Discussion of chess software programming and technical issues.
Moderators: hgm , Rebel , chrisw
Evert
Posts: 2929Joined: Sat Jan 22, 2011 12:42 amLocation: NL
Post by Evert Fri May 01, 2015 10:29 pm
wgarvin wrote:
Google Translate made me laugh out loud with this one:
The main components of virtually any protein-free chess are searching and assessment of the position .
But still, its extremely readable. Its amazing how good automated translation is these days.
I have found it to be very hit-and-miss. See here for a good laugh:
http://www.talkchess.com/forum/viewtopic.php?t=56128
I'm sure it depends a lot on how much effort has gone into translating between particular languages, and the text in question.
hgm
Posts: 27787Joined: Fri Mar 10, 2006 10:06 amLocation: AmsterdamFull name: H G Muller
Post by hgm Fri May 01, 2015 11:11 pm
Dann Corbit wrote: 1-0 is the damage if you are white
If a game is fully solved, then indeed, there is no place for a heuristic evaluation. Mate solvers have no use for piece values.
But for an efficient search (rather than a 32-men tablebase), you would still need a heuristic evaluation to allow the engine to focus on the 'promising' branches through NMP or LMR, and for iterative deepening to make sense. There is no hope at all that you could solve chess through a depth-first fixed-depth search without iterative deepening.
Ajedrecista
Posts: 1966Joined: Wed Jul 13, 2011 9:04 pmLocation: Madrid, Spain.
Post by Ajedrecista Sat May 02, 2015 7:01 pm
Hello:
I managed to compile Vladimir's tool and I tested it through 234 PGNs of
this site . There are some extreme cases as you will notice:
Code: Select all
www.pgnmentor.com/files.html
10,000 iterations for each player with pgnlearn.exe by Vladimir Medvedev.
Pawn = 100 centipawns for each player (234 players in total).
Each piece value has been rounded to the nearest centipawn.
PLAYER Games N B R Q
Adams (2767 games) 214 226 386 798
Akobian ( 961 games) 220 235 339 734
Akopian (1734 games) 202 223 320 783
Alburt ( 776 games) 247 323 435 857
Alekhine (1661 games) 214 272 391 710
Alekseev (1239 games) 273 300 407 818
Almasi (1739 games) 220 234 383 754
Anand (2917 games) 216 220 355 762
Anderssen ( 681 games) 286 289 531 1014
Andersson (2641 games) 189 208 307 612
Andreikin ( 907 games) 151 182 282 584
Aronian (1887 games) 255 334 441 866
Ashley ( 414 games) 255 256 381 729
Averbakh ( 885 games) 157 194 254 463
Azmaiparashvili (1318 games) 280 296 467 926
Bacrot (2175 games) 190 229 387 857
Bareev (1844 games) 236 291 432 820
Becerra Rivero ( 967 games) 212 208 365 701
Beliavsky (2920 games) 239 295 400 855
Benjamin (1628 games) 189 225 354 714
Benko (1376 games) 201 238 331 675
Berliner ( 64 games) 230 186 407 1227
Bernstein ( 261 games) 199 247 347 746
Bird ( 353 games) 212 285 416 786
Bisguier (1186 games) 204 243 348 644
Blackburne ( 738 games) 213 234 332 740
Blatny (1751 games) 298 333 446 971
Bogoljubow ( 973 games) 263 273 388 858
Boleslavsky ( 651 games) 235 311 503 1005
