movei need testers to find a better personality

[Uri Blass]

It seems that progress 10 10 10 is better than the default version
but not enough games were played and it can be good if other testers also test it and play progress 10 10 10 and default against the same opponents.

progress 10 10 10 is now number 9 out of the free engines that were tested

http://www.computerchess.org.uk/ccrl/40 ... ons_only=1

Here is a comparison between movei versions

http://www.computerchess.org.uk/ccrl/40 ... _length=30

I did not play enough games to know if it is better based on my own results.

It is probably possible to get more rating by changing more parameters.
Note that some parameters are set to 0 that mean that they are not used not because they do not help but because I did not have time to test them.

extreme_pawn_array 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
rook_factor 0
a_weaker 0
practically mean that they are not used

note that I think that special_bonus_endgame 0 together with
something like
extreme_pawn_array 1 3 6 10 15 21 28 36 45 55 66 78 91 105 120 136 153 171 190 210 231 253 276 300 325

May give better results but I did not test it.

practically 240 in extreme_pawn_array mean 0.1 pawn bonus for having extreme passed pawn.

movei is using 0.1 pawns bonus for extreme pawn passed pawn only when the stage is the endgame and I believe that it should be better to have bonus that is going more slowly from the opening to the endgame.

I did not test it and decided to leave the default parameter.

rook_factor means better value for the rook in the endgame.
I did not test it and I do not know if it helps.

a_weaker mean that movei evaluate the a pawn as relatively weaker in the opening and relatively stronger in the endgame.
I did not test it so default value is 0.


I suspect that the difference between numbers in old_mobility_table should be reduced but I did not test it

The reason is that I already have some code that is similiar to fruit about mobility so using the number of legal moves for mobility is using a similiar idea twice.

default numbers are:

old_mobility_table -92 -86 -80 -74 -68 -62 -56 -50 -44 -38 -32 -26 -20 -14 -8 -2 4 10 15 20 25 30 34 38 42 46 49 52 55 58 60 62 64 66 68 70 72 74 76 78 80 82 84 86 88 90 92 94 96 98 100 102 104 106 108 110 112 114 116 118 120 122 124 126 128 130 132 134 136 138 140 142 144 146 148 150 152 154 156 158 160 162 164 166 168 170 172 174 176 178 180 182 184 186 188 190 192 194 196 198 200 202 204 206 208 210 212 214 216 218 220 222 224 226 228 230 232 234 236 238 240 242 244 246 248 250 252 254 256 258 260 262 264 266 268 270 272 274 276 278 280 282 284 286 288 290 292 294 296 298 300 302 304 306 308 310 312 314 316 318 320 322 324 326 328 330 332 334 336 338 340 342 344 346 348 350 352 354 356 358 360 362 364 366 368 370 372 374 376 378 380 382 384 386 388 390 392 394 396 398 400 402 404 406 408 410 412 414 416 418 420 422 424 426 428 430 432 434 436 438 440 442 444 446 448 450 452 454 456 458 460 462 464 466 468 470 472 474 476 478 480 483 484 486 488 490 492 494 496 498 500 502 504 506 508

practically only difference between numbers are important so
you can reduce it by reducing the second number by 1 reducing the third number by 2 and continue it this way(or even reducing the first number by 2 the second number by 4 and continue in this way).

Alternatively it is possible to increase
mobilitydivider but in this case the part in movei that is responsible to evaluate bishop mobility or knight mobility as more important than rook mobility and also more important than queen mobility(like fruit) is going to be reduced (very big numbers like mobilitydivider 16384 mean that it is practically not going to be used).

I plan to test search ideas in the near future and I have not time to test changing evaluation weights so it may be good if I find testers who help me to test evaluation changes.

You only need to change movei_changes438.ini for changing parameters(less_reduce 1 should always be there when you can add lines like
rook_factor 10 ).


Uri

[Uri Blass]

Here is a possible personality for movei that may be better or not better than the default(I did not test it in games)

less_reduce 1
old_mobility_table -92 -86 -80 -75 -70 -65 -60 -55 -50 -45 -40 -35 -30 -26 -22 -18 -14 -11 -8 -5 -2 0 2 4 6 8 9 10 11 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12 12

I simply reduced the mobility score by that personality

This personality cause movei not to blunder after
1.Nc3 d5 2.Nf3 when the default version simply blunders by 2...e5

Another example is that this personality does not blunder by long castling
as it did in the following game against SpiderChess.

long castling simply improved the mobility and movei considered the mobility as more important than king safety.

