ELO bugged when K factor reaches 20?

[Chamnix318]

When a game is over, you can see a breakdown of your change in ELO based on how you did against each opponent. Normally your total change in ELO is the sum of the changes from each opponent, but sometimes it is not. In particular, when your K factor is 20, the total change is often not the sum of the changes. Is there a bug in this calculation?

[sourisdudesert]

Hi,

There is no rules that the sum of all ELO changes is zero on BGA.

Especially, if there are different K factor, this is not true.

You can check this for more details:
https://boardgamearena.com/#!faq?anchor=faq_account_elo

[Chamnix318]

Sorry, I guess I wasn't clear. I'm not looking at anyone else's change in ELO, just mine. I am looking at the Details of ELO Computation for me. It says, for example (with made up numbers):

Victory over Player X
Probability of win: 25% 5.22

Victory over Player Y
Probability of win 35% 7.48

Total ELO variation: 12.70

In that example, the Total ELO variation equals the sum of 5.22 + 7.48. However, that is not always the case. That is what I am questioning.

Thanks for the quick reply.

[sourisdudesert]

The best would be to give a link to a table where the total sum is incorrect so I can check.

[Chamnix318]

https://en.boardgamearena.com/#!table?table=47464370

[Chamnix318]

https://en.boardgamearena.com/#!table?table=47641629

[Chamnix318]

https://en.boardgamearena.com/#!table?table=48128837

ELO rank before this game: 253.01

K factor: 20

Victory against Lunalol: 6.4
Victory against sesselfussel: 6.15
Victory against ysfx: 10.17

Total ELO variation: 7.58????

ELO rank after this game: 260.59

[eoc]

The three victories individually would sum up to a gain of 22.72 points. What gets displayed instead as "total variation" is a third of that sum (22.72 / 3 = 7.57), which probably makes sense (using your average over all players you played against) but indeed has a somewhat confusing label attached to it.

[Chamnix318]

Thank you for looking at that. That would make sense, but this game is also a four player game, and the total ELO variation is equal to the sum of the individual results without dividing by 3.

https://en.boardgamearena.com/#!table?table=47061317

Edit: Actually, it was a five-player game, but it still isn't divided by anything.

[qwertasdfg]

The three victories individually would sum up to a gain of 22.72 points. What gets displayed instead as "total variation" is a third of that sum (22.72 / 3 = 7.57), which probably makes sense (using your average over all players you played against) but indeed has a somewhat confusing label attached to it.

nice theory, but even for k=20 that's not it, either: take this one, f.e. https://en.boardgamearena.com/#!table?table=48550722