Man vs Machine, Zap!Chess vs Erwin L'Ami 1:1

Zap!Chess vs Erwin L'Ami: 1-1

Report by Eric Reem

The seventh Match in Maastricht was opened by Prof. Dr. Jaap van den Herik, the editor-in-chief of the ICGA Journal and tournament director of many computer chess World Championships. In his opening speech, Van den Herik spoke about the sensational tournament Zappa played in Reykjavik in 2005. Zappa, developed by Anthony Cozzie, won the world championship with 10.5/11 games. After a slightly disappointing result during the Computer Chess World Championship in Turin last year, Cozzie retired from computer chess as an active programmer in order to concentrate on his PhD work. Much to the surprise of computer chess fans, his new engine, Zappa Zanzibar was released early this year and proves to be about 100 Elo points stronger than his older engines, Zappa Reykjavik and Zappa Paderborn.

The organisers of the event have promised to continue staging such matches until 2016, and with new ideas. Recently the chess engine Rybka played a match against GM Jaan Ehlvest in which the computer had white in all eight games and was missing a different pawn in each game (you can read about it in Jan Ehlvest's blog). The final score: 5.5-2.5 for Rybka.

The Ehlvest vs Computer handicap match was certainly an interesting concept, but the “think tank” in Maastricht (van Reek, Gijssen, van den Herik, Van Reem, Brorens and Uiterwijk) thought that the idea of being a pawn up would be disconcerting for the human player. Therefore they chose a formula in which the incredibly fast chess machine had less thinking time than the human player. In Maastricht, Zappa has only 30 minutes for the whole game, with the permanent brain mode (thinking on the opponent's time) switched off. GM Erwin L'Ami (Elo 2594) has 2 hours for the first 30 moves, another 2 hours for the next 30 moves and after that 15 minutes + 15 seconds per move for the rest of the game.

L'Ami misses chances to win the first game

At 13.30 referee Geurt Gijssen (who checked the toilets carefully before the game) started the clock and Zappa opened 1.e4. The Dutch GM selected a quiet opening and although Zappa showed a slight plus after about 20 moves, L'Ami never had problems equalising the position and even had a better position in the rook ending. An interesting position arose after 36…d6. What would have happened after 37.exd6? Start your engines! On move 40 Zappa made a mistake by playing 40.f5? Suddenly the engines in the press and analysis room showed a plus for black. But L'Ami did not manage to bring home the point, and his 53rd move obviously was a mistake.

The start of game one, with operator Jos Uiterwijk and GM Erwin L'Ami

After the game the human player said: “I was pleased with the opening, and good easily reach an equal position. The 12th and 15th move of the computer were not very good, I think that 15.f4 would have been better for white. I thought that the game would soon end in a draw, but Zappa made a few inaccurate moves. Unfortunately I made some mistakes in the endgame as well and could not win the game. I am not completely sure where I made the wrong decision, I will take a look at the game in my hotel. I think that I had good chances to win the game somewhere".

The grandmaster sinks into concentration

The second game was played behind bars! The playing venue, which is located in the historical heart of the city of Maastricht, used to be a women’s prison until the mid-seventies…. Erwin played pretty fast, and showed a typical plan in the Queens Gambit Declined Exchange Variation: he decided to open the game on the Queens side by playing 14.b4 and 15.a4. Zappa’s move 18…cxb5 seems to be new (18…axb5 was played before in the game van Wely-Hjartason, 1995) and now Black can create a passed pawn by playing 19…a5. After that move Erwin, who did not like his position, invested a lot of time. Zappa won a pawn (Qd7-Qc6-Qxb6), and played the very active move 25…Ne4! However, White could solve his problems, won back the pawn (with a little luck, Erwin said) and swapped some heavy pieces. The position that arose was equal, but both Zappa and L’Ami had to play carefully to keep the balance. After 42 moves Erwin offered a draw, which was immediately accepted by Zappa operator Jos Uiterwijk.

Checking the game with the computer operator

Reaction from Erwin L'Ami

“I did not play very well in the second game. The opening went the wrong way, because I had prepared something completely different. My goal was to get a set-up with c4-d3-e4-g3-Bg2 etc. and than slowly improve the position, but Zappa played unexpectedly 2…e6. In my test games Zappa often played 2…e5. And my plan with a4-b4 and b5 was bad, because I opened the position, which is not a good idea against the computer. I think 18.b5 was wrong, maybe 18.a5 is an option, followed by Na4-c5. Zappa could have tried 24…Ne4, instead of 24…Qxb6, I have to check that at home. I was not satisfied with my position at all and even in the endgame I had to play very accurate moves to get a draw, e.g. I completely missed 32…g5 and had to be careful not to get in a mating net! In the end I was happy with the draw and the final result. Yesterday I missed a win, today I was not happy with my play, but the result was OK.”

Jaap van den Herik, Jos Uiterwijk and Erwin L'Ami in postgame analysis

• Click to replay and download both games (annotated)
• Official match web site

Zap!Chess is the commercial version of Zappa, the 2005 World Computer Chess Champion. The style of the program reflects the background of the author as a computer engineer. Rather than developing clever search tricks that may or may not work, he concentrated on getting the most out of modern hardware. Zap!Chess contains one of the best parallel implementations in the world to run efficiently on multiple CPU systems, and it also uses 64-bit machines to their full potential. The program contains large amounts of chess knowledge, and like most modern programs it is tuned fairly aggressively – it knows where the opponent‘s king lives. This gives it an exciting style without being unsound. While the program is optimized for long time controls and big hardware, don‘t despair if you own a smaller system. The program comes with an implementation of Singular Extensions, the famous Deep Blue search algorithm, which can increase the tactical strength of the program at the cost of positional strength.

System requirements: Minimum: Pentium II 300 MHz, 64 MB RAM, Windows Me, 2000, XP, DVD ROM drive, Windows Media Player 9. Recommended: Pentium IV 2.2 GHz or more, 256 MB RAM, Windows XP, GeForce5 graphics card (or equivalent) with 128 MB RAM or more 100% compatible with DirectX, sound card, Windows Media Player 9, DVD ROM drive.

Price: €49.99 incl. VAT;
€43.09 without VAT (for customers outside the European Union); US $51.71 (without VAT).
Buy now.