[Event "5th Amateur D2"]
[Site "Howick"]
[Date "2007.08.31"]
[Round "1.3"]
[White "Movei 0.08.438"]
[Black "SpiderChess 070603"]
[Result "0-1"]
[ECO "E15"]
[Annotator "-0.14;0.11"]
[PlyCount "54"]
[EventDate "2007.08.26"]
[EventType "tourn"]
[Source "Banks"]
[TimeControl "40/4080:40/4080:40/4080"]

{AMD Athlon(tm) MP 1666 MHz W=13.7 ply; 250kN/s; HS-10moves(2).ctg B=13.9 ply;
515kN/s; HS-10moves(2).ctg} 1. c4 {-300.02/0 0} Nf6 {-300.02/0 0} 2. d4 {
-300.02/0 0} e6 {-300.02/0 0} 3. Nf3 {-300.02/0 0} b6 {-300.02/0 0} 4. g3 {
-300.02/0 0} Ba6 {-300.02/0 0} 5. b3 {-300.02/0 0} Bb4+ {-300.02/0 0} 6. Bd2 {
-300.02/0 0} Be7 {-300.02/0 0} 7. Bg2 {-300.02/0 0} c6 {-300.02/0 0} 8. Bc3 {
-300.02/0 0} d5 {-300.02/0 0} 9. Ne5 {-300.02/0 0} Nfd7 {-300.02/0 0} 10. Nxd7
{-300.02/0 0} Nxd7 {-300.02/0 0} 11. Nd2 {-0.14/14 126} O-O {0.11/14 137} 12.
e4 {(0-0) -0.03/14 188} dxc4 {(Rc8) 0.35/12 87} 13. Nxc4 {0.04/14 72} c5 {
(b5) 0.43/13 129} 14. e5 {0.17/14 107} cxd4 {0.29/13 87} 15. Qxd4 {-0.30/14 125
} Rc8 {0.08/14 195} 16. O-O-O {(0-0) -0.35/13 143} Bxc4 {(Rc7) -0.67/13 174}
17. bxc4 {-0.74/14 133} Qc7 {-0.47/13 88} 18. Kb1 {-0.74/13 177} Rfd8 {
-0.47/14 133} 19. Qe4 {-0.75/12 74} b5 {-0.56/14 127} 20. Rd4 {
(f4) -0.96/12 226} Nb6 {-0.96/13 146} 21. Qb7 {(Rxd8+) -1.37/12 88} Qc5 {
-1.48/13 91} 22. Rxd8+ {-2.04/13 165} Rxd8 {(Bxd8) -2.17/13 426} 23. Qxa7 {
(Rc1) -2.31/13 149} Rd7 {(b4) -2.11/13 105} 24. Qb8+ {-3.38/14 294} Bf8 {
-2.70/14 166} 25. Rc1 {-3.38/13 128} Nxc4 {-4.46/13 141} 26. f4 {
(Be4) -4.96/12 129} Qe3 {(Na3+) -8.27/13 77} 27. Bb4 {-7.82/12 55} Qd3+ {
-12.65/14 142} 0-1


Uri

[IgorGorelikov]

Hi Uri,
I have here in the ini file:

progress 0 0 0 0
(ie 4 zeros)

Should I change it to
10 10 10 0
or to
10 10 10 10
(to test a new personality)
?

Regards,
Igor

[Uri Blass]

Hi Uri,
I have here in the ini file:

progress 0 0 0 0
(ie 4 zeros)

Should I change it to
10 10 10 0
or to
10 10 10 10
(to test a new personality)
?

Regards,
Igor

The 4th number has practically no influence so 10 10 10 0 and 10 10 10 10 are the same.

Uri

[Dann Corbit]

I have a suggestion:
0. Describe what each parameter means
1. Describe what you think is the minimum sensible value
2. Describe what you think is the maximum sensible value
3. Describe the current default value

Armed with that information, the users will know what is sensible to try.