Im just testing different books for Crafty. I am using the book.bin and bookc.bin from Bob Hyatt's site. Also I've build my own book based on a collection of GM games. In my first tests I found out that Crafty was performing slightly better playing _without_(!) the computer book. To test this I played about 100 games at fast time controls. All games was played on my Dual Cel 2*520 MHz. Ponder was on. Crafty.rc included "book random 1", "book width 2" and "computer" allways. All 3+4 and some 5-men tablebases were in use. Hashsize was 24+8 MB.
For my personal amusement and to test the different books against strong opponents I played a little tournament after this. Time control was game in 30 minutes for each engine. Computer was the same, each engine got about 32 MB hash. Crafty_hb was using the book.bin from Bob Hyatt's site, Crafty_hbc was using the bookc.bin also, Crafty_nb was using my book. Amy07_4 is the executable dated 000327 from Dann Corbit's site. Comet_B18-2 was used because of the popup window B20 is opening at the beginning of a match. This tournament was controlled by a *.bat file and I wasn't around all the time when Comet started to close that window. Here are the results:
No. Name 1 2 3 4 5 6 7 8 9 10 Score
-------------------------------------------------------
1 Amy07_4 xx =0 11 0= == == 11 11 10 00 10.0
2 AnMon506 =1 xx 0= 0= 10 == 11 00 =1 11 10.0
3 Comet_B18-2 00 1= xx 00 0= 00 00 00 =0 10 3.5
4 Crafty1710_hb 1= 1= 11 xx =1 == 11 =0 10 11 13.0
5 Crafty1710_hbc == 01 1= =0 xx == =1 0= 01 =0 8.5
6 Crafty1710_nb == == 11 == == xx 11 10 01 1= 11.5
7 Gromit_30 00 00 11 00 =0 00 xx 01 11 0= 6.0
8 LG2000V25 00 11 11 =1 1= 01 10 xx 0= 00 9.5
9 Phalanx22 01 =0 =1 01 10 10 00 1= xx 00 7.5
10 SOS_991103 11 00 01 00 =1 0= 1= 11 11 xx 10.5
-----------------------------------------------------------------
No. Name Win Draw Loss Unf. Score Games % SOP
-----------------------------------------------------------------
1 Crafty1710_hb +10 =6 -2 *0 13.0 18 72.2% 154.0
2 Crafty1710_nb +7 =9 -2 *0 11.5 18 63.9% 157.0
3 SOS_991103 +9 =3 -6 *0 10.5 18 58.3% 159.0
4 Amy07_4 +7 =6 -5 *0 10.0 18 55.6% 160.0
4 AnMon506 +7 =6 -5 *0 10.0 18 55.6% 160.0
6 LG2000V25 +8 =3 -7 *0 9.5 18 52.8% 161.0
7 Crafty1710_hbc +4 =9 -5 *0 8.5 18 47.2% 163.0
8 Phalanx22 +6 =3 -9 *0 7.5 18 41.7% 165.0
9 Gromit_30 +5 =2 -11 *0 6.0 18 33.3% 168.0
10 Comet_B18-2 +2 =3 -13 *0 3.5 18 19.4% 173.0
-----------------------------------------------------------------
Start Elo: 2400
Program Score % Av.Op. Elo + - Draws
1 Crafty1710_hb : 13.0/ 18 72.2 2386 2552 145 149 33.3 %
2 Crafty1710_nb : 11.5/ 18 63.9 2392 2491 157 104 50.0 %
3 SOS_991103 : 10.5/ 18 58.3 2396 2455 167 165 16.7 %
4 Amy07_4 : 10.0/ 18 55.6 2398 2437 172 129 33.3 %
4 AnMon506 : 10.0/ 18 55.6 2398 2437 172 129 33.3 %
6 LG2000V25 : 9.5/ 18 52.8 2400 2420 178 156 16.7 %
7 Crafty1710_hbc : 8.5/ 18 47.2 2404 2385 100 178 50.0 %
8 Phalanx22 : 7.5/ 18 41.7 2408 2350 165 167 16.7 %
9 Gromit_30 : 6.0/ 18 33.3 2414 2294 198 153 11.1 %
10 Comet_B18-2 : 3.5/ 18 19.4 2427 2180 224 135 16.7 %
Total Games: 90
White Wins: 34 (37.8%)
Black Wins: 31 (34.4%)
Draws: 25 (27.8%)
Unfinished: 0 (0.0%)
-----------------------------------------------------------------
Amy was really surprising me. It seems to be much stronger than I expected. I'm going to play another tournament with same settings but some changes: Comet_B18 will be replaced by DerBringer. B20 is stronger but I cannot use it when I'm not around. Amy will be updated by the new version which is using some assembly routines.
BTW: does anybody know what is necessary to build a bookc.bin for Crafty properly? "Bookc create ..." seems to be not enough. I have allready downloaded the brainy.zip file from Dann's site but I don't know what to do with it
.To download the games please visit my games archive.
Best wishes
Volker
my games archive
Die Erläuterungen sind vielleicht hilfreich, ist ja mehr ein Test. In ca. 100 Stunden kommt dann Teil 2.