PvP Discussion: Matchmaking and Leagues

[Exedore.6320]

> @Ithilwen.1529 said:

 

> You can't get good conclusions starting with an unsound premise. Here you have three strong reasons that team average rating is not sound. **The only effective way to get even matches would be to lock classes, force solo queue and set up matches on class based ratings. You would also need to match player-by-player. Team average would throw You off badly.**

Have fun getting two thieves on your team and not being able to do anything about it.

 

[Ithilwen.1529]

> @Exedore.6320 said:

> > @Ithilwen.1529 said:

>

> > You can't get good conclusions starting with an unsound premise. Here you have three strong reasons that team average rating is not sound. **The only effective way to get even matches would be to lock classes, force solo queue and set up matches on class based ratings. You would also need to match player-by-player. Team average would throw You off badly.**

> Have fun getting two thieves on your team and not being able to do anything about it.

>

 

If it were under the circumstances I have described, I'm reasonably confident it would be a good match.

[ReaverKane.7598]

Why not Ranked Team/Flex Queues? Single/duo is clearly not working, or at least not resulting in any improvement. You still see the exact same problems you did before queue restrictions, with the added loss of people that play less or nothing at all because they only played with their friends. (and Unranked isn't the answer until you equate Unranked and ranked rewards, and tournaments don't work... Haven't been able to register ONCE).

Not only that, but team/flex queues will lower the toxic practices of losing on purpose so they can continue duo playing.

 

I do feel that matchmaking is skewed a lot of the time, maybe due to the play hours. So why not give some sort of indication on the PvP menu that there's not enough people queueing to give you a proper matchup. (like Status:O, with the O changing color based on the average deviation of rank in the last X matchups)

 

I'd say that Leagues need to be LONGER, much longer. The current state of leagues is reminiscent of Living World Season 1. We're being rushed into progressing, which ends up discouraging people that don't have the time availability.

 

I wouldn't mind having rank points being attributed more closely to how the match progressed. I mean it sux to lose 13 points on a 500-490 match. And i wouldn't really mind not winning 13 points (which you generally don't) if the match was close.

Also, why no rank point rewards for winning streaks? It's really discouraging to win 7-10 matches in a row, and looking at your match history seeing points go lower instead of higher.

 

PS: Will it ever be possible to click on a match in our history and see the details? (Like score and the new graphic)

Also, it would be awesome if we could see the difference between our stat and top stat on your team (like show our score and then _- xxxx_ in red with the difference)

 

EDIT: Was going to add this to the rewards thread, but after reading it, i think it makes more sense here.

Like i mentioned earlier, there's way too much incentive to play ranked vs unranked because of pips and reward chests, this also kind of makes the overall ranking kind of immaterial (which leads to people more happily engaging in match manipulation to lower their rank so they can keep duo-ing).

I'd suggest removing the whole pips/chests reward structure entirely, and moving the ticket and glory shards as a per-match reward on ranked.

 

Give players a reward whenever they rank up to a higher tier (akin to the current reward in a chest). These rewards should also be retroactive after placements (so if you get placed on gold 1 you get all the rewards from bronze to gold 1), but one time only (so if you drop from gold 1 to silver 3, you don't get the reward again when you go back to gold).

 

Then at the end of the league, give a MAJOR, unique reward, according to rank (yes, just like on League of Legends - it works, no shaming in learning from others). If you do this right, and especially if the reward is 100% unique and exclusive per season, not repeatable. You'll have a LOT more incentive for people to play ranked. Pair that with longer leagues, and you give players time to get the rank for the reward, and give yourselves time to design the reward.

[Griever.8150]

> @"Cal Cohen.3527" said:

> These two ideas are tied together. We do actually track profession specific mmr, but it isn't currently used **because** we don't lock professions in queue. Without the character lock, players could queue on a profession with lower mmr, then swap to their main profession after the queue pops for easier wins. Some time ago we held a poll asking players if they would prefer using profession mmr and locking characters on queue, but it did not pass.

 

Locking characters on queue would be fine if you could define your role before queuing. Would guarantee everyone has a support for example and make scenarios where you get in teams with 2 Mesmers and 2 Thieves against a balanced comp impossible.

