Warum ist der unterschied in der elo so groß

When reading today's update I noticed something strange.

Paul Inouye Project Update - Jan 21/2014 said:

We've seen numerous reports of players saying they're being mixed with players far outside of their Elo rating. This is odd because 1) no player knows their Elo rating and 2) metrics is proving this to be a misconception. As of yesterday (Jan 20), the average deviation between team Elo ratings was approximately 175 points. The extreme maximum was around 750. What this means is that match maker is finding people of relative skill compared to your personal Elo with a maximum tolerance of 750.

So we've got a 175 point difference. Lets find out what that means.

For this I'm using Matchmaker Phase 3 Update to calculate what the expected outcome would be between two teams that have a 175 point difference.

This means that 175 points is a huge difference. The lower ranked team is only expected to win 26% of the time. So it seems that on average we don't have a balanced match at all. So it's not a misconception at all that people are matched out of their Elo range. A straight up contradiction of what Paul concludes.

FAQ

Why is this bad?

The purpose of the match maker is to provide fair matches. Yet right now on average you'll either have a 26% or a 74% chance of winning. This is not fair at all.

I am still winning 50% of my games

The average difference of 175 points can work against you or in your favour with equal chance. So you'll have an equal number of games that are biased towards and against your team. One game you'll have a 26% chance of winning, the other a 74% chance. This comes down to 50% overall but the individual matches are not balanced.

I don't think Elo match making works (for teams)

I do think it works. And so does Paul. So please assume it works too because even when you assume it works you can see that matches are biased right now.

I don't think Elo accurately predicts match outcomes

You might be right there. Right now the rating system is setup such that a 400 point difference means that the lower players is expected to win 10% of the time. This value is used in chess and may not be correct for MWO. However we've got no data to check that.

edit: Extra sentence for conclusion, expanded quote a bit, added F.A.Q, rewrite intro, added extra F.A.Q.

Summary of conversation

by Hasek10

For those who want to keep track of the conversation here: If we assume that Paul's number was an un-averaged version of "Team" ELOs (More like a Gross ELO difference), then we get an average team ELO difference of approximately 14.5, which puts the probability of a player winning any given match at 47% (well within a reasonable definition of "Fair"). On the other hand, if Paul did not in fact mis-speak, I would like to direct everyone's attention to LukeHauser's and Levitas' excellent statistical analysis in this comment section. As Levitas points out, if the matchmaker is in fact as unfair as we assume it to be with Paul's Number of 175, then it becomes very difficult for players who rely solely on the matchmaker (read: do not confound our statistical analysis with externalities like voice chat and premade blobs of matchmaking entities) to deviate from a strict 50/50 win/loss ratio. Levitas provides evidence that his stats seem to be exceptional if in fact the matchmaker is as biased as Luke claims, but agree's with Luke that not enough information is accessible to truly resolve the situation.

Welche Elo

Wie wird Schach Elo berechnet?

Wie groß ist der Elo?

Wer hat die höchste Elo