Played in a local tournament this weekend and was surprised to encounter so many underrated players. My first opponent was rated around 1150 when his true strength was really about 1400-1500. My second opponent was around 850 when he played like an 1100-1200 player.
My last opponent was rated a little over 600. I happened to glance at the pairing and he had defeated a 1400 and a 1500 player. So I was careful and beat him. It turns out he had just hired a chess coach who told him he played at a 1700 level. Here’s my game against him. Not pretty, but I won.
This is a problem when there aren’t consistent opportunities to play rated over the board chess against players near ones rating. Losing to players rated several hundred pont higher loses one a few points, but doesn’t address one’s real ability level.
You also point to another issue, the player who has started using a coach but who hasn’t competed much since commencing studying with the coach. In some areas where there aren’t many tournaments available for U1200 or U1300 players, the accuracy of a player rated at that level or lower may well be inaccurate, as you discovered.
Your last round opponent played 3 games seven years ago. He was officially unrated for your tournament. That 600 rated based on three games seven years ago was a very poor estimate of his current strength.
Mike Regan
I would say a rating is only accurate if a player has played consistently against other players with a similar rating. If a player has mostly losses against players who are several hundred points above him then his true strength is probably lower than his rating, and if he has mostly wins against lower rated players his strength is probably higher. An unrated who loses to a 2200 would be provisionally rated 1800, but we don’t know if he is really an 1800 unless he plays another 1800 and can get a win or a draw at least sometimes.
Provisional ratings are notoriously unreliable, which I’m sure is why they are only “provisional.”
You’re right that there is no immediate win. But after something like 17.fxe6+ fxe6 18.Nf3 Qxd1 19.Rfd1 Bf6, Black has an extra passed center pawn supported by more active, much better minor piece, and possession of the only open file. See link
What is a routine win for you may not be routine for a less experienced player, of course.
You’ve correctly identified the flaw in provisional ratings, especially for players with all wins or all losses, but not their interpretation. All we really know about someone who consistently loses to players rated above him is that he isn’t their equal, similarly all we know about someone who consistently beats players rated below him is that they aren’t his equal.
The special formula, used for players with fewer than 9 games or all-wins or all-losses, tries to deal with this, but the reliability of those ratings is low.
This is separate from the point at which ratings are considered ‘established’, though provisional ratings are in general less reliable than established ratings.
As my stats professor told me years ago, a man with one bare foot in a pot of boiling water and the other on a block of dry ice is, on average, at a comfortable temperature.
There is a significant flaw in the calculations using provisional players. This past weekend, I won three out of four games in the U1250 section of our tournament and lost a game in the final round to a player a little higher than my listed rating. Two of the three provisional players I defeated were actually decent players from St. Johnsbury Academy’s chess club. Still, my rating dropped 8 points despite winning three out of four games.
It’s not so much a flaw as a limitation in what can be done with the available data. I haven’t looked at the specifics of this event, but in general with unrated or provisionally rated players there’s not much reliability in the data we have.
You’d have to as the PhDs how much a Glicko2-like system could impact situations like this.
The ratings estimator says you should have lost 10 points, so you did a little bit better than it estimated, but it isn’t given nearly enough information to deal with unrated/provisionally rated players.
3 of your opponents were unrated, one played just 2 games, losing both, another played 3, losing all 3, and both of them wound up with unpublishable post-event ratings of 399 and the third earned a post-event rating of 787 based on 4 games.
So the only game that had much impact on your rating was the loss to a player rated within 100 points of you.
I understand that. However, as I said, two of the new players were much better than one would have thought. They lost good fighting chess games to me and others including side games with Nita and Sasha who won the U1250 section. The bottom line for now is that their ratings certainly don’t reflect their actual ability and the points that I received, if I did, for beating them likely does not reflect accurately their strength in the games played.
Amongst lower class players who for whatever reason are only occasional rated tournament players, the ratings are very often not accurate indicators of playing strength. Another problem is limited access to tournaments with decent sized classes in some areas so that players don’t consistently have to play opponents 400 to 600 points higher than them. A loss with that disparity won’t lose the lower rated player many rating points, but if that’s all that you have as an opportunity for rated play, one’s rating will simply do a slow death spiral to well under 1000 when the player’s actual ability is significantly higher.
I’ve created a new topic, How can US Chess support the growth of clubs? in the Chess Clubs forum, to address the question of how US Chess can support the growth of clubs and encourage more people to play chess.