What is the formula used for elo change?

[please join me]

My ELO in Trajan is 2073 and I just defeated someone with an ELO of 1617, but didn't gain any ELO points. I have nothing to gain, but lots to lose? How do I know exactly how low to set my bracket minimum when setting up a new game in the waiting room, as to make sure I will gain at least some points if I win?

[covadis]

Your game was probably friendly (for at least one player )
About ELO you May do a search for an answer

[please join me]

What do you mean by "friendly"?

[Josquin]

That's a premium option: A game can be created without counting for the ranking.

[jouons]

You can set a minimum ELO to play with good players (and/or a maximum one).
You don't need to be premium to do that.

I do that for Agricola I'm so bad I limit to <1400

[peyo_fr]

My ELO in Trajan is 2073 and I just defeated someone with an ELO of 1617, but didn't gain any ELO points. I have nothing to gain, but lots to lose? How do I know exactly how low to set my bracket minimum when setting up a new game in the waiting room, as to make sure I will gain at least some points if I win?

I'm not surprised since I think that, with an Elo gap of +456 (2073 - 1617), you are likely to win 1 point at best for any victory, and most probably 0 through the rounding algorithm...

In a 5-years old dedicated post (the title can be translated as "Modification in the Elo calculation"), unfortunately written in French, Diplojak writes (excerpts from 3 different coms) :

... C'est un calcul elo classique, le même qui était utilisé aux échecs en France avant le calcul FIDE. C'est juste un peu long à détailler. ...
... [Un joueur d'Elo 1500 bat un joueur d'Elo 1400] : en gros, pour modifier l'Elo pour une partie à 2, on donne a l'Elo actuel un poids 15, et à la perf[ormance] [*] liée à la victoire (qui vaut dans cet exemple quelque chose comme 1800) un poids de 1 : ((1500 x 15) + (1800 x 1)) / 16 = 1519 ...
... Les points gagnés, pondération ou pas, x joueurs ou pas, tendent vers 0 si l'écart [**] se rapproche de 400 points...

Translation (hopefully correct) :
... It's a classical Elo calculation, the same that was used in France for chess tournaments (by FFE, French Chess Federation) before the FIDE formula (= WCF, World Chess Federation). That's just a bit long to be detailed here ...
... A player ranked Elo 1500 beats a player ranked 1400 : roughly, to modify the Elo in a 2-players game, you give the current Elo a weight of 15, and to the gain [*] linked to victory (evaluated to something like 1800 in this example) a weight of 1 : ((1500 x 15) + (1800 x 1)) / 16 = 1519 ...
... the points you win, using a ponderation or not, and whichever the number of players is, tends towards 0 when the gap [**] gets closer to 400 points ...

[*] FR: performance / EN: gain : my understanding about this "gain" value is roughly your opponent's Elo + 400 in this case : 1400 + 400 =1800.
[**] FR: écart / EN: gap : my understanding is that this "gap" refers to the difference between both players Elo's : thus reaching 400 (and even worse exceeding 400) means you have little chance --if any-- of winning 1 single Elo point...

Conclusion: YES, in that kind of game, you have nothing to win but lots to lose...


Additional links :

• Link to FIDE Elo rating system
• Link to the FFE French national Elo rating system (the one Diplojak refers to ???)

[frederic.moulis]

That's a premium option: A game can be created without counting for the ranking.

It may happen if someone is replaced in a game as well : he will not count for ELO variations. si if in a 2p game, one is a replacement, the other will not count for ELO because he has no ELO opponent.

[peyo_fr]

That's a premium option: A game can be created without counting for the ranking.

It may happen if someone is replaced in a game as well : he will not count for ELO variations. si if in a 2p game, one is a replacement, the other will not count for ELO because he has no ELO opponent.

Sorry, but with an Elo gap >>+400, I don't think you need to be in a "friendly game" to win 0 Elo points in case of victory (see my com above)

[covadis]

That's a premium option: A game can be created without counting for the ranking.

It may happen if someone is replaced in a game as well : he will not count for ELO variations. si if in a 2p game, one is a replacement, the other will not count for ELO because he has no ELO opponent.

Sorry, but with an Elo gap >>+400, I don't think you need to be in a "friendly game" to win 0 Elo points in case of victory (see my com above)

Against Mat76 I win 2 ELO at each victory at Rallyman (around 400 ELO diff)
I think, that the minimum is 1, not 0

[peyo_fr]

I think, that the minimum is 1, not 0

Definitely NO ! I may send you back multiple mails I received from BaJ about games I won with an Elo gain of 0...
You say "around 400 ELO diff", the devil is in the details of this "around" ...

[lynkowsky]

Many plays vs bajbot at diaballik, to test : 447 difference -> +2 Elos ; 448 difference -> +1 Elo. And I do not think that in a 2 players game one can win 0 Elo...