Our snailgrades appear to be giving a list of snailgrades capable of providing a seeding list suitable for future club tournaments. They are not, and will not be a reliable measure of performance until we record a lot more games. While there are so few games, they give very different results depending on the order in which games are played.

In the column next to our snailgrades I have now added Glicko ratings. These provide a measure of reliability of each rating, the ‘Ratings Deviation’, RD. Thus today, Brian M, who has played only one game has been given a glicko rating of 1567(299). He has a glicko rating of 1576 with a RD of 299. We can interpret this as meaning that we are 95% sure Brian M has a rating between 969 and 2165. ( 1567-2*299 and 1567+2*299 ) – Well yes, we could have guessed that.

Michael, who has so far played as many games as anyone else has a glicko rating of 1910 with a RD of 182, meaning his strength probably lies between 1728 and 2092. As the ranges overlap we cannot yet be sure Michael is any stronger than Brian M.

Ratings of a player with a high RD change more than those with a low RD after each game.

The RD is a measure of uncertainty in the grade calculated. As Michael has played more games than Brian M, Michael has the lower RD. As we record more games our RD will decrease.

Some online tournaments require a RD of less than 100 before they feel they can use them on a seeding list. This probably needs 20 or so games.

Glicko2 gradings are designed to decrease your RD if you stop playing for a while. This is because we will become less uncertain of the playing strength if the player stops playing for a while. I have not yet found a way to implement this feature on our site.



The source code: https://github.com/sublee/glicko2

A lucid but detailed explanation: http://www.freechess.org/Help/HelpFiles/glicko.html