Bologan (2363 games) 256 291 474 901
Botvinnik ( 891 games) 260 305 410 956
Bronstein (1930 games) 229 288 403 855
Browne (1620 games) 254 264 403 874
Bruzón ( 690 games) 204 220 371 625
Bu Xiangzhi (1307 games) 242 272 368 739
Byrne (1013 games) 215 256 373 656
Capablanca ( 597 games) 254 257 372 702
Carlsen (1730 games) 213 243 352 786
Caruana (1305 games) 282 305 452 974
Chiburdanidze (1323 games) 204 231 366 657
Chigorin ( 688 games) 219 300 382 770
Christiansen (1592 games) 246 268 410 890
DeFirmian (1683 games) 219 254 355 774
De La Bourdonnais ( 101 games) 307 80 433 1155
Denker ( 404 games) 143 232 337 528
Ding Liren ( 481 games) 206 236 344 678
Domínguez Pérez ( 750 games) 312 336 490 894
Dreev (2742 games) 224 267 393 861
Dzindzichashvili ( 780 games) 250 276 325 692
Ehlvest (2449 games) 195 224 336 702
Eljanov (1265 games) 196 257 367 665
Euwe (1122 games) 183 213 341 636
Evans ( 685 games) 227 254 437 915
Fedorowicz (1357 games) 257 310 449 834
Fine ( 305 games) 215 291 319 526
Finegold ( 835 games) 258 283 355 797
Fischer ( 827 games) 240 273 394 805
Fishbein ( 448 games) 177 250 357 639
Flohr ( 986 games) 198 219 316 640
Gaprindashvili (1075 games) 218 228 448 833
Gashimov ( 953 games) 235 257 380 927
Gelfand (2748 games) 247 296 431 912
Geller (2198 games) 199 238 333 691
Georgiev (2362 games) 260 280 373 742
Giri ( 912 games) 297 358 438 808
Gligoric (2898 games) 265 335 452 932
Goldin ( 898 games) 268 308 461 880
Granda Zúñiga (1510 games) 225 296 422 737
Grischuk (2190 games) 237 270 376 759
Gulko (1874 games) 135 189 257 652
Gunsberg ( 319 games) 238 300 480 1085
Gurevich, Dmitry (1535 games) 239 273 386 833
Gurevich, Mikhail (2372 games) 199 243 340 800
Harikrishna (1431 games) 301 309 478 798
Hort (3070 games) 193 235 360 736
Horwitz ( 113 games) 558 592 967 1810
Hou Yifan (1190 games) 260 271 378 766
Huebner (2087 games) 186 214 350 717
Ibragimov (1435 games) 291 327 488 1019
Illescas Córdoba (1389 games) 272 320 404 924
Ivanchuk (3554 games) 234 253 391 801
Ivanov, Alexander (1218 games) 229 246 369 878
Ivanov, Igor ( 784 games) 311 307 404 831
Ivkov (2505 games) 250 274 369 719
Jakovenko (1335 games) 269 326 449 740
Janowski ( 769 games) 223 247 369 770
Jobava (1161 games) 238 301 372 807
Jussupow (1920 games) 193 215 331 709
Kaidanov (1085 games) 245 295 444 877
Kamsky (1744 games) 257 264 408 787
Karjakin (1793 games) 248 289 419 842
Karpov (3351 games) 275 294 426 945
Kasimdzhanov (1383 games) 202 207 319 604
Kasparov (2027 games) 159 228 334 639
Kavalek (1328 games) 268 311 438 886
Keres (1571 games) 162 205 293 575
Khalifman (2127 games) 178 227 373 799
Kholmov (1900 games) 209 227 317 712
Koneru (1196 games) 233 301 466 888
Korchnoi (4565 games) 240 272 417 806
Korobov (1021 games) 242 284 378 765
Kosteniuk (1477 games) 282 320 522 1087
Kotov ( 641 games) 192 220 399 857
Kramnik (2487 games) 213 262 358 722
Krasenkow (2212 games) 214 243 339 646
Krush (1407 games) 233 251 376 758
Kudrin (1186 games) 293 324 446 1080
