Talk:Glicko rating system

This article was nominated for deletion on 12 July 2019. The result of the discussion was keep.

If you are developing code based on this...

Base it off of the glicko-2 paper that is linked in the references. I was attempting to do it off of the algo specified in the wiki article, but a lot of important details are left out. Also glicko two has lots of improvements. — Preceding unsigned comment added by 134.134.137.71 (talk) 18:55, 11 April 2012 (UTC) More detail would be appreciated, about these important details left out, and the improvements contained in Glicko 2. How much is "a lot" ?? Why is Glicko 2 worth the extra effort? Thanks for your input. --Twixter (talk) 18:58, 25 April 2019 (UTC)[reply]

Yes, thanks for mentioning this. the article is very brief and has left important parts out. i personally also struggeled with this until i found the paper 240F:79:57A:1:582F:1D39:8023:2850 (talk) 06:22, 30 July 2023 (UTC)[reply]

Good message. I do t know why this wasn’t highlighted in the article Fewcada (talk) 11:58, 30 July 2023 (UTC)[reply]

Additional sources

Some relevant links [1] [2] SunCreator (talk) 23:40, 22 December 2009 (UTC)[reply]

Public domain

The article states that "Both Glicko and Glicko-2 rating systems are under public domain". Is this a meaningful statement? Glicko is just a series of mathematical formulas, and mathematical formulas are not subject to copyright. JakubMarian (talk) 21:39, 23 September 2014 (UTC)[reply]

Dr. Mark Glickman patented Glicko-2 and attempted to license its use. After that did not work out he moved it to the public domain. Source: Dr Mark Glickman on the Perpetual Chess Podcast, EP 249, ~minute 34. https://www.perpetualchesspod.com/new-blog/2021/10/26/episode-249-dr-mark-glickman 4.31.102.58 (talk) 22:20, 7 June 2022 (UTC)[reply]

About the value of t

In the first step, if a player played in the previous rating period, should t be 0 or should it be 1? --Twixter (talk) 18:58, 25 April 2019 (UTC)[reply]

Evidently it's 1 because if it were zero, RD would not change and "If several games have occurred within one rating period, the method treats them as having happened simultaneously." Earnstein (talk) 19:30, 17 March 2022 (UTC)[reply]

Introducing a lower limit for RD in step 3?

Quoting from Glickman's pdf, "One practical problem with the Glicko system is that when a player competes very frequently,his/her rating stops changing appreciably which reflects that the RD is very small. This may sometimes prevent a player’s rating from changing substantially when the player is truly improving. I would therefore recommend that an RD never drop below a threshold value,such as 30, so that ratings can change appreciably even in a relatively short time."

This being the case, perhaps the formula in step 3 should be

$RD'=\max \left(30,{\sqrt {\left({\frac {1}{RD^{2}}}+{\frac {1}{d^{2}}}\right)^{-1}}}\right)$