XXXIX Chess Olympiad (Khanty-Mansiysk 2010) performance ratings: women, on board prizes

How should one decide the board prize winners? The traditional method is to reward those who have reached the best percentage score out of those who have played at least a certain predefined number of games. This is a very simple and intuitive criterion but unfortunately it has a couple of drawbacks. First of all, the highest percentage scores are commonly reached by players who are considerably stronger than the rest of their team, or who play a lower board they should. Of course, this may even be seen as desirable in order to increase the motivation of players whose teams are not strong enough to fight for the team prizes. Still, this may not be the best way to achieve such a goal, since the method seems to suggest that if one truly wants to fight for the board prize, (s)he should play a lower board than the strength suggests and make sure the teammates perform as poorly as possible. Furthermore, a player with a high percentage score and a sufficient number of games is bound to be reluctant to play any further games since that would severely risk getting the prize. This seems all but encouraging.

Recently the percentage score based prizes have often been replaced by rewarding the players with the best performance ratings. This certainly fixes the first problem since the strength of the opponents is now considered. However, since the performances are commonly calculated by the method advocated by FIDE, the second problem may have gotten even worse. When the performance ratings are calculated using the average rating of opponents, wins over poor opponents regularly lower the performance. Thus, players interested in performance based prizes are encouraged to avoid facing weak opponents. Also, the handling of perfect scores is purely arbitrary and the performances calculated for those are really not based on anything. Furthermore, since it is relatively easy to score a high performance on just a couple of games, a fixed minimum number of games is still needed. Finally, exactly the same argument than before holds: a player with a high performance may not wish to endanger it by further games.

Now, there seems to be a reasonably straightforward way to fix these problems. Instead of the average based performance formula, one would calculate the performances the proper way, finding such a rating that the expected score equals the actual score. However, first one would reduce all the scores by a single point. This equals to adding a fake starting round where everyone loses against a very weak player. Thus, everyone needs to start working their performance up from the lowest possible. It will likely take a few games before even the best have reached performances that correspond to their rating. Thus, playing actually helps and each and every win is valuable. Of course, the selection of a single point reduction is arbitrary as well but it's the only arbitrary selection here and it seems natural enough, although it may not always be the best possible. For example, reducing two points instead of one would increase the significance of the number of games thus bringing the risk that someone might be tempted to sit on the performance reached even further down. The flipside would be the reduced importance of the actual performance strength.

But enough chit-chat, let's see who would be the top ten players at each board with the suggested one point reduction. The modified performances can be found in the column m.perf. If you didn't bother to read any of the above, read this: no prizes will be given based on these lists. Unfortunately.

Juha Kivijärvi