Lahno ( 919 games) 265 263 388 688
Larsen (2383 games) 232 254 307 603
Lasker ( 900 games) 231 260 486 744
Lautier (1796 games) 217 234 359 722
Le Quang Liem ( 996 games) 265 303 439 842
Leko (2202 games) 234 242 412 838
Levenfish ( 354 games) 129 175 388 747
Li Chao ( 670 games) 290 293 542 988
Lilienthal ( 649 games) 268 303 443 823
Ljubojevic (1944 games) 225 241 357 772
Lputian (1552 games) 202 234 374 773
MacKenzie ( 198 games) 737 828 1187 2728
Malakhov (1422 games) 213 253 369 630
Mamedyarov (1707 games) 167 180 264 675
Maroczy ( 756 games) 177 242 390 821
Marshall (1027 games) 237 255 390 798
McDonnell ( 106 games) 288 141 660 1338
McShane (1134 games) 306 299 446 913
Mecking ( 569 games) 237 268 481 934
Mikenas ( 497 games) 218 281 404 636
Miles (2614 games) 254 292 457 982
Milov (1342 games) 260 321 445 902
Morozevich (1905 games) 282 316 442 941
Morphy ( 211 games) -228 326 77 1557
Motylev (1377 games) 260 247 374 777
Movsesian (2202 games) 237 263 405 846
Najdorf (1604 games) 213 269 395 732
Najer (1450 games) 184 188 327 662
Nakamura (1623 games) 290 296 425 838
Navara (1605 games) 284 285 436 813
Negi ( 996 games) 249 292 457 866
Nepomniachtchi (1238 games) 203 243 368 819
Ni Hua (1279 games) 223 232 413 810
Nielsen (1978 games) 218 268 385 759
Nikolic (2224 games) 200 244 357 737
Nimzowitsch ( 512 games) 274 262 351 804
Nisipeanu (1545 games) 225 224 431 1058
Novikov (1318 games) 283 314 480 1035
Nunn (1655 games) 175 229 343 645
Olafsson ( 916 games) 224 254 361 851
Oll (1034 games) 197 236 296 578
Onischuk (1459 games) 233 269 409 874
Pachman (1462 games) 200 236 329 631
Paehtz (1626 games) 268 305 437 951
Panno (1873 games) 225 265 390 744
Paulsen ( 322 games) 210 279 436 837
Petrosian (1893 games) 187 197 294 571
Philidor ( 6 games) 3058 3034 -837 0
Pillsbury ( 388 games) 218 233 327 800
Pilnik ( 899 games) 235 295 419 783
Polgar, Judit (1825 games) 210 271 420 1008
Polgar, Sofia ( 741 games) 187 215 387 714
Polgar, Zsuzsa ( 909 games) 199 227 349 856
Polugaevsky (1890 games) 197 244 354 765
Ponomariov (1640 games) 240 264 443 828
Portisch (3030 games) 202 247 382 821
Psakhis (1963 games) 205 201 353 711
Quinteros (1136 games) 159 220 363 693
Radjabov (1551 games) 246 272 434 786
Reshevsky (1267 games) 201 257 440 843
Reti ( 646 games) 238 274 433 935
Ribli (2112 games) 228 276 417 929
Rohde ( 445 games) 280 277 506 915
Rubinstein ( 797 games) 230 291 336 699
Rublevsky (1543 games) 307 312 441 1022
Saemisch ( 551 games) 244 287 432 947
Sakaev (1425 games) 278 339 497 937
Salov ( 766 games) 272 290 457 1003
Sasikiran (1311 games) 203 249 368 712
Schlechter ( 739 games) 182 239 360 704
Seirawan (1483 games) 221 247 380 795
Serper ( 848 games) 221 233 352 782
Shabalov (2000 games) 294 354 505 1008
Shamkovich ( 977 games) 187 213 282 675
Shirov (3193 games) 208 247 395 817
Short (2689 games) 229 262 395 825
Shulman (1191 games) 266 262 439 874
Smirin (1988 games) 183 256 378 719
Smyslov (2627 games) 210 247 377 789
