Page 1 of 2
K Factor Variations in ELO Ranking ?
Posted: 26 October 2022, 22:05
by Altandor
Hi there!
Does any one know per chance how is it that the K Factor (used in ELO ranking) varies that much?
I mean, most of the times it's 20. I've read that it's supposed to be higher during a certain amount of times you play a game for the first times.
But every once in a while I see something like a 45 K Factor on a player whose score is already high, and I wonder about the reason (and the value of classment at the same time XD).
Re: K Factor Variations in ELO Ranking ?
Posted: 26 October 2022, 23:30
by Meeplelowda
Altandor wrote: ↑26 October 2022, 22:05
But every once in a while I see something like a 45 K Factor on a player whose score is already high, and I wonder about the reason (and the value of classment at the same time XD).
You can get a high ELO in a small number of games, especially if you only play high-rated players. Those two things aren't mutually exclusive.
Re: K Factor Variations in ELO Ranking ?
Posted: 26 October 2022, 23:35
by frogstar_A
It's all explained in the FAQs
Re: K Factor Variations in ELO Ranking ?
Posted: 27 October 2022, 00:33
by Altandor
Thanks for the answers
From the Faqs, depending on the number of games you've played, K factor is supposed to be 60, 40, or 20 ... Nothing said about a 45.
Any clue ?
Re: K Factor Variations in ELO Ranking ?
Posted: 27 October 2022, 03:47
by Meeplelowda
Altandor wrote: ↑27 October 2022, 00:33
Thanks for the answers
From the Faqs, depending on the number of games you've played, K factor is supposed to be 60, 40, or 20 ... Nothing said about a 45.
Any clue ?
Provide a link to the game in question.
Re: K Factor Variations in ELO Ranking ?
Posted: 27 October 2022, 04:32
by ufm
Re: K Factor Variations in ELO Ranking ?
Posted: 27 October 2022, 05:12
by sorryimlikethis
Altandor wrote: ↑27 October 2022, 00:33
Thanks for the answers
From the Faqs, depending on the number of games you've played, K factor is supposed to be 60, 40, or 20 ... Nothing said about a 45.
Any clue ?
Games 1-11 it's 60, games 12-21 it's 40 and 22+ it's 20.
When you play in games with more than 2 people there is another multiplier used to scale how much you win/lose for each player. This multiplier is not shown in the calculations and causes the K Factor to display incorrectly in some cases. But it is being calculated correctly.
I made a small video on it here:
https://youtu.be/on1Qvnuug2c
Re: K Factor Variations in ELO Ranking ?
Posted: 27 October 2022, 12:06
by Altandor
Thanks for the answers
The FAQs is unclear and incomplete indeed, which is why I voted for the report linked, thanks.
I've watched the video which helps to explain reasons behind variations, thanks (thought I might say that it doesn't go to the bottom of calculations and just asks viewer to "rest assured it is calculated correctly" XD -which frankly I believe, but always appreciate full comprehension of this kind of calculation mecanism).
I would suggest to find a way for what appears in the ELO calculations after each game to fit the explanations given in FAQs (or vice versa), as a distorted K Factor can't be good for general trust toward this system
Re: K Factor Variations in ELO Ranking ?
Posted: 27 October 2022, 12:54
by Altandor
Someone asked for a link ... Just went by the "45" factor again ... Here's the link
https://boardgamearena.com/7/itsawonder ... =312198518
Re: K Factor Variations in ELO Ranking ?
Posted: 27 October 2022, 13:03
by sorryimlikethis
Altandor wrote: ↑27 October 2022, 12:06
Thanks for the answers
The FAQs is unclear and incomplete indeed, which is why I voted for the report linked, thanks.
I've watched the video which helps to explain reasons behind variations, thanks (thought I might say that it doesn't go to the bottom of calculations and just asks viewer to "rest assured it is calculated correctly" XD -which frankly I believe, but always appreciate full comprehension of this kind of calculation mecanism).
I would suggest to find a way for what appears in the ELO calculations after each game to fit the explanations given in FAQs (or vice versa), as a distorted K Factor can't be good for general trust toward this system
Right, the whole reason I made the video is because I was annoyed the FAQ wasn't clearer. I was looking at my ELO calculations after a game and seeing K factors of 30, 45, etc and wondering why they weren't higher when they should have been. So, yes, also +1 for them displaying the full calculations and fixing up the FAQ a bit.
The calculation is simply ELO gain/loss from players ((a+b+c) x K factor) x Hidden factor.
But we don't know what the hidden factor is because it changes depending on the game and the player count.
For example in 4 player For Sale it is x0.66 but in 4 player Conspiracy it is x0.33.