Discussion of anything and everything relating to chess playing software and machines.
Moderators: hgm , Rebel , chrisw
Laskos
Posts: 10948Joined: Wed Jul 26, 2006 10:21 pmFull name: Kai Laskos
Post by Laskos Wed Jan 10, 2018 1:13 pm
The main conclusions:
Komodo overperforms against other engines in disbalance Queen - Two Rooks, both with Queen and with Rooks. Stockfish and Andscacs underperform with the Queen. Houdini overperforms with the Queen.
1/ Regular opening file 2moves_v1.epd:
Code: Select all
Games Completed = 1500 of 1500 (Avg game length = 12.890 sec)
Settings = RR/0MB/100ms per move/M 700cp for 3 moves, D 70 moves/EPD:C:\LittleBlitzer\2moves_v1.epd(32000)
Time = 2888 sec elapsed, 0 sec remaining
1. Komodo 340.0/750 161-231-358 (L: m=36 t=0 i=0 a=195) (D: r=117 i=15 f=0 s=1 a=225) (tpm=109.1 d=15.50 nps=1473269)
2. SF 486.0/750 308-86-356 (L: m=6 t=0 i=0 a=80) (D: r=108 i=13 f=0 s=2 a=233) (tpm=109.6 d=17.24 nps=1763634)
3. Houdini 450.0/750 268-118-364 (L: m=14 t=0 i=0 a=104) (D: r=131 i=16 f=0 s=1 a=216) (tpm=108.6 d=14.44 nps=1974323)
4. Andscacs 224.0/750 52-354-344 (L: m=354 t=0 i=0 a=0) (D: r=68 i=6 f=0 s=2 a=268) (tpm=118.9 d=0.00 nps=1232327)
1500: +423 =711 -366
W B
51.9% 48.1%
Komodo: 47.2% 43.5%
SF: 66.0% 63.6%
Houdini: 62.9% 57.1%
And: 31.4% 28.3%2/ An artificial very balanced opening position of disbalance (from which I built a small 3-mover book). The pawn was removed because Houdini would not accept 3 rooks and 8 pawns:
[D]1nbqkbnr/pp1ppppp/8/8/8/8/PP1PPPPP/RNBRKBNR w KQk - 0 1
Code: Select all
Games Completed = 1500 of 1500 (Avg game length = 11.816 sec)
Settings = RR/0MB/100ms per move/M 700cp for 3 moves, D 70 moves/EPD:C:\LittleBlitzer\Queen_Rook_Book1.epd(135)
Time = 2665 sec elapsed, 0 sec remaining
1. Komodo 394.5/750 232-193-325 (L: m=26 t=0 i=0 a=167) (D: r=167 i=12 f=0 s=1 a=145) (tpm=108.7 d=15.89 nps=1526926)
2. SF 458.5/750 296-129-325 (L: m=10 t=0 i=0 a=119) (D: r=146 i=6 f=0 s=1 a=172) (tpm=109.5 d=17.44 nps=1797084)
3. Houdini 466.5/750 314-131-305 (L: m=13 t=0 i=0 a=118) (D: r=145 i=8 f=0 s=0 a=152) (tpm=107.4 d=15.08 nps=2033063)
4. And 180.5/750 48-437-265 (L: m=437 t=0 i=0 a=0) (D: r=90 i=8 f=0 s=0 a=167) (tpm=118.3 d=0.00 nps=1288886)
1500: +455 =610 -435
W B
50.6% 49.4%
Komodo: 52.0% 53.2%
SF: 62.8% 59.5%
Houdini: 60.6% 63.8%
And: 27.2% 21.0%3/ Over/Under performance:
Code: Select all
W B
Komodo: 6.1% 8.4%
SF: -1.9% -5.4%
Houdini: -1.0% 8.0%
And: -2.9% -8.6%
Multiply those numbers by 7, and you will get Elo over/under performance.
cdani
Posts: 2204Joined: Sat Jan 18, 2014 10:24 amLocation: Andorra
Post by cdani Wed Jan 10, 2018 3:08 pm
Thanks! The new version I should publish this weekend of Andscacs has a change precisely on this imbalance.
CMCanavessi
Posts: 1142Joined: Thu Dec 28, 2017 4:06 pmLocation: Argentina
Post by CMCanavessi Wed Jan 10, 2018 5:23 pm
Interesting. In the Elite League of my tournament, this game just happened:
cdani
Posts: 2204Joined: Sat Jan 18, 2014 10:24 amLocation: Andorra
Post by cdani Wed Jan 10, 2018 10:26 pm
Nice game, thanks! Is clear that black could do nothing as its king was too exposed. So Andscacs evaluation was apparently better.