Sokolov (2742 games) 203 267 364 777
Soltis ( 370 games) 175 186 277 811
Spassky (2231 games) 142 176 276 604
Speelman (2122 games) 173 211 350 653
Spielmann (1057 games) 216 308 374 809
Stahlberg ( 964 games) 152 187 291 545
Staunton ( 284 games) 285 331 643 1172
Stefanova (1924 games) 288 319 452 866
Stein ( 699 games) 246 329 407 871
Steinitz ( 590 games) 216 315 399 795
Suetin (1284 games) 205 264 350 653
Sutovsky (1095 games) 262 233 362 773
Svidler (2570 games) 198 245 353 756
Szabo (1777 games) 191 259 375 791
Taimanov (2117 games) 232 285 441 821
Tal (2431 games) 148 171 261 572
Tarrasch ( 704 games) 211 235 407 726
Tartakower (1290 games) 206 252 328 713
Teichmann ( 536 games) 275 327 439 810
Timman (3402 games) 242 266 402 861
Tiviakov (2507 games) 211 233 381 767
Tkachiev (1529 games) 191 240 359 767
Tomashevsky ( 882 games) 277 299 473 929
Topalov (2085 games) 178 215 308 645
Torre Repetto ( 175 games) 329 296 513 931
Uhlmann (2537 games) 223 253 342 738
Unzicker (1265 games) 245 309 516 1019
Vachier-Lagrave (1332 games) 213 225 315 485
Vaganian (2255 games) 204 234 375 839
Vallejo Pons (1763 games) 220 249 357 833
Van Wely (3233 games) 232 252 378 758
Vitiugov (1009 games) 245 238 344 766
Volokitin (1050 games) 221 252 451 854
Waitzkin ( 385 games) 223 329 427 707
Wang Hao (1047 games) 200 251 343 720
Wang Yue (1195 games) 210 218 326 648
Winawer ( 241 games) 302 297 529 1026
Wojtaszek (1318 games) 203 281 348 648
Wojtkiewicz (1564 games) 268 325 489 907
Wolff ( 614 games) 293 278 437 949
Xie Jun ( 701 games) 332 365 501 980
Ye Jiangchuan ( 655 games) 243 271 367 875
Yermolinsky (1407 games) 169 182 353 541
Yu Yangyi ( 619 games) 215 239 424 796
Yudasin (1426 games) 187 210 339 613
Zhu Chen (1041 games) 266 286 411 878
Zukertort ( 265 games) 80 141 295 402
Zvjaginsev (1441 games) 239 263 359 698
min.(N) = 80 (Zukertort); max.(N) = 737 (McKenzie).
min.(B) = 80 (De La Bourdonnais); max.(B) = 828 (Mackenzie).
min.(R) = 77 (Morphy); max.(R) = 1187 (Mackenzie).
min.(Q) = 402 (Zukertort); max.(Q) = 2728 (Mackenzie).
N < P for:
Zukertort.
B < P for:
De La Bourdonnais.
B < N for:
Becerra Rivero.
Berliner.
Ivanov, Igor.
Lahno.
McShane.
Motylev.
Nimzowitsch.
Nisipeanu.
Psakhis.
Rohde.
Shulman.
Sutovsky.
Torre Repetto.
Vitiugov.
Winawer.
Wolff.
R < P for:
Morphy.
Some curiosities/singularities:
Knight (MacKenzie) > Knight (Horwitz) > Queen (Averbakh) > Queen (Zukertort).
Knight (Horwitz) ~ 7 Bishops (De La Boudonnais) ~ 7 Knights (Zukertort).
Bishop (Mackenzie) > 10 Bishops (De La Bourdonnais) ~ 10 Knights (Zukertort).
Knight (Morphy) < 0 (not included in min.(N)).
Queen (McKenzie) > 35 Rooks (Morphy).
Philidor's values have not been included in min.(j) and max.(j) (too few games? Material odds (it seems that there are not material odds in those six games)?).
There are not all the current top players but it is enough to show the possible weaknesses of this approach. It would be curious to see the values of each one of us... mine are surely joking.
