BGonline.org Forums

ELO in chess

Posted By: Nack Ballard
Date: Wednesday, 2 April 2014, at 1:16 a.m.

In Response To: The Donkey Corner #6 - Rollout + Analysis (Robert Chow)

This strays from the topic of backgammon PR (and I'm not much interested in rating systems in general), but since you brought it up...

My understanding is that a player stronger by 100 ELO is expected to score 64%, and a player strong by 200 ELO to score 76%. (In chess, draws exist, which count half a point, so 76% might mean 76 wins 24 losses 0 draws in 100 games, or 52 wins, 0 losses, 48 draws, or you can use linear interpolation to find other distributions.)

As I recall, the chances of a stronger player beating a weaker player is the same regardless of absolute level. However, the "K factor" (which addresses how many point are at stake) might make it appear otherwise. For the USCF, the K factor is (or was) 32 below 2100, 24 from 2100 to 2400, and only 16 above 2400. For FIDE, the numbers (and parameters) are tighter, maybe just K = 10 at the top, but I think the 200 ELO = .76 is the same for both systems and regardless of level.

For example, Magnus Carlsen, if playing at his (peak) rating of 2881, should score .76 (something like 64% wins, 12% losses and 24% draws) against a lowly 2681 player.

Nack

Post Response

Subject:
Message: