| |
BGonline.org Forums
ELO
Posted By: Rich Munitz In Response To: ELO (Jake Jacobs)
Date: Wednesday, 16 December 2009, at 7:35 p.m.
First of all, there is a great deal of variance in Elo ratings, so that a 1700 player might at any moment in time have an Elo rating between say 1600 and 1800. But on average, players will be above their true rating as often as they will be below and with enough of a sample, this should balance out.
The FIBS Elo formula gives the 200 Elo favorite a 62.6% chance of winning and apportions rating points in this ratio (i.e. giving fair odds) so that if two players that were truly 200 Elo apart played an infinite number of matches and did not gain or lose any actual skill over the period, their ratings would remain on average 200 points apart. The Elo formula does not care in the least what the absolute ratings are - only relative ratings, and these two players must produce win percentages in the long run to match the Elo formula by definition.
So a question is why does the Fish MET's 61.9% at 5-away, 5-away differ slightly from the FIBS rating formula prediction of 62.6%?
I think that where the Elo system breaks down somewhat is that you don't have ratings come about because of the same two players playing against each other. The ability to factor skill difference into one's play is a skill itself. Some 1900 players may have that rating because they are able to achieve more success than predicted against 1700 players while having losing records against other 1900 players that are not as good at beating 1700 players.
And of course, there is always the issue of ratings manipulation, or players willingness or unwillingness to play new players with low experience that may have ratings far from their proper levels. Or players that only play with closely rated players vs those that will play anyone.
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.