Page 1 of 1

ELO calculating bug (?)

Posted: 05 May 2018, 16:09
by Jeffrey Harpen
I lost a game against player with almost the same ELO (#951016) and lost 15 ELO points. Then I won a game against slightly stronger player (#951022) and earned only 13 ELO points. How is it possible? Shouldn't it be +16 in the second game?

Re: ELO calculating bug (?)

Posted: 05 May 2018, 17:43
by pier4r
We cleared this somewhere already in the forum but I don't find it, also on telegram people noticed that.

The two games are:

Code: Select all

2018-05-05 13:42:11 UTC   951022   mrchris (1481)   1460 +13
2018-05-05 13:40:19 UTC   951016   IRRbertPC (1472)   1475 -15


For the elo formula, refer to wiki.gladiabots.com and the FAQ in the manual section.

The point is that each ghost keeps the kfactor of the associated player and the kfator for the result is determined by the player deploying first, normally a ghost.
Mrchris is actually in grandmaster, kfactor 20.
Irrbert is actually in diamond, kfactor 30. This means that with very similar scores, the points exchanged are different.

It is a bit of a bummer, but since it is true for everyone, it shouldn't be that bad. Surely if every ghost would be bind to the kfactor of the league in which it would have been, it would be better but even without that one can survive.

Re: ELO calculating bug (?)

Posted: 05 May 2018, 18:29
by Jeffrey Harpen
I was looking for explanation in the wiki and here in the forum, but didn't found, so thanks for the answer. I was suspecting problem with difference between current ELO of the player and his ghost's ELO, but I see it's the K-factor, not the ELO itself. Now it's clear for me and yes, I can survive, but I think GFX will keep getting questions like this, so it would be reasonable to fix it.

Re: ELO calculating bug (?)

Posted: 05 May 2018, 18:45
by GFX47

Re: ELO calculating bug (?)

Posted: 05 May 2018, 20:04
by pier4r
but... but... I noted it already in March!

Well https://en.wikipedia.org/wiki/Stigler%2 ... of_eponymy

Re: ELO calculating bug (?)

Posted: 05 May 2018, 20:21
by GFX47
Are you sure it was the same bug?

Re: ELO calculating bug (?)

Posted: 05 May 2018, 20:32
by pier4r
yup.

Re: ELO calculating bug (?)

Posted: 05 May 2018, 21:14
by GFX47
Sorry then. I must have misunderstood your report at the time :[

Re: ELO calculating bug (?)

Posted: 08 May 2018, 19:28
by Jeffrey Harpen
pier4r wrote:https://en.wikipedia.org/wiki/Stigler%27s_law_of_eponymy

Don't worry, I don't want it to be called "Harpen's bug". It's all yours! ;)