 

Some roles would of course get shorter queue times (such as Support) but having that as an incentive to play the less "shiny" roles isn't necessarily a bad thing in my opinion.

 

 

[Mthe mystery.4615]

> @"Ben Phongluangtham.1065" said:

> This thread is to discussion matchmaking and league play. To kick off the thread, I wanted to talk a bit about the current state of the matchmaker.

>

the difference between the average skill rating of each team was less than 50 points. The matchmaker is doing a good job in most cases.

>

> One thing to keep in mind is that just because the average skill rating of each team is close, that doesn’t mean you won’t have a blowout match.

>

I have a problem with these two statements. Just because the skill rating between players is "less than 50 points" does not mean that the players are of relatively similar skill. I have not competed in ranked the last three seasons because I am sick of the people i'm matched win (on my team and enemy team). For example, in one of my match placement games this season (only played a few), two players on my team at the start of the match went home and stood on the point. Ummm Same skill level? I proceeded to contest far for the majority of the game as i was successfully kiting 2v1s and 3v1s. I even won a 3v1 (yes got each player down and finished), but my team loses a 4v2 at mid.. How? Same kill level? No. Whatever the algorithm is (for me) it is wrong. Im at a much higher skill level then my so called "mmr" and its getting on my nerves, and I assume i'm not the only one with this issue. I do not intend to hurt anyones feelings, however, I've seen some of the T2 platinum players this season play in twitch, or in unranked, they do not belong in Plat to begin with, some I thought were honestly below gold level players, which is insane.

 

[Bazooka.3590]

Its happening right now:

[https://imgur.com/lhWDOeh](https://imgur.com/lhWDOeh "https://imgur.com/lhWDOeh")

It's late afternoon/evening here and after 1min 50sec waiting time MM came up with this match in lower plat. Luckily I was with blue but I watched them to play on the map and they had no afker or leaver.

I can post screenshots like this daily I just don't want to.

My experience is only 1 game out of five is balanced right now.

 

[Ben Phongluangtham.1065]

> @brannigan.9831 said:

> Team based MMR averages for matchmaking is a terrible system. I know you responded to questions already about possible changes but it needs to be done. Let me give you an example and explain what happens with your system sometimes. Team One:1700, 1600, 1550, 1450, 1400 average MMR 1540. Team two 1600,1600,1550 1500, 1500 = MMR 1550. You matcher will say this is a great match. Im here to tell you that the later team will win easily the vast majority of the time because 1450 and 1400 MMR players will lose any fight where they are together unless they get extremely lucky. There is a weakest link element in this game that can't be overlooked. What ever team has the most good players even if the other team has the highest MMR player or two in the batch will win most of the time. Im far from a pro player but I've beat pros plenty of times because they got saddled with the majority of the weakest players in the match. Not to mention the idea of whether its fair or not to have people forced to group with people far below there skill level and I would argue anything over about 150 MMR difference is pretty big difference in skill level if we are talking about true established true MMRs.

 

We actually have been looking at this as well. In the 100,000 match sample we've been studying, we also looked at standard deviation from the average on each team to see its effects. We found that while there is some effect, overall team average had a much higher effect on the outcome of the match. Cal isn't in yet, and all our spreadsheets are on his computer, so I might be able to respond with more details around this later.

[Faux Play.6104]

> @"Ben Phongluangtham.1065" said:

> This thread is to discussion matchmaking and league play. To kick off the thread, I wanted to talk a bit about the current state of the matchmaker.

>

> In a recent random sample of 100,000 matches, we found that in approximately 95% of matches, the difference between the average skill rating of each team was less than 50 points. The matchmaker is doing a good job in most cases. Things get more problematic at the very low and very high skill ratings. Our change to duo queue for 1600+ ranked players is part of our efforts to address this. In addition to that modification, we’re working on some fine tuning on the matchmaker. Our simulation with the proposed changes extended the favorable difference ratio mentioned above from 95% to over 99% of matches. I can’t give you specific dates on when these changes go live, but we’ll be looking to trial them on the unranked queue somewhat soon™.

