Anet Please restrict Max MMR indifference in your MM algorithm.

I know this will fall on deaf ears but seriously.Bad enough I was over valued and placed in near Plat 2. Having a nice steady win rate of 2/10, 3/10 to put me back were I'm normally in (Mid gold) Its fine, I'm used to losing in droves. But being placing against with players that are 300 rank points above me every time does not make for a healthy match or PVP mode!

A good example is my last match, I'm currently rated at 1400. Matched up against Vallun, rank 12 with a rating of 1700... The skill indifference in unfathomable... I just becomes a matter of which Plat 3 can carry harder. I'm certainly contributing nothing to team fights, let alone trying to side node against a person 1 division above me...I pretty much ready to quit cold turkey on spvp :/

I know some of us just wants a even match, not a train wreck of 7-8 blowouts per 10 matches.

I definitely agree with that for the sake of everyone involved. It needs to be restricted in DuoQ as well. Nothing stopping an actual silver player from queing up with a top 25 legend for the sole purpose of breaking the matchmaking to be worse than it already is.

Here's the dilemma:At off-hours, population near the extremes is low. Do you force those players to sit in queue forever by capping the maximum skill difference, or eventually let them play games?

Also, a rating difference of 300 isn't that large. 200 is about normal; 300 for off-hours isn't unreasonable. The problem is that although the rating difference between the top 50 or so players and the rest of the population isn't that large, their skill difference is huge. This is mostly due to all the power creep making player skill largely irrelevant except for the very top.

Imagine getting a msg like this in off-hours:

"Sry but we currently are unable to find a match for you. Please try again later".lmao

https://wiki.guildwars2.com/wiki/PvP_Matchmaking_Algorithm

looks like they changed it to not include (or reworded it to something tricky) rating difference. last time I checked it was 200-300 max.

Another thing to consider is a 300 skill difference is actually larger then it seems. Those at top get single digit rank gains for each win, and can lose 50 points for a loss. The 1400 skill player gains 20+ for a win and only loses 10-15 or so for a loss. It tends to keep everyone around a 1400-1600 skill level.

And as someone mentioned people just duo que up with their friend who tanked an account to 1200 ranking to fix the matchmaking system. Two 1700 skill players, but one on a tanked account of 1200 plus the 1700 legit account gives an average rating of 1450 for the duo team. Unless they get paired against another duo doing exactly the same thing they will have a huge advantage. Two 1700 skilled players will just murder two 1450 rating players.

@DAN.7314 said:And as someone mentioned people just duo que up with their friend who tanked an account to 1200 ranking to fix the matchmaking system.There are a few ways to fix this:

• Only allow group queuing within a certain rating differential. May have to restrict it to platinum+ though as to not alienate more casual players
• When a player DCs or throws, everyone who queued with that player gets a penalty. This prevents an alt account who group queues from invalidating a game if it goes south.
• Match players at the higher of the pair's rating rather than average. Or use a weighted average to move it closer to the top player's rating.

@Exedore.6320 said:Here's the dilemma:At off-hours, population near the extremes is low. Do you force those players to sit in queue forever by capping the maximum skill difference, or eventually let them play games?

Also, a rating difference of 300 isn't that large. 200 is about normal; 300 for off-hours isn't unreasonable. The problem is that although the rating difference between the top 50 or so players and the rest of the population isn't that large, their skill difference is huge. This is mostly due to all the power creep making player skill largely irrelevant except for the very top.

Yes. Queue times need to be much longer. I played an almost-dying MMO in the past and didn't mind waiting 20+ minutes for a game nor did the remaining players mind.