I do not know anything about C++, but it looks like the number of iterations can be changed at
regression.cpp file before compiling, exactly here:
Code: Select all
for (iter = 0; iter < 10000; ++iter)
I hope no typos in this long post.
Regards from Spain.
Ajedrecista.
Dann Corbit
Posts: 12537Joined: Wed Mar 08, 2006 8:57 pmLocation: Redmond, WA USA
Post by Dann Corbit Sat May 02, 2015 8:36 pm
Berliner and De La Bourdonnais both love the queen and hate the bishop.
Dann Corbit
Posts: 12537Joined: Wed Mar 08, 2006 8:57 pmLocation: Redmond, WA USA
Post by Dann Corbit Sat May 02, 2015 8:38 pm
Philidor really did believe that "Pawns are the soul of chess!"
Gerd Isenberg
Posts: 2250Joined: Wed Mar 08, 2006 8:47 pmLocation: Hattingen, Germany
Post by Gerd Isenberg Sat May 02, 2015 11:17 pm
WinPooh wrote: My article on a polular Russian IT portal:
http://habrahabr.ru/post/254753/
It is in Russian, but I believe the text is quite understandable even without translation.
Courtesy of Vladimir, I made a CPW page trying a translation with the help of google translate, improvements and suggestions welcome.
https://chessprogramming.wikispaces.com ... n+Analysis
Cheers,
Gerd
SuneF
Posts: 127Joined: Thu Sep 17, 2009 11:19 am
Post by SuneF Sun May 03, 2015 11:37 am
Dann Corbit wrote: These weights are derived from correspondence chess:
How are these derived from correspondence chess?
Granted the pawn is particularly difficult to estimate an average value of, since it could be worth anything from 0 to a queen. IMO though, in practice, a minor piece is worth more than 3 pawns "on average" based on the fact you don't see many "piece for 3 pawns" exchanges in actual games. I would consider such a trade to be more of a sacrifice.
Vinvin
Posts: 5228Joined: Thu Mar 09, 2006 9:40 amFull name: Vincent Lejeune
Post by Vinvin Sun May 03, 2015 12:59 pm
Ajedrecista wrote: I managed to compile Vladimir's tool and I tested it through 234 PGNs of
this site . There are some extreme cases as you will notice:
Code: Select all
www.pgnmentor.com/files.html
10,000 iterations for each player with pgnlearn.exe by Vladimir Medvedev.
Pawn = 100 centipawns for each player (234 players in total).
Each piece value has been rounded to the nearest centipawn.[/quote]
As I see, it's more the value of the pawn who changed more over time (pieces values between them don't change a lot). In my sense, it would be better to set the value of the knight to 300 to have a more clear view of the evolution over time.
Could you do that, please ?
Thanks,
Vincent
Vinvin
Posts: 5228Joined: Thu Mar 09, 2006 9:40 amFull name: Vincent Lejeune
Post by Vinvin Sun May 03, 2015 1:12 pm
Nice page, Gerd !
I recompute the table at the middle at the page with knight = 300 instead of pawn = 100
Code: Select all
Anand Carlsen Andersen
Pawn 139 141 105
Knight 300 300 300
Bishop 319 342 303
Rook 493 496 557
Queen 1058 1107 1063
King ∞ ∞ ∞
Gerd Isenberg
Posts: 2250Joined: Wed Mar 08, 2006 8:47 pmLocation: Hattingen, Germany
Post by Gerd Isenberg Sun May 03, 2015 2:36 pm
Vinvin wrote:
I recompute the table at the middle at the page with knight = 300 instead of pawn = 100
Code: Select all
Anand Carlsen Andersen
Pawn 139 141 105
Knight 300 300 300
Bishop 319 342 303
Rook 493 496 557
Queen 1058 1107 1063
King ∞ ∞ ∞
Looks much better since pawns are so volatile ...
„Tout soldat français porte dans sa giberne le bâton de maréchal de France.“
see also
http://www.talkchess.com/forum/viewtopi ... 43&t=45512
