Guild Wars 2

Justin ODell.9517:

We’re working on this right now. Hopefully it will be available in the next PvP drop.

One thing to keep in mind is we perform matching based on a Glicko v2 rating, not player rank. So even though higher ranked players also tend to have a higher rating, it isn’t always the case.

@Cheri Wolf
I wrote a simulator to test several different ways of solving the problem, one of which is the way you suggested, so I’m glad we’re on the same page.

While that method improved things it also tends to weigh the match in favor of the first team. The current iteration we’re working on now is very similar, but greatly reduces the bias.

So your saying that if your high up on the leaderboards it doesnt mean you have high mmr? If so why do you guys do it that way? Shouldn’t your leaderboard position be based on your MMR?

Or are you talking about our pvp rank level (wolf, tiger, bear, etc.)?

He was talking about pvp rank level. The context was people complaining that they had rank 20 teammates vs. a team of rank 40’s (which certainly does seem to happen sometimes…) so it’s safe to assume that there is no difference between leaderboard rating and matchmaking rating.

Your probably right, but why does matchmaking always have one team stacked with top 50 players against rank 300 and above. Thought there was a real reason for it. Seems like just they wanted to release something quick for the community because it seems too flawed. Honestly, I just want fair games. You can just scramble the teams between the 10 players based on MMR.

Also isn’t there people who lose games and rank up? SchiTown.7598 was rank 66 and lost a game, which sent that person to rank 2. Stoddles.2350 was rank 56 and lost a game, which sent that person to rank 18. I’m sure there is more.

Kultas lost and went to rank 1 on the EU lb

Beastjk.6531:

Your probably right, but why does matchmaking always have one team stacked with top 50 players against rank 300 and above.

this is why https://forum-en.gw2archive.eu/forum/pvp/pvp/Are-Anet-aware-that-solo-q-is-broken/first#post2923482

Yeah they’re system makes it very difficult to progress or know what to expect with players you go against

Actually, in a completely accurate matching system you would “progress” quite slowly up the ladder. Why? Because you are always matched with teams that give you a tough game, and you only tend to win 50% of the time in that case.

The only way to raise your position in such a ladder would be to play consistently better every match you play. In other words, you would have to win lots of matches in a row against tough opponents. That isn’t likely to happen.

But we all hit a “plateau” where performing better takes a lot of work, and it does not happen fast.

@beastjk
Currently yes you can not play and easily become rank 1 after losing a game if you havent played for a week-month.

The reason for this is as follows:
You stop playing and your mmr stays at lets say 2000 rank 101
Decay beings lets say after 1 week and you lose 10 points (Now remember decay loss is only temporary and will refresh those 10 points when you play another game.)

Now those facts seem to be true. And heres the sad part:

Those top 100 are currently anywhere between 2001-2050 hidden mmr.
So if those 100 players keep playing/fighting over the top spots in reality over the week their mmr is slowly decreasing because they win 1 game and getting lets say 5 mmr points but everytime they lose 1 game they lose 6 mmr points.

So thats why azshene and the Gotmight are rank 1. Since they don’t play and dont lose mmr when decay is only temporary and only need to play 1 game played every 2 weeks or so.

That means its smarted to not play which alot of people are doing but never caught onto the fact quick enough so, Azshene and Gotmight will forever have the highest mmr’s in SoloQ

Also currently the system matches up 5 similarly ranked players who could happen to be rank 1-5 and then matches together then next group of 5 which happen to be rank 100 or beyond.
They said they are currently working on grabbing 10 players and trying the randomization idea or some other fix.

Krayiss.4926:

Those top 100 are currently anywhere between 2001-2050 hidden mmr.
So if those 100 players keep playing/fighting over the top spots in reality over the week their mmr is slowly decreasing because they win 1 game and getting lets say 5 mmr points but everytime they lose 1 game they lose 6 mmr points.

I also believe this is the heart of the problem. Of course, none of us knows the exact formula, but it appears to work this way based on what we can observe. No matter how severe the decay is, this strategy will still work as long as you play only enough matches to avoid decay.

I think if matches themselves fell off of your record and stopped influencing your rating after a couple of months, it would be better.

@NevirSayDie
That sounds like a possible solution with an alteration or 2. Something like playing 10 matches per week to stay on the LBs to refresh decay, cause 1 should not be able to refresh DECAY thats silly.

Right now changing the LBs to show the top 1000 with highest glory rank would be better then having this LB that encourages you to not play. X.X

& I dont think I made it sound as drastic as it is.
You can win 8 games and go up 20 LB ranks and lose 1 and go down 21 LB ranks.
So its more like win 8 games and gain 40 mmr and losing 1 game you can lose 41 mmr

Krayiss.4926:

@NevirSayDie
That sounds like a possible solution with an alteration or 2. Something like playing 10 matches per week to stay on the LBs to refresh decay, cause 1 should not be able to refresh DECAY thats silly.

Right now changing the LBs to show the top 1000 with highest glory rank would be better then having this LB that encourages you to not play. X.X

& I dont think I made it sound as drastic as it is.
You can win 8 games and go up 20 LB ranks and lose 1 and go down 21 LB ranks.
So its more like win 8 games and gain 40 mmr and losing 1 game you can lose 41 mmr

I don’t think Anet will introduce a mandatory number of matches per week to stay on the leaderboard. Wouldn’t it be enough if matches from longer than two months ago just didn’t count? Let’s say you play exactly two matches per day. Every day, two matches would fall off your record, and two new matches would get added to your record.

Even if you only played once a week, your rating wouldn’t stay inflated because all the matches against people with 2200 ratings (ratings that don’t even exist any more) would disappear. If you played 30 matches over two months, those would be the 30 that determined your rank. You’d even be forgiven eventually for losses in your past, like a random 4v5 vs. a really low rating.

Anet should just copy sc2 ledder system, which i think is really good, but requires tons of constant attention and tuniing. Dont know anything about LoL ladder system.

Urejt.5648:

Anet should just copy sc2 ledder system, which i think is really good, but requires tons of constant attention and tuniing. Dont know anything about LoL ladder system.

Yes. I believe starcraft has some kind of bonus to rating based on winning a certain ammount of games over a time period among other things. Something like a bonus pool. That would keep the people that don’t play off the leaderboard eventually. So you have your glicko-rating plus a bonus for playing and winning x games over a time period. Then the decay doesn’t even matter that much because you will get passed by people actually playing the game eventually. Any way you slice it the way they did it was a kluge job and hastily implemented and they have had like half a year to fix it (hidden rating and decay has been around a long time when was the original team leaderboard April?) and we still have no estimate when these fixes will be implemented it should be embarrasing to them but isn’t.