Role Queue Update


(Scott Mercer) #1

We’re now over halfway into the Role Queue Beta Season, and the introduction of Role Queue into Quickplay and Season 18 are just around the corner on September 1. We wanted to take the opportunity to address feedback we’ve been reading on our forums and other social media, as well as provide some insights into how these systems work.

The primary goal of this feature has always been to increase the quality of Overwatch matches, and it’s been great to see so much positive feedback about match quality in a role queue world. Creating high quality online multiplayer games of Overwatch has multiple components and Role Queue has helped us do this in several important ways.

We believe the most exciting, fun games are between players of similar skill. Role Queue allows us to track your skill with the three roles separately. So, when you play as a 2100 SR Tank we try to match you with 2100 SR Damage and Support players, both on your team and on the enemy team. We can’t guarantee that all your games will be super close 3-2 games that ended in overtime. Stompy games sometimes happen even with great matchmaking and well balanced teams. Improving our skill-based matchmaking with Role Queue has been a clear net positive for match fairness so far. We’ve also seen feedback from players who enjoy having their three separate SRs as goals to drive personal improvement.

We also believe that games of Overwatch are better when both teams have balanced team compositions. This has been the loudest, most noticeable positive feedback from everyone. Players can no longer get matched into a game where no one wants to play a tank. You can also feel better queueing as a tank knowing that there will be 2 healers to help keep you alive. With both teams guaranteed to have 2 tank, 2 healer, and 2 damage roles the games immediately feel fairer and more winnable. Players are more engaged, perform better, and ultimately have more fun when the games are balanced.

Preventing discussions about who plays what role from overheating into arguments or worse is also having an immediate impact on player’s enjoyment of the game. Overwatch is at its best when teammates are working together in a positive fashion. Negative communication not only feels bad, it makes you play worse.

With the generally positive feedback we also seen some a few concerns from players.

Why are queue times for Damage role players so high?

We’re monitoring the data very closely, and I can share some information here. Across all skill ranks, we do see higher queue times for players who specifically queue for Damage Only. Most damage players usually see 3-6 minute queues, unless they’re in the masters or grandmaster skill tiers. Players with very high SRs will always see higher queue times because there’s fewer players of similar skill to match against them, but we’ve seen reports and have data showing some players are waiting 20-30 minutes. We’re investigating this and are already working on some matchmaker adjustments for future updates to help address the issue. If you can find a friend who plays Tank or Support, queuing together as a Damage/Tank or Damage/Support duo drastically reduces the queue times you’ll face.

Why are damage role players getting unfairly blamed for losses?

Now that the team composition is not an easy reason to blame for a team’s loss, I’ve seen community feedback and even players in my own games feel that they’re losing because no one on the enemy team is dying. They then blame the players in the damage role. Everybody should always remember that Overwatch is a team game, and everyone is responsible for the team’s success. Losing a game doesn’t mean that your team necessarily played particularly poorly, it might just mean that the other team played particularly well that game. Winning or losing often comes down to who made one more big play.

If you’re trying to improve your level of play over time, blaming a teammate for a loss doesn’t help. Instead ask yourself if there is a lesson you can take away from the match (win or loss) that will help you improve in the next one. Similarly, before a game starts think about what aspect of the game you can work towards improving.

Why did I complete placements and receive an SR that was too high/low?

If you’re a new player, placements are straight forward. You start at a “new player baseline” SR and when you win games your SR goes up, and losing games lowers your SR. New players are also set at a higher variance level, so the amount of SR you gain and lose is multiplied higher. The matchmaker does this because as a new player it knows very little about your skill and wants to determine it as fast as possible. The increased variance slowly decreases as you play more and more games, and the matchmaker is more confident in identifying your skill.

If you’re a returning player, placements are still simple. Instead of starting you out as a “new player”, we start you out at your “historic baseline” SR that covers more time than just the previous season. Your SR then changes as you play your placements, but since you’re not a new player we use the base variance level. This means that you’re unlikely to see radical change in SR from one season to the next, but that reflects the fact that your actual skill at playing Overwatch doesn’t radically change either.

We’ve actually tried in a few seasons doing something that amounts to a partial SR reset across seasons, but the results were pretty horrible. When you queue for a match with an SR that doesn’t accurately reflect your skill, the quality of the match drops tremendously. Therefore, we try to quickly change your SR as a new player to more accurately determine your SR, and we can’t ever really consider implementing a full SR reset. Match quality would be harmed for months until everybody played a lot of games and the matchmaker could properly identify everyone’s skill again.