@Exedore.6320 said:There are a few ways to fix this:

• Only allow group queuing within a certain rating differential. May have to restrict it to platinum+ though as to not alienate more casual players
• When a player DCs or throws, everyone who queued with that player gets a penalty. This prevents an alt account who group queues from invalidating a game if it goes south.
• Match players at the higher of the pair's rating rather than average. Or use a weighted average to move it closer to the top player's rating.that looks pretty good to me!

after the first year in this game i didnt play much spvp, however recently for building guildhalls were you need those daily potions i did go into there again and to speed it up i did so with multiple accounts in a row. now aside from my main account the others did not have pvp games on them. on two of them before reaching the rank to be able to play ranked, i was matched against a duo with the titles 'Best of the Best' + 'Unyielding Legend'. i didnt think this was possible to get matched against such experienced players on an account with pretty much no spvp matches. if the ranked match maker is as bad..then thats one more reason for me to stay away from the mode, aside from being rated for teamperformance in a solo/duo queue.

@DAN.7314 said:

And as someone mentioned people just duo que up with their friend who tanked an account to 1200 ranking to fix the matchmaking system. Two 1700 skill players, but one on a tanked account of 1200 plus the 1700 legit account gives an average rating of 1450 for the duo team. Unless they get paired against another duo doing exactly the same thing they will have a huge advantage. Two 1700 skilled players will just murder two 1450 rating players.

Except that is never how it worked.If you duoQ with anybody your duo's average rating is automatically equal to that of the higher player.Always been like that. To avoid exactly what you are describing.

@Zenix.6198 said:

@DAN.7314 said:

And as someone mentioned people just duo que up with their friend who tanked an account to 1200 ranking to fix the matchmaking system. Two 1700 skill players, but one on a tanked account of 1200 plus the 1700 legit account gives an average rating of 1450 for the duo team. Unless they get paired against another duo doing exactly the same thing they will have a huge advantage. Two 1700 skilled players will just murder two 1450 rating players.

Dan.7314 is correct; Zenix.6198 is wrong. It's easy to see by looking at rating gains from wins as a disparately ranked duo compared to each player solo. The higher ranked player in the duo earns very few points, but the lower ranked player earns more than if he had queued alone.

ANet for a time matched at the rating of the highest rated player in a group, but that was abandoned years ago.

@Exedore.6320 said:

@Zenix.6198 said:

@"DAN.7314" said:

And as someone mentioned people just duo que up with their friend who tanked an account to 1200 ranking to fix the matchmaking system. Two 1700 skill players, but one on a tanked account of 1200 plus the 1700 legit account gives an average rating of 1450 for the duo team. Unless they get paired against another duo doing exactly the same thing they will have a huge advantage. Two 1700 skilled players will just murder two 1450 rating players.

Yes, rating gain/loss obviously is handled separately and in consideration of respective ranking.

Matchmaking however, where the system pairs teams based on average rating, still uses the rating of the higher rated player as the "average value" for duos.

EDIT: can we get some DEV comment on this, please. @Ben Phongluangtham.1065

Last season during prime time (sunday around 5pm?) in the last week I had a game where two enemy players were below 1400 while I was aroung 1700. I asked in mapchat because... it was kind of obvious.

Regarding duo: I didn't find anything regarding that "average rating of dues" topic except:

Filter/Power/@percent The percent a roster's rating is inflated due to the number of players in the roster.

and

Power curve="1" percent="1"/

(Source: https://wiki.guildwars2.com/wiki/PvP_Matchmaking_Algorithm)

Since there are other factors - curve for example - where "1" means a neutral factor, I think it is not actually a percent but a factor. Can't know for sure though.

Clarification would be welcome, but we had lots of discussions on the MM here and never really got an official answer on more complicated matters.

I find match quality has been very bad.

@"Zenix.6198" said:Yes, rating gain/loss obviously is handled separately and in consideration of respective ranking.

Matchmaking however, where the system pairs teams based on average rating, still uses the rating of the higher rated player as the "average value" for duos.Do math, and you'll see that I'm correct.

