07 February 2009

Adventures with Arena & Rybka

A week after my Arena / Rybka Analytical Upgrade, I continue to be happy with my choice. Rybka is everything I expected it to be. It's fast, super strong, and an excellent assistant for chopping through complex tactical positions.

Arena appears to have been conceived for running engine vs. engine tournaments, although it's also useful for playing against a specific engine. It's somewhat less useful for general analysis. Its handling of PGN files is clumsy and its display of variations is difficult to take in at a glance. It more than compensates for these drawbacks with an excellent interface to the internals of the engine while it's running. I especially like being able to review the previous iterations as the software calculates deeper and deeper.

As I mentioned way back when in Where I Play, my main chess competition is via email servers. I decided to put my new toys to a practical test and joined a tournament on Playchess.de: ACL-M116. (NB: Playchess.de shouldn't be confused with Playchess.com, the Chessbase online site for crossboard play.) Unlike many online email servers, the site allows players to use engines in its Advanced Chess League (ACL) tournaments.

As for other sites where I play, I've stopped using the ICCF server. Its high cost per event and its policy of requiring players to join a national federation make it a more expensive option than the free IECG server. The IECG players are just as good as those in the ICCF, and many excellent players compete on both services. It's generally assumed that most players on both services use engines, so why should I pay extra to play the same opponents using the same engines?

Lately, my results have been less satisfactory on IECG, and my rating has declined 200 points over the past few years. Opponents rated 400-500 points lower than me have been offering stiff competition even though they seem to spend little time on their moves. I've had several recent games where an opponent's strong reply typically came only a few hours after I sent my move, and where I had to work hard for a draw. I compared these games with Rybka's analysis and found that most of my opponents' moves matched Rybka's top suggestion. If I allowed Rybka to run longer on the exceptions, I bet I'd see close to 100% matching moves.

Now that I know why my results have been off, I can do something about it. First I have a general problem to solve: What is the best anti-engine strategy for use against a specific machine?

No comments: