Formula for mmr

#2 - Jan. 28, 2014, 8:26 p.m.

We use Glicko2 with very minor alterations. When updating your rating, we adjust based on the average MMR of the other team.

We want to do further alterations and tuning to our MMR system, but not until we deprecate the current leaderboards.

#5 - Jan. 29, 2014, 1:06 a.m.

Harbinger.8637:

Is there anyway to tell or see what your individual MMR is?

No, not currently.

Ryan.9387:

Does mmr change merely from win/loss or your preformace during the game?

Win/loss only. If you have any thoughts on how you would like to see it done, feel free to share.

#7 - Jan. 29, 2014, 1:34 a.m.

kirito.4138:

Since decay is reset/undone basically when LB updates, which MMR is actually used in the formula when you decay? The current fake decayed value or your original.

Original. The decay is exclusive to the leaderboards. It’s applied to the data when we take an hourly snapshot.

#13 - Jan. 29, 2014, 10:27 a.m.

Errant Venture.9371:

Just out of curiosity, what made you guys decide to switch from the the simple and robust Elo system from GW1 to Glicko2 in GW2?

Elo seemed to work very well and was an entirely transparent way to rank + matchmake players.

I wasn’t with ArenaNet when the decision was made, so I can’t say for sure.

My guess would be because of the ratings deviation available with Glicko.

Professor Glickman

The problem with the Elo system that the Glicko system addresses has to do with the reliability of a player’s rating. Suppose two players, both rated 1700, played a tournament game with the 1st player defeating the second. Under the US Chess Federation’s version of the Elo system, the 1st player would gain 16 rating points and the second player would lose 16 points. But suppose that the 1st player had just returned to tournament play after many years, while the second player plays every weekend. In this situation, the 1st player’s rating of 1700 is not a very reliable measure of his strength, while the second player’s rating of 1700 is much more trustworthy. My intuition tells me that (1) the 1st player’s rating should increase by a large amount (more than 16 points) because his rating of 1700 is not believable in the 1st place, and that defeating a player with a fairly precise rating of 1700 is reasonable evidence that his strength is probably much higher than 1700, and (2) the second player’s rating should decrease by a small amount (less than 16 points) because his rating is already precisely measured to be near 1700, and that he loses to a player whose rating cannot be trusted, so that very little information about his own playing strength has been learned.

http://www.glicko.net/glicko/glicko.pdf
http://www.glicko.net/glicko/glicko2.pdf

#18 - Jan. 29, 2014, 11:39 a.m.

One of the ways we plan to change matchmaking, not the MMR system, is to match on more than just MMR. E.g. ranks, ladder tier, etc..

This should improve the quality of matches and should also make MMR more accurate.

#20 - Jan. 29, 2014, 11:44 a.m.

BrunoBRS.5178:

that sounds reasonable and all, but at the same time, it sounds like the best recipe to improve your rating is to stop playing, let it decay, and then come back to improve your rating on leaps, instead of crawling your way up.

at least that’s the impression i got from the quote you highlighted.

Assuming you had reasonable confidence that you would win most of your games, possibly. It’s also a recipe for having your rating drop quickly. Lower confidence also means an increase in the range of players you can be matched against.

The biggest issue we have now seems to be pairing people around the starting rating. Player experience is very diverse at that rating level and it can often leader to poor quality matches when new players are paired with veterans. When we introduce ladders, climbing the ladder would be the best way to improve the quality of opponents and teammates.

#28 - Jan. 29, 2014, 2:06 p.m.

Pixels.6532:

… why, in a team game, do you have individual Leaderboards?

We have Solo Arena leaderboards because people asked for them. I agree though, but think this will be much less an issue if we decide to have a ladder for individual players.

Pixels.6532:

Why not have teams … play as a registered team

That’s a good idea.

Pixels.6532:

… is something like this already in place?

No, but we do plan to make adjustments after the leaderboards are deprecated.

style.6173:

After all, this is a ladder. Outside factors such as your overall PvP level should not be included.

I agree, and we’re taking that into consideration. Unless we start requiring a full roster for rated play, we need the matchmaker to consider more data when creating/merging rosters. When we do use this extra data, we will likely use it more as a preference rather than a requirement.

style.6173:

The changes that need to be made to matchmaking are to ensure games do not start with a 4v5.

Definitely a priority.