• Rating adjustment is done as player vs. composite team. That is, your rating gain or loss is based on your personal rating compared to the composite (i.e. average) rating of the enemy team. Glicko2 has additional factors which adjust the magnitude of that rating gain or loss. Note: player vs. each enemy player rating adjustment ends in the same outcome; only composite team vs. composite team would be different. However, we know that it's not composite team vs. composite team because the adjustment of each rating-stable (i.e. they played a large number of games) player on a team can be a significantly different amount.
• A duo of rating-stable players with very disparate ratings have significantly different gains and losses. As I and many others have, if you duo queue with a friend who is well below (or well above your rating), you can look at your rating change after a game. In a win, the higher rated player earns much less rating than the lower rated player. In a loss, the higher rated player loses a lot and the lower rated player loses very little.
• Rating gain/loss as a solo queue is significantly different than in the previous bullet. If one of the two players' ratings was chosen as the matching pivot (as opposed to average), then in the duo queue example, one player would gain and lose approximately the same amount of rating as if they had solo queued. This is certainly not the case based on a few user samples. A typical gain/loss is 15-20 points for a rating-stable player in a well-matched game. In a disparate duo queue, the higher player will only gain <10 while the lower player gains >30.

a) The system tries to use ''some'' players(1-2) that have waited for more than 6-7 min (because they have high mmr).

b) The system then suffles the players , so the average TOTAL TEAM mmr (winning chance) is the same .So the first team will consist of : 1600mmr player+ 1400+1300 = total 3300 TEAM MMR .While the second team : 1500 + 1450 + 1350 = 3300 Team MMR .Its 50-505 chance who would win .

c) rating gained =/= OVERALL team rating .If you Team up with a low mmr player , he will get the highest party's members MMR .Team 1 : 1600 MMR (party leader) + 1600 (party member - his actually mmr is 1200 , but got boosted because of Leader) + the system will assign you a low (1300) MMR playerWhile Team B = 1600 + 1500 + 1500

d) If you get teamup with higher mmr player , it doesnt mean you will gain the same ratings points ..Its a ''fake'' Rating , just like you use a lvl 3 PvE character and get transformed into 80 in the PvP zone .It doesnt mean you have an actual 80 character , nor you are an actually Platimun player with only 1200 ratings in your '' Rating Panel''

I would say 100 rating difference is already very large and can become very one-sided.And 200 difference is just insane. It is equal to a 1600(plat 1-2) vs 1400(gold 2-3) matchup.Larger difference than that is really unfair to both sides.

I am don't care about max mmr or something ... for me it ok.If i want play spvp I take pica, lemon juice, and press button to join rated. And worst thing just wait and not be connected.

@Exedore.6320 said:

@"Zenix.6198" said:Yes, rating gain/loss obviously is handled separately and in consideration of respective ranking.

Matchmaking however, where the system pairs teams based on average rating, still uses the rating of the higher rated player as the "average value" for duos.Do math, and you'll see that I'm correct.

• Rating adjustment is done as player vs. composite team.
  That is, your rating gain or loss is based on your personal rating compared to the composite (i.e. average) rating of the enemy team. Glicko2 has additional factors which adjust the magnitude of that rating gain or loss. Note: player vs. each enemy player rating adjustment ends in the same outcome; only composite team vs. composite team would be different. However, we know that it's not composite team vs. composite team because the adjustment of each rating-stable (i.e. they played a large number of games) player on a team can be a significantly different amount.
• A duo of rating-stable players with very disparate ratings have significantly different gains and losses.
  As I and many others have, if you duo queue with a friend who is well below (or well above your rating), you can look at your rating change after a game. In a win, the higher rated player earns much less rating than the lower rated player. In a loss, the higher rated player loses a lot and the lower rated player loses very little.
• Rating gain/loss as a solo queue is significantly different than in the previous bullet.
  If one of the two players' ratings was chosen as the matching pivot (as opposed to average), then in the duo queue example, one player would gain and lose approximately the same amount of rating as if they had solo queued. This is certainly not the case based on a few user samples. A typical gain/loss is 15-20 points for a rating-stable player in a well-matched game. In a disparate duo queue, the higher player will only gain <10 while the lower player gains >30.

What does this exactly have to do with "math"? And how does this (very common knowledge) prove anything?As I already said above "Yes, rating gain/loss obviously is handled separately and in consideration of respective ranking".So of course a higher rated player gets less rank from a win than a lower rated player even though they are on the same team.

