Guild Wars 2

Has Anet released the formula? Do we know all of the factors that contribute to our rating? I am very curious.

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

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

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.

Is it possible to rate people by MMR score ?

Also, i saw that ANet was really good at collecting data, would it be interesting to have some data about the sPvP past year that we can stare at ?

Justin ODell.9517:

Harbinger.8637:

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

No, not currently.

I would actually like to know slightly more about this. Is it likely that we might be able to have a visible rating at some point?

EoNxBoNx.9213:

Justin ODell.9517:

Harbinger.8637:

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

No, not currently.

I would actually like to know slightly more about this. Is it likely that we might be able to have a visible rating at some point?

Word of warning: having played Heroes of Newerth, publicly visible rating is a bad, bad bad idea – it should be personal and mutual friends only, HoN players would trashtalk the hell out of and blame the lowest MMR player on the team frequently.

Justin ODell.9517:

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

I think this is smart, while the ‘game points’ can indicate the amount of work a player is doing- it’s correlation to skill is less than that of winning the game.

A small suggestion (which might be already implemented) is to lessen the mmr gain/loss of 4v5s by a small amount- unless you are the person that forsakes the team.

Justin ODell.9517:

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.

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.

Justin ODell.9517:

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.

I have been playing a bit of both team and solo queue and I am kind of thinking the rating needs to go beyond win/lose but that the formula will likely be different between the queue type and eventually the game type.
For solo queue, I believe that finding a way to measure skill is needed. My idea for this is tracking kill/death ratio and tracking unique skill utilization including successful dodges and blocks. If a player is running into a 1v2 situation at the far point and always dies, but the team still wins, it is likely they have low skill and I believe their rating should go down. That person is continuously giving an easy 5 points to the opposition. They would have a low kill to death ratio.
For fighting skill, I see some players coming in not knowing how to use their builds and just spam 1 and their heal skill but their timing is off. For example, I play necro and have been using dagger/dagger; if when I use dagger 4 and it transfers conditions from me to the enemy it shows more skill than me just going down through every skill as they come off cooldown. Not sure if Anet can track how successful each skill is, but someone who knows their build and wins the match should get more rank than someone who wins the match and spammed 1-5 off of cooldown and didnt use their elites or utility skills.
For team match, win/lose may be good enough because really what is being measured is the effectiveness of the team rather than the individual. An individual on one team may shine, but on another team drag the team down depending on the team’s build synergy.

Justin ODell.9517:

snip

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.

We’ve asked this question so many times, but why can we not see our MMR? Ranking without being able to see what your current MMR is doesn’t mean anything.

i feel the system is working good with some minor flaws.
i just dont think that two guys, one with 340/170 and the other with 21/1 ratio deserve the same positon on leaderboards.
it feels very odd. to me it is clear that the more games you play the higher you are ranked.
maybe it is working as intended, but as far as we have zero insight, it is not ours to judge.

to me people with less than 100 games shouldn’t be on leaderboards.
same should be considered when (if) ladder/season is going to be implemented

Justin ODell.9517:

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.

Yeah ladders would be nice, although just being able enter your intended role when entering the queue would make a big difference too.

i.e. roamer, back point or bunker

If anyone abused this system they’d shoot down the LB anyway as it would kitten their team.

Justin ODell.9517:

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.

i’m just picturing this as a situation from someone stuck in the so-called “MMR hell”. if you’re getting good games, then you shouldn’t want to game the system to jump ranks faster, but when you’re stuck on a bad MMR (as in, you feel you’re better than the average player on your MMR level, but bad luck/bad games make you keep losing, because you can’t do anything alone), suddenly the idea of just outright stopping and seeing if you can leap out of it in a stroke of luck a month later seems more tempting.

The best improvement would be if more people were playing in pvp.

A real better balance, new game modes and better rewards are the keys.

Justin ODell.9517:

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

This is a great response that I’ve been looking for for months. Thanks Justin!