>

> One thing to keep in mind is that just because the average skill rating of each team is close, that doesn’t mean you won’t have a blowout match. Some maps just tend to snowball, some players tend to give up when they get a bit behind, etc. This can lead to a blowout even if the average skill rating of each team was fairly close.

>

 

I would rather see a bigger gap in the average MMR between teams, and a tighter gap between players on a team. It is not uncommon to have 200-300 point gaps on the same team. This ends up driving players toward the same rating vs separating them. It also makes for toxic teammates when there is a huge skill difference on the same team. On the math side of it, assuming everyone's deviation is around 60 on a team, once you sum the squares the actual deviation of the team is ~134. Based on that scenario I'd want teammates that are within 120 rating of each other and average team ratings that are within 268 pts of each other.

 

The other thing I would like to see go away is the huge deviation people start with for placements. Once someone has played a full season, there is no reason for them to have a deviation that high. It ends up putting people in matches they don't belong in, causing more frustration.

 

One last thing I think is important to get across is the the deviation term in the MMR algorithm. I don't think it is clear to a lot of people how large the deviation term is. Swings of +- 100 pts are within the the 95% confidence interval of the system. When I did the math based on numbers in the wiki I couldn't get a deviation below 60.

[zHasgard.9827]

Matchmaking is doing a good job in 99% because you are looking at the numbers, you still can have bad players in 1500 +.... The teams can have the same rating, but the skill play is what change the game, and this is the second biggest problem in spvp in my opinion, you win or you lose. And the game not even look at you individually, it doesnt matter if you are 20/1, you hold 2 on far for 5 minutes and your team still can't win the other nodes. You can do your best, but if you get a ridiculous team, you are done. -20 points. That's why i dont care anymore about rating and i still get nervous playing spvp. To be honest,i only keep playing spvp because it's faster than roaming in wvw, even with 5 minutes queue. Matchmaking isnt that bad, what needs to change is look more individually to the players to put then in the right place. Also, the leaderbord system is not that cool at all. Most of us come to the forums to give opinions when we get horrible matches in game. I think the leaderbord should be separeted per class, and death/kills/heals/damage/points holded should give points or make you lose less points. And the most important, the pvp team must do the class balances,it's ok if the balance class are made every 2 mounths, but most of them dont even bring big changes to pvp

 

[Namless.4028]

> > @"Cal Cohen.3527" said:

> > > @Ario.8964 said:

>

>> We've talked internally about reducing the maximum rating range of the matchmaker, but the tradeoff is that queue times will increase. We will likely do some testing in unranked to see how much of an impact certain ranges have, and look to make some adjustment for ranked in the future. 50 rating might be an unrealistic goal, but I do think there's some adjustment that can be made here without blowing out queue times.

 

When testing tweaks on the matchmaking it maybe wiuld be a good idea to implement a way to rate the matchmaking on the way.

Maybe a small pop up asking how the MM was (one to five stars) and some space to leave a comment.

[Bazooka.3590]

> @zHasgard.9827 said:

> Matchmaking is doing a good job in 99% because you are looking at the numbers, you still can have bad players in 1500 +.... >

If you have bad players in 1500+ that means Match Making don't do a good job.

 

[Cynz.9437]

> @Bazooka.3590 said:

> > @zHasgard.9827 said:

> > Matchmaking is doing a good job in 99% because you are looking at the numbers, you still can have bad players in 1500 +.... >

> If you have bad players in 1500+ that means Match Making don't do a good job.

>

 

Or placements.

 

The issue i have with looking at pure numbers is that they take pool of 100k matches. Please show me a player that has time to play 100k matches in one season. Theoretically if everyone did it would all even out and everyone would be where they belong. But in reality i would say most people play around 100-300 matches per season so it is by far not enough matches to even out all the extremes.

[EvilSnowflake.1453]

> @Morwath.9817 said:

>

> A reason Quaggan has stopped playing Heroes of the Storm, with his dynamic IP, any IP restart would temporary DC Quaggan, and then he would have to play with toxic people...

