Mark Glickman's World

USCF Ratings:

 

I am currently the chairman of the U.S. Chess Federation (USCF) ratings committee, a position I have had since 1992. Besides helping to maintain the integrity of the USCF rating system, much of my research is devoted to ratings-related issues.

Below are some recent USCF ratings reports and other assorted ratings-related items.

Important: Most of the following documents are pdf files (unless otherwise noted).


Here is the specifications of the current USCF rating system. This revision, current as of July 21, 2014, incorporates the increase in the bonus threshold to B=10, corrects a minor error in the description about rating floors and the initialization of Blitz ratings, corrects two typos, and eliminates the antiquated concept of "provisional" ratings. The committee developed a set of approximating formulas that can be used for paper and pencil updates. This document has been revised as of March 2014.

The USCF title system is now currently implemented. The linked document (current as of October 2011) spells out the details of the new system. The current version makes some clarifications in the underlying approach. An Excel spreadsheet performing the norm calculations was created by Robert N. Bernard who let me post it on this site.

In 1998, in response to the title system not being implemented, the Ratings Committee developed the life achievement system which for unclear reasons to me had been shelved.

The feature article in the October 2006 issue of Chess Life Magazine was an interview of me (the article is posted here with permission of the US Chess Federation and of author Howard Goldowsky).

In the November 13, 2006, Boston Globe, Harold Dondis and Patrick Wolff wrote a flattering article about the Chess Life interview.

The Ratings Committee submitted a report to the USCF Executive Board responding to the ratings proposal by the DDDE committee. The report was submitted at the end of September 2000.

A plot showing the distribution of ratings for players aged 35-45 over the 1990s indicates that ratings for this arguably stable group have been generally declining over time. This may be some evidence that there has been rating deflation. This analysis was performed by me and Ken Sloan.

Click here for a plot showing the relationship between USCF rating and frequency of drawn games. This analysis was based on 1997 USCF data between established players whose ratings are within 100 points. It seems to show that the probability of drawing is higher (for closely rated players) when players' ratings are higher, in general.

I am the inventor of the Glicko system, a rating system that extends the Elo's by incorporating a measure of uncertainty of a player's rating. If you are considering implementing the system, you should take a look at my description of the Glicko system for details. If you are interested in the mathematical derivation of the system, check out my technical paper on the Glicko system.

I have invented another system, called the Glicko-2 system, that improves on the original Glicko system. It is also in the public domain (i.e., you are free to use it).

A slightly less ambitious version of the Glicko-2 system I developed is the Glicko Boost algorithm, which I used when entering the Deloitte/FIDE Chess Rating Challenge.

Some ratings-related research can be found on my research page.