Justin,

Thanks for the replies in this thread, good stuff.

I’m curious about a couple things and was hoping I could get an answer from you here, but regarding Glicko 2 (or Glicko or Elo), these are all systems designed to rate players in a 1v1 setting.

Individual Leaderboards
GW2 meta has no 1v1 setting. I get using it the way you do, and I think over time it probably does its job, assuming a healthy player base, in getting you into matches with opponents of roughly the same skill, but something that has always confused me is why, in a team game, do you have individual Leaderboards?

I’d rather see a Team Leaderboard that I could click on to see the individual players who have contributed to that team’s ranking. Why not have teams register with an NPC and give them a roster that they can draw from to play as a registered team?

For solo queue, i think at best, because removing the leaderboard there might rustle so many jimmies that all the world would cry out in anguish, you could have a soft leaderboard that showed A ranked players but didn’t assign a 1 or 2 or 3 value to them. Something you could sort by wins, server, losses, win %, etc.

W=1, L=0
Also, have you given any consideration to changing the W/L value in Glicko from 1 = W, 0 = L to something like 1 = W, 1-(Winning team Score – Losing team score)/500 = loss? You could even add a hard base by subtracting a set value from that result so that the losing team couldn’t gain too much rating in a close match (say subtract .25 from that result or take 0 if that puts them below 0).

This would keep the losing team in close matches from getting the same ratings drop they would’ve received had they been completely blown out. Or is something like this already in place?

Anyway, thanks for your time.

Justin ODell.9517:

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.

I’m not sure this is a good idea. After all, this is a ladder. Outside factors such as your overall PvP level should not be included. The ladder should be the only factor for matchmaking. I’ve seen plenty of rank 40 people who gained that through sPvP who are horrible in solo queue. Similarly, I’ve seen rank 10 people who are amazing. Again, the matches needs to be based on the ladder.

The changes that need to be made to matchmaking are to ensure games do not start with a 4v5. There also needs to be faster moves up and down the ladder.

Matchmaking should consider in game rank as well.

Justin ODell.9517:

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.

Will you also do something against the afk/leave aspect? I lose most of my matches cause of this..
Just not really because of the MM, but i really get ALOT of these afkers and leavers…

Please, would be a big plus

I really feel that for solo queue at least starting MMR is one of the biggest fixable problems. Imo, true noobs start out way too high. I know when I started playing even though I was good at killing people I really knew very little about conquest and things like rezzing and stomping.

@Justin. Thanks for responding in here. Glad to see you are actively working on improving the matchmaking.

No matter how you look at it, MMR is an appropriate marker of your skill. Your skill within the boundaries of the game. If your decision making and execution is better than the average player within the boundaries of the game, you win more games than the average player. You increase in MMR.

There is no ELO/MMR/blabla hell.

Justin ODell.9517:

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.

There is a problem with the deviation. Most of the time when I lose a game by a huge margin like 50-500, I check the ranks of the people on my team and I find everyone has about the same leader-board rank, but some if not the rest of them have less than 20 wins.

I also noticed it is much easier to climb leaderboards during high activity hours. I am not sure why, but I often get HUGE losing streaks at night, and then I win it all back during prime time. If the system was working correctly it shouldn’t make someone fluctuate from r200 to 85% on a regular basis.

I know some people who start playing after a long break, and they almost always boost their ranking by a huge amount, sometimes landing in the top 25. However after that they all start on a huge losing streak.

While Glicko is mathematically constructed to be accurate, it just feels like there is something wrong with it or how it is implemented.

If you know that the current system is flawed wouldn’t it be reasonable to tweak it now before ladders come out? I don’t think anyone is using the current leaderboards for any contest and you could test the changes before ladders come out.

Writetyper.1985:

EoNxBoNx.9213:

Justin ODell.9517:

Harbinger.8637:

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

No, not currently.

I would actually like to know slightly more about this. Is it likely that we might be able to have a visible rating at some point?