What you are missing entirely, is the fact that the matchmaking-system =/= rating-system.

@Zenix.6198 said:What you are missing entirely, is the fact that the matchmaking-system =/= rating-system.

What if the lower rated account in question was an alt created solely with the intent of tanking ranked games? Wouldn't their MMR be fairly close to their actual rating if that's all they did?

@Multicolorhipster.9751 said:

@Zenix.6198 said:What you are missing entirely, is the fact that the matchmaking-system =/= rating-system.

I dont quite get what you mean by that. Please clarify...maybe im just misunderstanding you.

But generally speaking:

a) Duo-Qs DON'T have an average rating but simply use the one of the higher player for both members for matchmaking purposes.Any other system would be totally exploitables. Like a 1.8k player Qing with a 1.1k player (=which is an alt account from an actual 1.8k player as well) to easily farm 1.4k players....WHICH IS NOT A THING.

b) rating gain/loss is calculated on an INDIVIDUAL basis.Where basically each player individually gains/loses rating in comparison to the Enemy's Team average rating.So if Team A has an average rating of 1.5k, a 1.6k player from Team B would get less from a win than a 1.4k player from team B. And vice versa for a loss obviously.That's why you see players in very high brackets generally getting lower rating for a win (<10)while players in lower brackets can get >20 rating for a win.

@Zenix.6198 said:

@Multicolorhipster.9751 said:

@Zenix.6198 said:What you are missing entirely, is the fact that the matchmaking-system =/= rating-system.

What I mean is; if someone created an account and only played ranked and only did so to throw games and joke around with the intention of dropping rank, having only played ranked their hidden MatchMaking Rating should be close to their Ranked Rating which would mean Matchmaking-system could potentially = rating-system.

And, I mean; I wasn't trying to accuse anyone there, but I find that quote marked "a)" a little hard to believe considering back in Season 9 when DuoQ was first removed, it was stated by Ben that the decision to do was made to improve matchmaking at higher ranks, and because of the difficulty in matching people of great ranked disparity together. If you aren't convinced, go Duo up with someone leagues below you and ask around in map chat for people's rating(Or add them to check, for science. Yeah science!) and you can visibly see the imbalance a high-disparity Duo creates. Duos most definitely have a sort of average rating, and even if they didn't it could still be exploited. The real solution is similar to what the OP suggests in introducing caps on how far apart DuoQs can Q together or by simply not matching people 200-300 even 400 points apart ever. Like, ever.

@Multicolorhipster.9751 said:

@Zenix.6198 said:

@Multicolorhipster.9751 said:

@Zenix.6198 said:What you are missing entirely, is the fact that the matchmaking-system =/= rating-system.

Oh ye, I remember that.The thing why it was an issue for matchmaking was a bit more intricate.Like when The Rank1 AND Rank2 player DuoQ together (maybe even at off hours) the matchmaker is kinda screwed, since those two have to be on the same team.If they couldn't Q together, they would just be split and the game would be a lot fairer. But when your potential player pool already is small (like in off-hours), there might be issues with actually creating a match where both team's average ratings are close-ish.

Edit:

Like for the sake of argument.Lets assume at a given time a maximum of 10 players are online.You have two people with 2100 rating and 8 with 1500 rating.Now imagine having the two 2.1k rating players in the same team ....big yikes.

@Zenix.6198 said:Like for the sake of argument.Lets assume at a given time a maximum of 10 players are online.You have two people with 2100 rating and 8 with 1500 rating.Now imagine having the two 2.1k rating players in the same team ....big yikes.

Big yikes indeed! I know that still happened sometimes in SoloQ, but at least back then it was just a chance. Now it's pretty much a guarantee with the option for them to DuoQ together. And like you say, the lower the population; the easier that is to exploit there.

They could go ahead and implement the cap between ratings like the OP suggests, but personally I think they'd accomplish the same thing by removing DuoQ entirely and splitting ranked into SoloQ and TeamQ and even maybe possibly making 2v2 a permanent arena. Well said though, props.