#### Only Chess Forum

1. 06 May '06 07:51
Anyone know what the median rating is on this site?
How about the different percentiles, like 75th percentile?
Thanks.
2. 06 May '06 08:12 / 1 edit
The median is around 1280.

If you go to the Player table you'll se that there ar 376 pages with active members (subs and no-subs). The page 376/2=188 have members around 1280.

If you call the page where you find yourself N, you can use the formula 1-N/376 to find out in which percentile you yourself belong to.
3. 06 May '06 09:06
Originally posted by FabianFnas
The median is around 1280.

If you go to the Player table you'll se that there ar 376 pages with active members (subs and no-subs). The page 376/2=188 have members around 1280.

If you call the page where you find yourself N, you can use the formula 1-N/376 to find out in which percentile you yourself belong to.
Such a clear and complete answer (and without a single wise crack). Thanks!
4. 06 May '06 16:56
Median or mean? Personally I would use mean.
5. 06 May '06 17:01
Originally posted by FabianFnas
The median is around 1280.

If you go to the Player table you'll se that there ar 376 pages with active members (subs and no-subs). The page 376/2=188 have members around 1280.

If you call the page where you find yourself N, you can use the formula 1-N/376 to find out in which percentile you yourself belong to.
Wouldn't it be more accurate to go by the number of players instead of the number of pages, as there are more than 1 players per page?
6.  DeepThought
06 May '06 19:52 / 1 edit
Originally posted by prosoccer
Median or mean? Personally I would use mean.

Originally posted by LanndonKane
Wouldn't it be more accurate to go by the number of players instead of the number of pages, as there are more than 1 players per page?
The problem is that it would be a total pain to calculate the mean, the easiest way to do it would be to write a robot to hit every page in the player tables, read the rating of each player, and get the mean from that. You'd probably find it came out at 1,200...

No, people's ratings vary by more than a page on a daily basis. If you look at page 250 of the ratings table (1,200 - 1,201) +16 for a win against another 1,200 rated player moves you to page 239. People finish games on a daily basis. Essentially it's only accurate within the variance of an individual players rating. I bounce around between 1,700 and 1,800; assuming that range (100 points) is fairly typical the percentile calculation gives you a result which is only meaningful within that range, so for me that's 1,750 +/- 50 = page 13 +- 3 = 97% +- 1%. For a 1,200 player that's about 40 pages [40/376 =10%] so 1,200 = 50% +/- 10%. So in terms of percentiles the error is between 1%and 10%, there is no point in doing a calculation to an accuracy of 1 part in 10,000; it doesn't give you any extra information.

Edit: Changed brackets from round to square to avoid
7. 06 May '06 19:59
The 50'th percentile is the median.
To know what percentile you're at you don't need the mean.
In this context percentiles are more interesting.
8. 06 May '06 20:20
Originally posted by prosoccer
Median or mean? Personally I would use mean.
A number of casual players sign up for the site, start a game or two and never return, perhaps timing out 6 games and winding up with a 700 rating. Average that with a person whose played 500 games and has a rating of 2300 and you get an average of 1500. Of course you could go for median in which case 100 people who timed out 6 games and never game back balance out 100 people who've played 10,000 games combined.

The most meaningful number, in my opinion is to count the starting rating of each player in each game and average that out (or take the median). In this method, if a player has played 100 games, his rating would count 100 times toward the mean/median. If a player had only played 6 games his rating would only count 6 times toward the mean/median.