Talk:Elo rating system

Elo in Snooker??

In the introduction it is claimed that the elo rating system is used as a rating system in snooker (amongst others). I have removed that until someone can provide a source for that. For me it's just plain wrong. The Snooker World Rankings are based on the prize money won within the last 2 years ("Rankings FAQ - WPBSA".), i.e. where your rating is based on how far you've come in tournaments, and not on the outcome of individual games and the rating of the opponent.

"Suggested modification" section is misleading

I believe the section Suggested modification, proposing 480 as a better rating difference divisor, is misleading. Any number should leave the underlying mathematics the same, resulting just in rescaled ratings. It is my understanding that the issue in the referenced article by Jeff Sonas arises only in the presence of the 400 points maximum difference cutoff, as is presented in the article. This cutoff is not integral to the Elo rating, it is not discussed elsewhere in the Wiki article and without it the whole system should just converge to slightly less spread values on its own.

If I am correct, the whole paragraph can be removed without loss. If I am wrong, perhaps it can be enhanced to prevent future readers from following my line of reasoning. — Preceding unsigned comment added by 217.28.83.193 (talk) 17:28, 6 September 2024 (UTC)

Since that section was added, I have been wondering what happened to Sonas's ideas since his article was published, in 2011. If FIDE responded to his logic by either taking his suggestion, or by making some other change, then we should acknowledge that, citing some appropriate source. Otherwise, the whole discussion is something like WP:OR from our point of view, and doesn't belong in an encyclopedia. I don't know where to look to find out about changes to FIDE's rating calculations over the years. Bruce leverett (talk) 18:37, 6 September 2024 (UTC)

Agreed, mathematically this makes no sense, I removed the section. See section A3 (in the appendix) in the paper "Measuring skill and chance in games" as published in the European Economic review (2020) by Duersch, Lambrecht and Oeschssler, for a formal mathematical proof. 87.120.102.25 (talk) 21:28, 2 September 2025 (UTC)

Use Outside of Chess - Other board and card games

I'm surprised that Board Game Arena's use of (some variant of) this system isn't mentioned. Maybe someone could add it, if that's useful? 50.227.112.182 (talk) 18:42, 6 November 2024 (UTC)

Game playing sites are here today, gone tomorrow. Look at Wikipedia's definition of reliable source, then see if you can find significant coverage of BGA or its use of Elo ratings in such a source. Bruce leverett (talk) 20:38, 6 November 2024 (UTC)

Elo (normal distribution) vs Zermelo (logistic) Rating system in the "Theory" Section

It is my understanding that the mathematical model described in the "Mathematical Details" section is not the model actually proposed by Elo. It is instead the model which is actually used (which is the model written about first by Zermelo). This is hinted at lower down in "Most accurate distribution model," but it isn't actually said that this is the model being described. You'd have to know that the described model is a logistic distribution model, which is not obvious to those not familiar with the subject. Shouldn't this be said outright and before diving into the math? As it is, readers are mislead by what system is being described. I do agree that the logistic distribution is the one worth describing, as it is what is used today.

Another, larger, improvement that I think would be possible would be to go as in depth in Elo's system, as that's literally the name of the article and some readers may actually be looking for his system, rather than the one that replaced it, even if the common name doesn't refer to it anymore.

This is a more major of a change to the article structure and content than I have the time (or experience) to do at all right now. I would also like to know if I'm making an obvious misconception here before I put any large amount of time into it. Nebman227 (talk) 13:25, 23 December 2024 (UTC)

Yes, if you think there should be clarifying discussion about the choice of distribution, especially before we get to the math, then I recommend that you make the edits that you think won't be controversial. If some edits might be controversial then you could discuss the details here. But, no, I don't think that the general idea is controversial; go for it.

I'd like the article's emphasis to be on what the name means today. However, the historical meaning is also plenty relevant ... perhaps in a section that is named "History" or something that conveys that idea. —Quantling (talk | contribs) 14:28, 23 December 2024 (UTC)

I'm new to editing and have only done minor copyedits before, so I'm coming at this with an abundance of caution - I'll plan on giving this a look and checking/finding sources when I'm free from the holidays. Nebman227 (talk) 18:26, 23 December 2024 (UTC)

Indeed, a brief concise summary of Chapter 8 Rating System Theory in (Elo, 1986) would be very helpful.

KP (talk) 10:50, 7 January 2026 (UTC)

Short description

@MaxBrowne2: Short descriptions should be short, ideally < 42 characters long. They are not intended to be *complete* descriptions, but just enough to let a user know whether an article is in the right general area. The current description is unnecessarily long. My shorter version can surely be improved upon. Ideas? See WP:SHORTDESC --Macrakis (talk) 14:32, 25 September 2025 (UTC)

The test for sufficient shortness is as follows: working on a desktop computer (rather than a mobile phone), start typing the name of the article in the Wikipedia search box. A list of article titles will appear, each followed by the short description of the article. If the short description of the article you are interested in gets truncated, it's too long. (If you try this test on a mobile phone, even shorter descriptions get truncated.)

The present short description for this article passes the test. I'm not saying it can't be improved upon, but it doesn't have to be shortened just to fulfill its purpose. Bruce leverett (talk) 15:16, 25 September 2025 (UTC)

I like the latest version, "System for rating game players". That said, since "system" and "rating" are already part of the title, I wonder how we can take advantage of that, or somehow include chess (which is the original application ... but I don't know if it's still the most common). --Macrakis (talk) 10:02, 27 September 2025 (UTC)

Formal derivation for win/draw/loss games

Is this chapter relevant to the Elo rating system? In the Elo system, a draw is treated as half a win and half a loss. The expectation function predicts the result over a number of games, a tournament, a match, games played in a certain period, not an individual game result. See, for example, Glickman 1995, p. 32, Probability of a Drawn Game. KP (talk) 17:22, 10 January 2026 (UTC)