Elo Formula

[DayanG]

Hi, I was curious to know if the elo DE formula had changed from that of tad / ep. Does anyone know which is the new one or otherwise the old one from tad / ep, I'd like to make a comparison.
PS: Formula Elo both 1v1 and team.

[musketeer925]

I am not sure the formula has been publicized. The DE Elos displayed on this site are just pulled directly from Microsoft's official ladder and not independently calculated, unlike our legacy ladders.

I think the best way to try to figure it out would be to track how much elo you gain/lose and checking the elo of your opponent after a game.

[scarm]

Always was under the impression that the Elo-System is a definite system/way to calculate ranking that can't really be changed, is that not the case?

[musketeer925]

The formula ESOC used was tweaked a bit compared to what the JP Elorating site used to use. I am not sure how strictly that was a true "Elo rating" system or if some variant/derivative.

I am also not sure we know that DE's system is really an Elo system. Need to check, but I think in game it is just called "rating". It's called "ELO" on their stats site, but it could just be an incorrect label.

[musketeer925]

With five data points from me and a friend's games, it looks like DE's system is:

• The DE base formula is exactly the same as our legacy ladder: Awarded Elo = 16.0f - (winnerElo - loserElo) * (32.0f / 400.0f)
• Loser loses the same amount of Elo that the winner gains
• If the formula would give you more than 20 Elo for winning, you just get 20, and the loser loses 20.
• If the formula would give you less than 2 Elo for winning, you get 2, and the loser loses 2.

Conveniently, I happened to get a game with large enough Elo differences to exercise both the boundary rules, and other three games that match our formula exactly makes me fairly confident it's the same.

It looks like our legacy ladder just has a lower limit of 1 Elo instead of 2, and no upper limit, if I am reading our code right.

[DayanG]

In case a player earns 0 Elo this is due to what?
In teams, on the other hand, it is sufficient to add only the elo of the players such as winnerEloTeam1-winnerEloTeam2?

[musketeer925]

As far as I am aware it shouldn't be possible to gain zero Elo after a win, but I could be wrong about something.

For team games, assuming the DE ladder continues to work like our legacy ladder, I think you divide the elo difference by numbers of player per team:

Awarded Elo = 16.0f - ((winnerElo - loserElo) / playersPerTeam) * (32.0f / 400.0f)

You'd need to validate this with some data though to confirm that's the case.

[DayanG]

As far as I am aware it shouldn't be possible to gain zero Elo after a win, but I could be wrong about something.

For team games, assuming the DE ladder continues to work like our legacy ladder, I think you divide the elo difference by numbers of player per team:

Awarded Elo = 16.0f - ((winnerElo - loserElo) / playersPerTeam) * (32.0f / 400.0f)

You'd need to validate this with some data though to confirm that's the case.

Understood thanks.
It is possible that from 0 of elo, it has often happened but it is also right because for example a 2k of elo wins +1700 against a 1300 and 1200 elo (I always speak as a team) it is not right that they lose elo, already matchmaking sucks and neither team enjoys playing such a match.