Word of warning: having played Heroes of Newerth, publicly visible rating is a bad, bad bad idea – it should be personal and mutual friends only, HoN players would trashtalk the hell out of and blame the lowest MMR player on the team frequently.

Sure then it could be personal, but it wouldn’t hurt for me, my self to know my MMR. I want to know this. Not just know that I am x rank on a leaderboard, that has unknown ratings behind it.

Gewd.8125:

If the system was working correctly it shouldn’t make someone fluctuate from r200 to 85% on a regular basis.

There definitely is something very wrong with that. It seems team Q (even as as solo player) gives me far better results than Solo Q. In Solo, usually I’ll either win or lose in steaks of 5-6 games and those games will be absolutely one-sided, like 500 – 200 or so. Whereas in Team Q games seem to be much less frustrating and much closer as well. Not sure exactly why it is such…

Anyway, great to hear about those ladders. Excited to see how things unfold. Personally, I like the idea of registered team, but you need to still be able to swap people in/out, not so much like in WoW (maybe they changed it, only played until Burning Crusade).

EDIT: Agreed it would be nice to see one’s MMR. One could have a Solo Rating and a Team Rating, and then the team’s overall rating be decided by the average rating of all the team’s players. I think that would work well.

The best idea would be removing the soloq ladder
and match people totally random no matter what
and just balance them on classes.

Soloq ladder can never represent anything with four other people included
If you look for a representive ladder you have to take the other four pl in account and rate thm as a team and not individuals,…

Sidenote: if you play with pl not including your team it shouldn t affect your ladder position,this opens for more casual play whr you can also group up with some pve friends to show thm how exiting spvp is without losing 800 ranks on one single matchup.

Gewd.8125:

I also noticed it is much easier to climb leaderboards during high activity hours. I am not sure why, but I often get HUGE losing streaks at night, and then I win it all back during prime time. If the system was working correctly it shouldn’t make someone fluctuate from r200 to 85% on a regular basis.

It makes sense to be honest.

During peak hours more people are playing so the system will find enough people that are relatively close to your own MMR.

During off hours the system has a smaller pool of people to pick from, so it ends up being somewhat off a toss up Throwing people who are top 5 on the leaderboards in the same game as a rank 2 staff elementalist (yes this really happened).

Justin ODell.9517:

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.

Most 4v5’s I’ve noticed are from long queue times and the player akf’s 1 minute into the match and gets kicked which can be fixed with a working ready button. Some 4v5’s are players leaving before a match even starts yet continues to start as a 4v5 which can be fixed by delaying the game from starting. Additionally there are the occasional 4v5’s that result in a disconnect after the match has started which the dishonorable debuff takes care of from discouraging intentional disconnects.

Because there is neither a working ready button or a delay of game the dishonorable debuff is a band aid that does not work which skews a players mmr imo. The downside is it increases queue times.

Locuz.2651:

Gewd.8125:

I also noticed it is much easier to climb leaderboards during high activity hours. I am not sure why, but I often get HUGE losing streaks at night, and then I win it all back during prime time. If the system was working correctly it shouldn’t make someone fluctuate from r200 to 85% on a regular basis.

It makes sense to be honest.

During peak hours more people are playing so the system will find enough people that are relatively close to your own MMR.

During off hours the system has a smaller pool of people to pick from, so it ends up being somewhat off a toss up Throwing people who are top 5 on the leaderboards in the same game as a rank 2 staff elementalist (yes this really happened).

Well what I am saying is it doesn’t make sense for a rating system.

Unless your playing skill has really changed so much in a day, the rating system has failed because it is supposed to measure your skill. The fact you are coming up with a reason not based on personal skill means the rating system has failed.

I don’t know what the exact problem is but I think it is caused by one or both of these:

- The rating volatility is too high and there are more high volatility players during dead hours.
- MMR should not be plainly averaged. A team filled with players far below the standard deviation of everyone else in the game should have a penalty applied to their average MMR.