|
|
|
Author
| Message |
|---|
|
|
Aren't correspondence chess ratings usually higher by 200 points simply because you can think about each move longer? What effects a rating? Maybe the "think" feature on this site even adds another 50 points due to its extreme usefulness.
Also, ANY set of ratings is valid ONLY for the particular group being rated. It's all about relative strength within that group.
And please stop worrying about provisional ratings. This comes up on these boards all the time. Until they are established, just ignore provisional ratings when comparing them to just about anything. When you see that little "p" next to a rating, take that rating with several hundred grains of salt.
--p
|
|
|
Those first two paragraphs are contradictory - if ratings are only relative to the pool, then features available to everyone in the pool (think mode, correspondence time controls, etc.) shouldn't make any difference to the overall ratings distribution, even if they do have a positive effect on quality of play.
The standard of correspondence chess generally may be higher than OTB chess, but you can't use ratings to make that comparison because they aren't an absolute measure of performance.
|
|
|
I've long suspected that correspondence ratings are generally higher than OTB for most players becau8se a lot of people try it, get trounced, and then quit. These "easy" wins earned by those that stick with it serve to inflate the ratings.
Along similar lines, flaws in the calculation of provisional ratings will rapidly skew most ratings in the pool in the diresction of the flaw. The average rating may not reflect the inflation that becomes the norm for most active players.
Arpad Elo's formula is rooted in the assumption of an average rating of 1500. When the average is substantially higher, the usefulness of the ratings diminish. Established ratings are always useful within a given pool for identifying certain levels of competition, where a good game can be expected. But rating pools where 2200 represents a certain level of mastery are preferred, as here players may set meaningful goals for improvement.
|
|
|
| a lot of people try it, get trounced, and then quit. |
Unfortunately it's even worse than that, a large proportion of players start a few games and then never come back and finish them; resulting in a lot of easy wins and over-inflated ratings for the more persistant players.
Another confounding factor on a casual site such as this, is that you can choose your own opponents, and whether you play as white or black. Ideally, only randomly paired games that didn't end as a win on time should be rated at all - however this would probably be unacceptable to the membership and would also take a lot of the fun out of the whole thing.
| flaws in the calculation of provisional ratings will rapidly skew most ratings in the pool in the diresction of the flaw. |
Yep, every organisation which uses Elo for ratings has a different way of calculating 'provisional' ratings. As far as I know, Elo didn't propose a way around this problem.
A better way of dealing with this crudity is to drop the idea of a provisional period altogether, and instead attach a degree of certainty to each player's rating, which improves as the number of games played increases (so that the results of games against less established players are less influential on your rating).
|
|
|
I'm talking generally about casual correspondence chess servers here btw, not queenalice specifically 
|
| Previous 1 2 3 4 5 Next |
|
|
|
