22/6/08 10:30 | Mark Adams | Rating accuracy depends largely on the pool of data you are using to calculate ratings. The larger the pool the better. It’s another debate but to me, gradings will never be wholly accurate and should only be used as an indication of playing strength. For example, a rating could theoretically be based on games played 23 months prior to publication. This is an argument for regular publications for the rating list. FIDE (and Gwent!) is now doing almost a daily rating update for the top 100. Mathematically, players are still grouped together in bands of 200 points as used in the good old days with the class system (e.g. Class A players were rated between 1800 and 2000). So, for example, a player rated 1850 is more or less, the same strength as a 1950 player (you can get an 1850 rating by drawing against a batch of 1850 rated players, while you get a 1950 rating by beating a lot of 1700 rated players while never having been able to beat anyone above 1800)..
When you have several data pools you cannot say that a 2000 rated player in one pool is the same strength as a 2000 from another pool unless there is a sufficient cross over, if you like, between those pools. In Wales we have multiple data pools which will reduce accuracy to a certain degree (whether this is really significant though is open to question). Elo recommends a review of the rating pool after a period of time to check for inflation/deflation (we’ve seen inflation in FIDE ratings over the years. Super GM’s used to be 2600+, but now they’re 2700+). But what do you check this against? To have an inflation figure, you must have at least 2 references to compare. How to do this? Using historic data - maybe, but what historic data do you use? Do we compare, say, all 23 year olds in 1980 with 23 year olds on 2008? Do you look at average rating trends over a period of time? This is not straight forward as the total number of players have reduced so much over the last decade that the average Welsh rating then is incomparable to the average now (e.g. if most people who have given up were low rated players, then today’s average could be artificially higher). Do you say - Welsh Dai always beats English Cuthbert therefore even though Dai’s rating is lower, is he stronger than Cuthbert (maybe - but what about looking at Cuthbert’s other results - perhaps he had a bad day at the office when playing Dai. Or are Dai and Cuthbert in the same rating band anyway)?
Incidentally, an ELO rating usually means FIDE rating. I believe that WCU zones use the same rating formula as FIDE. |