 

You cripple your team by disconnecting randomly, yet you expect to be with players who don't do the same to you?

I'm sure short disconnects won't be punished harshly, however if you cause your team to lose by disconnecting, you are part of the problem I've encountered too often.

[Cal Cohen.3527]

> @brannigan.9831 said:

> Team based MMR averages for matchmaking is a terrible system. I know you responded to questions already about possible changes but it needs to be done. Let me give you an example and explain what happens with your system sometimes. Team One:1700, 1600, 1550, 1450, 1400 average MMR 1540. Team two 1600,1600,1550 1500, 1500 = MMR 1550. You matcher will say this is a great match. Im here to tell you that the later team will win easily the vast majority of the time because 1450 and 1400 MMR players will lose any fight where they are together unless they get extremely lucky. There is a weakest link element in this game that can't be overlooked. What ever team has the most good players even if the other team has the highest MMR player or two in the batch will win most of the time.

 

The data that we've been looking at suggests that larger advantages in average rating of a team result in a higher win percentage. The closer the average rating of each team, the closer to a 50/50 result. Obviously average rating is not an absolute measurement for matchmaking as you pointed out, but we can further tiebreak by comparing the standard deviation of the ratings on each team to look for a more even distribution of players. In your example here's a match with the same average ratings but much more even distributions:

Team One

1700

1550

1550

1500

1450

 

Team Two

1600

1600

1600

1500

1400

 

This is just one combination I came up with by eyeballing it; the matchmaker will consider all possible combinations of players and find the best one. So while it would consider your example a "good" match, it won't be the final selection as there are better ones.

 

One thing to keep in mind as the numbers get less pretty is that there will be buckets of average rating difference that are considered as "equal enough" to then compare standard deviations as a tiebreak. For example if we had one set of pairings that resulted in an average rating difference of 9.9 but had a huge discrepancy in standard deviations, this would be considered a worse match than one with an average rating difference of 10.1 and a much smaller deviation.

 

>Not to mention the idea of whether its fair or not to have people forced to group with people far below there skill level and I would argue anything over about 150 MMR difference is pretty big difference in skill level if we are talking about true established true MMRs.

 

As both Ben and I have said earlier in the thread, we're also going to be doing some testing on reducing the maximum rating range of the matchmaker and seeing what impact this has on queue times. I think it's fair to say that the range will be reduced by some amount, we just don't know by how much yet. We definitely agree that the smaller the range the better, but at some point the edges of the rating curve won't get matches in a reasonable amount of time, which is something we want to avoid.

[cyberwolf.5830]

please organize another poll to remove class stacking and implement class lock

[Lyger.5429]

> @EgyptRaider.3946 said:

> I was doubting whether to post this here or in rewards, but it's clearly related to leagues (but not matchmaking):

>

> -could we have a bigger tab of our league history? I notice myself that I'm slowly progressing over the seasons, but it's sad to only see my last season :( (even just a tabel that tells: season X, Rank Y, and MMR Z, would be amazing!)

 

Pretty sure gw2efficiency does this. Rating might be bugged at the moment though.

[Aden.9306]

This is mostly anecdotal evidence, but for the longest time I would have balanced matches 350 to 500, 450 to 500, etc. I get the occasional blowouts but not often enough. The funny thing was my team compositions were wacky and often had 3 of the same profession. I main Guardian so I would be on a team with three Guardians or the opposition would be Guardian stacked with more than two Guardians. Sometimes when I queue as thief the matchmaker gave me four thieves on one team but the matches weren't 500 to 100. Lately, I get balanced team comps but blowout matches and I don't get it. I noticed on the old forums 6-8 months ago everyone was complaining about class stacking; that Arenanet had to fix class stacking to fix matches. Now, from my perspective they did and because of that, the skill rating of teams are not being evenly distributed and most matches now are won or lost well before 500 points.

 

In short, My question is did you guys make changes to the matchmaker to even out the team compositions a couple months ago?

[Faux Play.6104]

> @"Mthe mystery.4615" said:

> > @"Ben Phongluangtham.1065" said:

> > This thread is to discussion matchmaking and league play. To kick off the thread, I wanted to talk a bit about the current state of the matchmaker.

