Thursday, January 29, 2015

Beware of Chess Engine Opening Books

 
Be more suspicious
  
 ...or any opening book for that matter. Engine opening books are made up from a database of games and engine play is significantly enhanced by the use of opening books. Most opening books are made by taking thousands of games and creating a database of the moves.
     Unfortunately many of the games included can be those of weak players, even near beginners, totally without regard as to the strength of the moves played. Even if the book was constructed using all Grandmaster games, the moves cannot always be considered good just because they were played by Grandmasters.  It's also quite possible the moves in the book are well past their "best if used by" date; recent innovations may have drastically changed a moves evaluation.
     When you build an opening book from a large collection of PGN games you run the risk of including a bad line that leads to an inferior position or even one that loses outright.  Of course a large collection of games is required to cover a reasonable number of possibilities that one is likely to meet, yet at the same time a large collection of games may well include serious blunders. This is why so many top level correspondence players spend an enormous amount of time fine tuning their opening books and looking for improvements. Most of us average CC players don't go into all that, but one still has to be careful when selecting opening moves in correspondence games. I ran into a good example of this recently in the Fritz 12 opening book.
     The line in the French Advance Variation is from Capa vs. Paredes and the book ends after 17.gxf5 and shows one game in the database that was won by white. Searching the database turned up the Capa game and so all his moves must be good, right? It turns out that the play after move 13 left a lot to be desired on both sides and so blindly following the opening book could easily result in disaster.
     This is an old story though. Back in the 1970's, before chess engines were even heard of, I was playing a correspondence game against an opponent who was one of the top OTB players in the US; he had even participated in a couple of US Championships. I was white and we were following the line in a popular opening booklet and I sent him my move copied directly from the booklet. Later I decided to look at the game because we were near the end of the book line and that's when I discovered a move that wasn't in the book for his next move. To me, it looked devastating. It turned out I was correct because that's exactly what he played. After the game (I lasted only a few more moves) he asked if I was using that particular booklet and when I said yes, he advised me, "There's a mistake in it."
     Always check things out for yourself and I am not talking just about chess. I am talking about everything. Never take someone's word for anything especially if the clues, or your gut, tell a different story. Someday I may learn this lesson myself.
     As for the game, there was something called the Havana "casual" tournament in 1901 that was apparently won by the 13 year old Capablanca. His opponent in this game was Leon Paredes, a prominent player of the day who served as president of the Havana chess club.


No comments:

Post a Comment