by Steve Lopez

Every chessplayer in the world (regardless of playing strength) likes to think that he or she is a pretty good player. It's what keeps us going: the idea that we're decent players and that we are improving, no matter how slowly.

Using computers as training tools gives each of us the opportunity for rapid improvement. Even a guy like me who is overburdened with work and family responsibilities can get a chess game any time of the day or night by using a computer (either against a program or an unseen on-line opponent). I usually have two or three e-mail games going at any given time as well. I'd like to think that the limited chess I do play is improving my game. I think most amateur players feel the same way.

That assertion is now pretty easily tested with the advent of statistical chess programs. You can load a bunch of your games, run a statistical function on them, and see how well you're really doing.

For example, I was playing in a chess tournament every month or two a few years ago. I was starting to win class prizes and saw my rating on the increase. Using the statistical functions in ChessBase, I was able to determine that I was winning about two-thirds of my games (both casual and tournament) over a two year period with very few draws. It wasn't my imagination -- I really was getting better.

Checking the statistics was pretty easy to do: I just did a search for all of my games as White, put them on the ChessBase clipboard, and clicked the button for statistics. I then did the same for my games as Black. I noticed that I was doing much better as White (which was not unexpected), and that I averaged about 65% of the total points available in the games I played.

That's all well and good; after all, it's nice to be able to pat oneself on the back occasionally. But what about my weaknesses? How could I improve what I was doing wrong?

The obvious approach was to have Fritz analyze my games and show me my errors. But there's another path to improvement that's a bit more subtle.

Chessplayers tend to spend a ridiculous amount of time studying the openings. I'd venture a guess that the average player spends twice as much time on openings as he does on all other aspects of the game combined (endgames, strategy, tactics, general technique). How can you tell whether or not this study time is paying off?

At this point, I'll turn the floor over to Fritz user Laurent Selvi, who sent me the following e-mail earlier this week:

I wish to contribute with a way I use Fritz's opening tree I find very useful : to make statistics on my own play -

I read that making statistics on your own games was a good way to appreciate the ease you had with any given opening variation (this advice is given by S.Webb in "Chess for Tigers", by the trainer Kevin O'connell and by Pierre Meinsohn, a 2300 French writer). For example, Pierre Meinsohn suggests that you can draw conclusions after ten games in any given variation and states that if you do not reach 70% of non-losses (draw or win = non-loss) with White or 60% with black, then you should seriously reassess your use of the variation. On the other hand, Kevin O'Connell states that percentages are not very relevant, but that you should calculate your performance with the opening.

Well, it's very easy to do either of these with Fritz : you just have to create two databases : one with your games played as white, the other with your games played as black. Then you create two separate trees from these two databases. Just reading the figures in the trees will give you your overall percentage, perf and "non-loss percentage" (just add winning and drawing % in the graph bars at the bottom of the tree) in any variation. This way you can assess your openings with some objectivity.

Updating the trees is fairly easy as well. As the number of games is bound to be low, you can just "delete the whole tree" and rebuild it each time you play a new game.

This sounds so easy! Isn't there a catch somewhere?

Not really. Simon Webb, in his book Chess for Tigers [Pergamon Press, 1990, ISBN 0-08-037788-2], recommends making a chart of the openings you've played, counting the games you've won, lost, and drawn, and using this chart to determine how successful you've been with each opening. If a particular opening or variation has a high success rate for you, keep it in your repertoire. If the success rate is low, either do some more work on that opening or else give it up altogether.

At the time Webb wrote this book, checking your opening preferences in this way involved some major grunt work. You had to flip through your old scoresheets and tally the games by hand. Nowadays with ChessBase or Fritz you can do the same work in just seconds.

I've been inactive as a OTB tournament player for nearly three years now, but I do keep records of the casual games I play against humans and computers. I went into each of my personal databases and sent all of my games as White for the last fourteen months to the ChessBase clipboard. I hit CTRL-A to highlight them and then SHIFT-ENTER to create the tree. In less than two minutes from the time I started I had obtained the statistical information I needed.

I quickly found out some interesting things, both positive and negative, about my play with the White pieces:

It's helpful, though, to do a bit more digging and not take these numbers at face value. For example:

So even though I'm doing 62% overall as White, there are definitely areas that need a ton of work.

Note that these are casual games for the most part (a few on-line rated games, as well as rated games against chess programs, are in the mix), so I usually don't have ratings available for myself or my opponents (therefore I shouldn't take that 2984 performance rating in the French Advance too seriously). If you're primarily looking at rated games, take the ratings of your opponents into account. If you're generally winning against lower-rated opponents with a particular variation, take that big percentage with a grain (or even a pillar) of salt. Likewise if you're getting your clock cleaned by significantly higher-rated opposition when you play a particular variation, don't make any major judgements about it until you get the chance to try it against players closer to your own level.

As an editorial note, I disagree with the recommended percentages attributed to Pierre Meinsohn earlier in this article. I think 70% for White and 60% for Black are awfully high. I think that 60% for White and 50-55% for Black are more realistic. Do a few trees and dossiers on strong players using ChessBase 7 and see if you agree.

In any case, it's easier than ever to evaluate your play using computers -- you just need to know how to do it. Hopefully this week's ETN has provided you with some ideas along those lines.

Well, now that everyone knows my strengths and weaknesses as White, I need to go now and alter my entire repertoire. A very special "thank you" goes out to Laurent Selvi for the valuable and thought-provoking e-mail. Until next week, have fun!

You can e-mail me with your comments, suggestions, and analysis for Electronic T-Notes. If you love gambits, stop by the Chess Kamikaze Home Page and the Yahoo Chess Kamikazes Club.