> >

> the difference between the average skill rating of each team was less than 50 points. The matchmaker is doing a good job in most cases.

> >

> > One thing to keep in mind is that just because the average skill rating of each team is close, that doesn’t mean you won’t have a blowout match.

> >

> I have a problem with these two statements. Just because the skill rating between players is "less than 50 points" does not mean that the players are of relatively similar skill. I have not competed in ranked the last three seasons because I am sick of the people i'm matched win (on my team and enemy team). For example, in one of my match placement games this season (only played a few), two players on my team at the start of the match went home and stood on the point. Ummm Same skill level? I proceeded to contest far for the majority of the game as i was successfully kiting 2v1s and 3v1s. I even won a 3v1 (yes got each player down and finished), but my team loses a 4v2 at mid.. How? Same kill level? No. Whatever the algorithm is (for me) it is wrong. Im at a much higher skill level then my so called "mmr" and its getting on my nerves, and I assume i'm not the only one with this issue. I do not intend to hurt anyones feelings, however, I've seen some of the T2 platinum players this season play in twitch, or in unranked, they do not belong in Plat to begin with, some I thought were honestly below gold level players, which is insane.

>

 

He said team average not player average. You can have two teams like this 1700,1650,1400,1350,1300, and the team average metric will say it is a good match.

[Aden.9306]

A separate topic: The Other Guys

 

With the new restriction that platinum two players can't duo queue. An given how much they can weight the outcome of a match. The reverse is worth considering too. Any amount of players below platinum should be able to queue in groups up to 5 players. ( Or at minimally groups of 2 and 3)

 

These players are not going to make leader boards. They're either learning, practicing to get better or some guild activity with friends. This would allow the bronze, sliver, gold player to get better games against groups of equally skilled players. For the sake of the community building and bring in new potential Spvp players. Now you could argue fairness but we need to rebuild the bottom so people can graduate into the top, and not always be the sacrificial lambs to slaughter of high MMR matches.

[brannigan.9831]

Thanks for addressing my scenario. I'm glad your going to test some new matchmaking criteria I just feel like it something you could have tried a long time ago and I just feel like its fairer to everybody when they play with people as close as possible to there exact skill level. Obviously queue times are a considertation. I personally would have no problems waiting 5-10 minutes on average if it meant everyone in the match would be within 50-100 MMR of each other and I dont care if that change didn't end up making the matches significantly more competitve globally. Its called peace of mind. I also wish you would consider showing everyones MMR at the end of the match. You don't have to put names with them which would stop any abuse of players. Why not just show the five MMR ratings of each player on each team with no names attached to them on a tab in the postgame stat window. More information the better imo and I guarantee players would be less angry on average.

[oEnvy.3064]

I honestly think that even if Anet were to somehow create a perfect matchmaking system there would still be plenty of people crying about it. Even with all the "improvements" being made using feedback from this thread, there will still be occasions where people feel like they lost unjustifiably. It is simply unavoidable, sometimes people dc, sometimes you got unlucky, sometimes you lost and it was your fault (gasp) - a lot of people I see complaining in my matches are those who will look to blame anything but themselves, rather than looking to improve they blame anything they can: teammates, matchmaking, match manipulators. Everyday I see average players with inflated egos and sense of entitlement who believe that everyone in the leaderboard is cheating or that you can't have a good rating or win tournaments without cheating. Personally I think it is a waste of time to dedicate developer resources towards something that will not resolve much. I would much prefer we let them focus on implementing actual game changing or exciting things(anything that will make pvp refreshing again for longtime players.) People need to understand that there is a limit on how much devs can do in one single patch, not everyone can be happy - would you rather them implement fun, exciting changes or waste time and resources on something that will have little to no impact, something that will ultimately be underwhelming/underappreciated even from those who wanted it the most. I just think it's sad that this is the first time in a while the devs have given us a chance to voice our feedback and we're not taking advantage of it properly due to shortsightedness, and lack of consideration for developer resources. I know II'm not the most literate person when it comes to writing but I hope that won't discredit my post.

