Really interesting post. I love reading about this. One of my students did his capstone project on whether the "Swiss gambit" is an effective tournament strategy (spoiler: it's not). He used Elo's formula to calculate expected scores for different rating distributions. It is surprisingly simple and mostly derives from the assumption that the winning odds only depend on the difference in ratings.
Something else I found interesting is that rating points are only exchanged. The higher rated player antes up more points but the total is always k, and the winner takes all (or 50/50 split for a draw).
I guess it's useful but I personally find it near meaningless. In addition to the issues you mention, for someone who doesn't play many rated games, it forever lags their actual rating (if that exists). And the obsession is probably unhealthy. I also hid the ratings on Lichess and love that feature, though I occasionally peek as well...
Chess is better suited for elo in some respects: The level of players is more consistent over the years, they play a lot of games and their actual value as a human being depends on it
I think elo ratings is more useful than the segments proposed by D.Heisman. But the only rating that reflects your OTB strength is the FIDE standard rating. All this national ratings like USCF, BCF, FCE and so on, are completely misleading and worth nothing.
Online the only rating I care is the Puzzle rating because It shows if I am improving or unproving my tactical awareness. So I use It for training purposes.
It always saddens me when a commentating GM stops playing to preserve his rating. You stop doing what you love because of your honor and marketing reasons, while the insiders know what's happening.
Considering the importance people attach to their rating, it is an economic miracle there hasn't been much inflation (in my experience over the past few years there's actual deflation, probably due too better training methods).
Also there is interesting article of chessbase you're probably familiar with that basically says players consistently underperform against opponents -100 to -200 points lower. That frustrates stronger players and disincentivizes them to play in some tournaments or leagues.
Enlightening post, thank you. I realize your primary objection to ratings is the impact on our psyche. But I also wonder if you think the Glicko ratings from Glickman solve statistical issues like draws and the White player’s advantage. (Awesome that you are playing Pokémon with your young nephew!)
I have mixed feelings. I think Glickman's ideas make sense from a statistical perspective, but the system is very well established at this point and there will be a lot of resistance to changes.
Some tournament organizers on the West Coast use section entry policies that unfairly prop up the ratings of higher-rated players because players below a certain rating are banned from entering the higher section. Also, achieving a final plus score in a tournament is always difficult, and I don’t think the Elo formula gives a player any points for a + score.
In theory there should be no particular advantage/disadvantage to playing higher/lower rated players. In practice, I've seen some evidence that lower rated players perform a bit better than they're supposed to in the system.
There is no bonus for having a positive score in a tournament, besides the usual reward for winning more games.
Really interesting post. I love reading about this. One of my students did his capstone project on whether the "Swiss gambit" is an effective tournament strategy (spoiler: it's not). He used Elo's formula to calculate expected scores for different rating distributions. It is surprisingly simple and mostly derives from the assumption that the winning odds only depend on the difference in ratings.
Something else I found interesting is that rating points are only exchanged. The higher rated player antes up more points but the total is always k, and the winner takes all (or 50/50 split for a draw).
I guess it's useful but I personally find it near meaningless. In addition to the issues you mention, for someone who doesn't play many rated games, it forever lags their actual rating (if that exists). And the obsession is probably unhealthy. I also hid the ratings on Lichess and love that feature, though I occasionally peek as well...
Chess is better suited for elo in some respects: The level of players is more consistent over the years, they play a lot of games and their actual value as a human being depends on it
Lol
! LOL
Great post. It has made me so much more knowledgeable about the mechanics of rating updates.
I think elo ratings is more useful than the segments proposed by D.Heisman. But the only rating that reflects your OTB strength is the FIDE standard rating. All this national ratings like USCF, BCF, FCE and so on, are completely misleading and worth nothing.
Online the only rating I care is the Puzzle rating because It shows if I am improving or unproving my tactical awareness. So I use It for training purposes.
Nice post.
Does everyone (USCF, FIDE, LiChess, CC, ICC) use the same k-value?
It always saddens me when a commentating GM stops playing to preserve his rating. You stop doing what you love because of your honor and marketing reasons, while the insiders know what's happening.
Considering the importance people attach to their rating, it is an economic miracle there hasn't been much inflation (in my experience over the past few years there's actual deflation, probably due too better training methods).
Also there is interesting article of chessbase you're probably familiar with that basically says players consistently underperform against opponents -100 to -200 points lower. That frustrates stronger players and disincentivizes them to play in some tournaments or leagues.
Enlightening post, thank you. I realize your primary objection to ratings is the impact on our psyche. But I also wonder if you think the Glicko ratings from Glickman solve statistical issues like draws and the White player’s advantage. (Awesome that you are playing Pokémon with your young nephew!)
I have mixed feelings. I think Glickman's ideas make sense from a statistical perspective, but the system is very well established at this point and there will be a lot of resistance to changes.
Some tournament organizers on the West Coast use section entry policies that unfairly prop up the ratings of higher-rated players because players below a certain rating are banned from entering the higher section. Also, achieving a final plus score in a tournament is always difficult, and I don’t think the Elo formula gives a player any points for a + score.
In theory there should be no particular advantage/disadvantage to playing higher/lower rated players. In practice, I've seen some evidence that lower rated players perform a bit better than they're supposed to in the system.
There is no bonus for having a positive score in a tournament, besides the usual reward for winning more games.