Coming soon - Get a detailed view of why an account is flagged as spam!
view details

This post has been de-listed

It is no longer included in search results and normal feeds (front page, hot posts, subreddit posts, etc). It remains visible only via the author's post history.

398
A fat tails flaw of the Elo System?
Post Body

All Elo rating systems rely on a fundamental assumption: that odds of winning are transitive. If player A beats B at a rate of 2:1, and player B beats C at a rate of 3:1, then A should beat C at a rate of 6:1. Without some sort of transitive assumption you wouldn't be able to say much about A's probability of beating C if they haven't played before.

I didn't find any study on this so I went to test this assumption and check if the Elo rating system underestimates the chances of an underdog. I used a month of blitz games on Lichess* to get this result:

https://preview.redd.it/7ob6ubtta3x81.png?width=1204&format=png&auto=webp&s=692837e07f5757cce4eda3179d3fb3635f77b030

On both sides of the graph, the underdogs win more often than expected on a consistent basis.

Why does it matter? Well this implies that you will end up with a higher rating than you should if you always play higher rated opponents, and a lower rating that you should if you play lower rated opponents.

Taking the "Actual" win rates above for granted, here is the implied inflation/deflation from playing opponents below/above your rating:

https://preview.redd.it/wdqbp5ewc3x81.png?width=1206&format=png&auto=webp&s=832063932ebd941006da04beeb48244e8f00a8b2

* https://database.lichess.org/ using 2016-07

Comments

What equation was used in the theoretical (orange) curve of the first figure?

Author
Account Strength
100%
Account Age
5 years
Verified Email
Yes
Verified Flair
No
Total Karma
11,874
Link Karma
1,667
Comment Karma
10,177
Profile updated: 20 hours ago
over 9000+

Subreddit

Post Details

We try to extract some basic information from the post title. This is not always successful or accurate, please use your best judgement and compare these values to the post title and body for confirmation.
Posted
2 years ago