[Loop.8106]

> @"Kuya.6495" said:

> Is it possible to get matchmaker to put together teams based more on what amulet archetypes they are using rather than what profession they are playing?

>

> That is, matchmaker, rather than trying to keep one of every profession in a team, tries to balance a team so there is 3 people with offensive (main stat condi damage or power) amulets, making sure only 2/3 max are either condi or direct damage focused, 1 with a defensive (main stat toughness)amulet and one with a healer (main stat healing power) amulet? Maybe you can test this algorithm in unranked first and see what the results are.

 

This, this and a hundred times this. I can't really reliably queue in on my Core Guardian ~ 1700-1800 rating because the enemy team will have a Firebrand 9 times out of 10. This means I'm basically cheating my team out of a fair match.

 

What I instead have to do is roll on the same class as the weakest player in my previous team, say Thief. This means the thief will likely be placed in the opposing team due to matchmaking trying to create a mirror match-up. In-game I then have to swap back to my Core Guardian build. I don't know if this is actually considering match manipulation or just match-making manipulation. But nevertheless, this is what I do before the majority of my games.

[Sihmm.9236]

> @"Cal Cohen.3527" said:

> This is just one combination I came up with by eyeballing it; the matchmaker will consider all possible combinations of players and find the best one. So while it would consider your example a "good" match, it won't be the final selection as there are better ones.

 

Thanks for taking the time to explain this, I just wonder though if you're really sure this is how things are working out in the game? Anecdotally the matchmaking this current season has felt really, really off - and I say this as someone who's played something like a thousand ranked games since season 5. Most matches just seem to be complete blowouts one way or the other, and not just because one team gets an early advantage and snowballs (I know how to identify this) - but because the teams feel dramatically mismatched. I [already posted](https://en-forum.guildwars2.com/discussion/16391/feels-like-the-game-thinks-my-mmr-is-something-other-than-it-is) about my really weird crazy loss streak earlier in the season after semi-deliberately losing my platinum ranking, and having talked to friends and others on the forums, it doesn't even seem that remarkable a story.

 

Since that post I've been steadily climbing by playing my absolute best on the professions/builds that best suit the map, team and my own experience, but so many games it feels like I'm having to bend over backwards to somehow dig out a win for my team, rather than - as I'm used to from previous seasons - having a feeling of genuinely working with my team to pull off a win together.

 

Perhaps it's just that there are so few players this season that MMR isn't being accurately pinned down, but it's deeply frustrating for veterans and new players alike to be matched in or against teams that just seem to make no sense in terms of relative player performance. It's sapped a lot of the joy out of PvP for me and so I'm playing far less this season... which I realise will contribute to the population being smaller and match quality being worse. And I'm sure I'm not the only one. Again, anecdotally, I know at least one friend who's played every season but quit this season because of the duo queue restrictions and the poor quality matches he was facing solo.

 

Perhaps this is going off-topic, but I really don't want PvP in this game to die - it's so much fun when it goes 'right' - but I just feel like with match quality tanking so hard we're at risk of entering a vicious cycle where more and more people stop playing and fewer and fewer newcomers will persevere long enough to become regulars.

 

OK, melodrama time over!

 

[Master Ketsu.4569]

> @"Xuazinegueri.3592" said:

> Can matchmaking analyze the classes before building the teams? I mean, as i said in a topic, sometimes, one of the teams get 2 of the same class and 0 of an other, while the other team has 2 of that other class and 0 of the first one. I think this could break some team compositions...

 

To be fair this is more of a balance issue. Class stacking wouldn't be nearly as big of a problem if FOTMs didn't exist.

 

 

[Hesacon.8735]

> @"Ben Phongluangtham.1065" said:

> > @Cynz.9437 said:

> > 4. Please consider match outcome by points loss. If match was close, player shouldn't lose as many points - this would encourage players to play to the end. A lot of matches ae winnable but players give up early since there is not enough difference between close match and simply lost match.

>

> In order for this to not be broken, we'd have to also not give as much rating increase for winning a close match. How do people feel about this?

 

I'm all for it. Why should someone gain/lose 20 points when they barely won/lost?