From: Steinar H. Gunderson
Date: Fri, 5 Oct 2007 21:34:08 +0000 (+0200)
Subject: More writing about FoosRank.
X-Git-Url: https://git.sesse.net/?a=commitdiff_plain;h=8ed5409ab738e8c4fbf30b3279b97c6807802169;p=foosball
More writing about FoosRank.
---
diff --git a/www/index.xml b/www/index.xml
index d645a04..315a7ac 100644
--- a/www/index.xml
+++ b/www/index.xml
@@ -160,17 +160,18 @@
About the ratings
The rating system in use is called FoosRank, and is specially
- designed for this system. It is a Bayesian rating system, inspired by
- Glicko 1 but re-worked out from scratch with a statistically sound model
- of non-binary results, and adjusted for teams (with some ideas from
- Microsoft's TrueSkill system). For those not familiar with Bayesian
- ratings, the most important parts are:
+ designed for this site. It is a Bayesian rating system, inspired by
+ Glicko but re-worked
+ out from scratch with a statistically sound model of non-binary results,
+ and adjusted for teams (with some ideas from Microsoft's TrueSkill
+ system). For those not familiar with Bayesian ratings, the most important
+ parts are:
- Your rating is a statistical estimation of your true skill.
It has a mean (the point estimate of your skill) and a deviation
(measuring the uncertainity of the estimate), called the RD. It is approximately
- Gaussian (actually logistic).
+ Gaussian.
- When you win or lose a game, your rating will change accordingly,
based on your score and your opponent. You do not get 'points'
for winning or losing, the estimate is merely getting more accurate.
@@ -187,6 +188,11 @@
you are ranked as an individual, as we do not usually play with fixed
teams.
+
+ For the more technically minded, a
+ very rough writeup
+ on the mathematics behind FoosRank is available in PDF format. It is,
+ however, by no means finished.