After seeing my performance-both in and out of tournaments–several people, including tournament directors, have commented on the fact that I perform much better than what my USCF rating shows. How accurate is the rating system, and is there a way to have my rating recalculated? Myself and other estimate that my rating ought to be about a class level higher.
How accurate is the rating system?
The rating system does what it does. It gives or takes rating points based on how well you do in RATED tournaments compared against your statistical expectation. The formulas are pretty straightforward as stats goes.
The only way to be under-rated or over-rated is if you’ve had a huge improvement or erosion in your skills independent of rated events. If you play enough your rating and your skill level will reach a dynamic equilibrium of sorts.
Yes. Play a lot of tournament games, and win a lot. Or lose a lot – that will also result in a recalculation, albeit a less satisfactory one.
Tongue out of cheek, my suggestion is to try, as much as possible, to play opponents your own age. Many young players improve much faster than their ratings reflect and consequently are significantly underrated. Losing against them will cause you to be underrated as well. Story of my life, currently.
You’ve played a total of 60 USCF rated games, but just 8 USCF rated games in the last year. (That doesn’t count your recent quick-only event.)
8 games is not enough games to make a significant impact in someone’s rating, especially since you drew 3 and lost 5 of those games. On that basis, other than the draws against higher rated players, there’s not much empirical evidence that you are seriously underrated.
If you want your USCF rating to be reflective of your current strength, you need to play more USCF rated games, especially against players with higher ratings than yours (but probably not 600 points higher.)
For example, using the ratings estimator page, if you played four 1200 players and went 4-0 against them, your rating would go up by over 300 points.
In other words if you play “higher than your rating” and still lose, then you do not gain rating points as we have no way to know that you played well in a loss as opposed to getting crushed.
“It is not enough to be a good player; you must also play well.” - Siegbert Tarrasch
The rating system measures your performance in tournament games. If you feel you are good at chess but underperform in tournaments, the rating system doesn’t know anything about it, and all you can do is learn to play as well as you and others think you are capable of when you are in a tournament situation.
In your last four rated tournaments you:
went 3/6 against 744-rated opposition
went 2/5 against 805-rated opposition
went 1/3 against 1021-rated opposition
went 1.5/5 against 1676-rated opposition
Those results in the first three tournaments are exactly what one would expect of a player rated around 800, so it is hard to claim that you are underrated at 925. Keep having more tournaments like that last one and your rating will rise quickly. (You will note that you gained a whole 133 points from that tournament, and about 60 of that came from “bonus points”, which is the rating system saying “wow, maybe he’s a lot better than his rating indicated”.)
Another illustration: A result of 2-3 against a field whose average rating was 1500 would be a spectacular result for someone whose rating was 1000 but a disaster against a 2000 rated player. Somewhere between those two a player of some rating would neither gain nor lose points. Whatever that rating is, that’s what would be given to an unrated player (turns out to be 1430).
Click for the USCF rating estimator.
Since we’re on the subject of ratings, here’s a question of no particular consequence, but which seems interesting to me:
What’s a good way to calculate a probable score against a field in a Swiss? Taking the average rating of the field, entering that average number in the rating estimator for each round in the event, and then tweaking the score until your “new rating” doesn’t change much (i.e., your performance rating is pretty close to your pre-event rating) – would this give a meaningful prediction? Or would you have to model the event using an automated pairing tool (assuming no upsets and every game performs to expectation)? I first assumed that because you get paired down when you lose and paired up when you win, that using the average rating wouldn’t be far off. But what if the strength of the field is lopsided, e.g., a few strong players and many weak ones, or the reverse?
There may not be a simple answer to your question, because the expected score function is not linear.
A 1500 player who plays three 1500 players would be expected to have a different total score than one who plays two 1000 players and a 2500 player, even though the average rating of his opponents is the same.
It’s approximately linear, though, isn’t it, for rating differences less than 350 or so?
Bill Smythe
That sort of depends on what you consider linear:
uschess.org/datapage/rating_curve.png
At around a 200 point ratings difference, the probability function says that the lower rated player should score around .23, a linear function between -350 and +350 says it would be around .27
That .04 difference. compounded over several games, might make a few points ratings difference, possibly even impact whether bonus points are awarded. (And even a few points in their rating means a lot to some people.)
Someone who is serious contention to win the tournament will meet a field stronger than the mean while someone heading towards the cellar will meet a weaker field. Taking your method and doing a weighted average against a break even score might be one method. Another method might be to take a (weighted?) average between a player’s rating and the mean rating and use that as the average of the field.
Crunch a number of players in a few rating reports and see if a pattern develops or if you can find the appropriate factors for weighted averages.
Some players are stronger at analyzing positions or playing blitz than they are at playing slow tournament time controls. They can see brilliant combinations but may miss a subtle refutation.
The best way to improve tournament play is to play more slow games, both rated and for fun. Don’t they say “practice makes perfect?”
Michael Aigner