprt
@ 10+0.1 th 1 |
Consider only if pawn is blocked by a pawn. |
| 19-08-14 |
Viz |
weakBlockers2 |
diff
|
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 13843 W: 3040 L: 3136 D: 7667
|
sprt
@ 10+0.1 th 1 |
Take 1 |
| 19-08-14 |
pro |
ps_mobility15 |
diff
|
LLR: -2.96 (-2.94,2.94) [-3.00,1.00]
Total: 2687 W: 565 L: 740 D: 1382
|
sprt
@ 10+0.1 th 1 |
simplify mobility = mg, mg/2. |
| 19-08-14 |
Viz |
unsafeQueenChecks3 |
diff
|
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 8937 W: 1913 L: 2033 D: 4991
|
sprt
@ 10+0.1 th 1 |
Take 3 |
| 19-08-14 |
pro |
ps_shelter48 |
diff
|
LLR: -2.95 (-2.94,2.94) [-3.00,1.00]
Total: 4901 W: 1029 L: 1205 D: 2667
|
sprt
@ 10+0.1 th 1 |
try to simplify shelter. |
| 19-08-14 |
Viz |
unsafeQueenChecks4 |
diff
|
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 8905 W: 1910 L: 2031 D: 4964
|
sprt
@ 10+0.1 th 1 |
Take 4 |
| 19-08-14 |
Viz |
unsafeQueenChecks1 |
diff
|
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 8121 W: 1759 L: 1884 D: 4478
|
sprt
@ 10+0.1 th 1 |
Take 1 |
| 19-08-14 |
sg |
passed_pawn |
diff
|
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 16847 W: 3680 L: 3761 D: 9406
|
sprt
@ 10+0.1 th 1 |
Use (k+2)*5/6 in endgame. |
| 19-08-14 |
Voy |
nmpFt1b |
diff
|
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 5836 W: 1252 L: 1388 D: 3196
|
sprt
@ 10+0.1 th 1 |
stc |
| 19-08-14 |
Voy |
nmpFt3 |
diff
|
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 25442 W: 5630 L: 5668 D: 14144
|
sprt
@ 10+0.1 th 1 |
Ver 3 |
| 19-08-14 |
sg |
passed_pawn2 |
diff
|
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 9998 W: 2175 L: 2290 D: 5533
|
sprt
@ 10+0.1 th 1 |
w_mg=6*r-16 and k+2 for endgame. |
| 19-08-14 |
Voy |
nmpFt2 |
diff
|
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 11418 W: 2468 L: 2576 D: 6374
|
sprt
@ 10+0.1 th 1 |
stc |
| 19-08-14 |
pro |
ps_knights8 |
diff
|
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 11494 W: 2552 L: 2672 D: 6270
|
sprt
@ 10+0.1 th 1 |
updated knights mobility values. |
| 19-08-14 |
sg |
passed_pawn |
diff
|
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 9021 W: 1995 L: 2115 D: 4911
|
sprt
@ 10+0.1 th 1 |
use k+3 for endgame. |
| 19-08-14 |
sg |
passed_pawn2 |
diff
|
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 20177 W: 4456 L: 4521 D: 11200
|
sprt
@ 10+0.1 th 1 |
w_mg=6*r-16 and k-1 for midgame. |
| 19-08-14 |
sg |
passed_pawn |
diff
|
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 38669 W: 8587 L: 8560 D: 21522
|
sprt
@ 10+0.1 th 1 |
use k+2 for endgame value. |
| 19-08-14 |
pro |
ps_bishopmobility1 |
diff
|
LLR: -1.25 (-2.94,2.94) [-3.00,1.00]
Total: 11353 W: 2507 L: 2600 D: 6246
|
sprt
@ 10+0.1 th 1 |
bishop mobility equation. For info only. Low tp. |