Now let’s talk about placements in the new Role Queue world. To help us seed your role’s SR after initially completing placements, we’ve been tracking your match results as you play the three different roles for several months. We want to help identify your skill at each role, so we can place you at the SR that reflects that skill. If we do a good job at that, you’ll immediately start playing in fair matches. The more inaccurate we are, the more games you’ll have to play on that role until your SR can adjust properly to reflect your skill. This also means that the more games you played on each role, the more accurate we could be with your new role SR.

What happens if you’re a returning player, but you never played one of the three roles? Since we don’t any have data to help us guide what your new role SR should be, we turned to some interesting data science. We looked at the population of players where we did have data for all three roles, and asked questions like “How well does a typical GM skilled player who primarily plays damage perform on the other two roles?” We then duplicated that question across all skill tiers, and the three different roles. This gave us some guidance on how to adjust players who hadn’t played a role yet. We also temporarily increase your variance level for SR gains and losses, so in cases we’re incorrect we can adjust you faster to the SR that matches your skill.

We appear with the Role Queue Beta to have done a decent job overall with seeding, but of course there will always be exceptions. We’re already looking at making some adjustments to the formulas to better match the live data we’ve been seeing, such as temporarily increasing variance and being more aggressive with our adjustment for roles with prior data.

We saw a few reports from very high skilled players who were incredulous that they seeded so high at an off role, but then we looked, and they had performed well on that off role in a significant number of games in the past. When we seeded their role SR, we used that prior data. Something else to remember is that we as humans are not particularly good at identifying our own skill at something. This is referred to as the Dunning-Kruger effect, and if you’re curious to learn more feel free to use your favorite search engine on that term.

In the end, the best way for you to improve our matchmaker’s ability to place you with other people of similar skill is just to keep playing. You’ll sometimes have some lucky win streaks, or unlucky streaks of losses, but if you keep playing your SR will continually adjust to you reflect your present skill and you’ll play in better games.

Your feedback is always important, and the Role Queue Beta season has been very helpful to us. There are still more exciting changes to come! On September 1, Season 18 of Competitive Play begins and we’ll introduce Role Queue into Quick Play. We’ll closely monitor queue times, match quality, and all your comments across our forums and other social platforms. We’re excited about the future of Role Queue!


2-2-2 lock is a disaster and here's why [UPDATED]
So, it has happened
SR, MMR, and Role Queue
Detailed writing on why Role queue will single-handedly eliminate a sizeable chunk of the playerbase
2-2-2 lock is a disaster and here's why [UPDATED]
Are we seriously going to ignore this?
Detailed writing on why Role queue will single-handedly eliminate a sizeable chunk of the playerbase
Please just reset MMR
SR, MMR, and Role Queue
The SR in SS18 will calibrate with RQ beta SR or from latest season? or will it start all over?
SR, MMR, and Role Queue
Quickplay does not affect comp mmr!
Reset MMR will fix comp
MMR RESET ? real or not plz answer
2-2-2 people just as unhappy but now with extra wait
The Hypocrisy of Anti 222 Players
Any official communications from Blizzard HQ regarding DPS Role Q Times?
Overwatch Blizzard/Developer Post Directory - Last Updated: 9/12/2019
Lack of MMR reset is lazy and ridiculous
Can people stop complaining for at least a day
Everyone said big streamers would come back for 2-2-2?
How Competitive Matchmaking and Rating Works (Season 18)
Most people want MMR reset
Can somebody give me a quick rundown on what happend?
It's chaos and I love it! :fire:
What? How did I lose this much sr
Misplaced players / Bad Placement System / "Too few avoid slots"
15-20 Minute Queues on DPS... In Plat
Role Q is deterring me from placing in my off-roles
Why aren't actual smurf accounts banned?
Role Q beta stats not counting
SR, MMR, and Role Queue
I lost 700 SR during Beta season
15-20 Minute Queues on DPS... In Plat
15-20 Minute Queues on DPS... In Plat
MMR Reset for Season 18
Trash m8s in matchmaking
2-2-2 broke the game (told you so)
Remove 222 from Comp
2-2-2 broke the game (told you so)
Remove 222 from Comp
So I'm *still* seeing pro 2-2-2 develop anti-2-2-2 threads
:white_check_mark: Why is a 3.5k DPS = to a 3.5k Tank?
I strongly expect them to revert role lock
Unpopular opinion: 222 was always the most boring team comp
2-2-2 proves that bullies win in 21 century
On my 2-2-2 poll results
Maps and MMR question
Maps and MMR question
It Hurts To Admit, But
It Hurts To Admit, But
(Molly Fender) pinned #2

#3

Alright… let’s read this.


#4

Hi dev! Why are you nerfing Symmetra? Isn’t she supposed to be good vs double shields? Shouldn’t we wait to see how the meta would evolve before nerfing her too much, I’m afraid she would go back to a bad spot again.


#5

Scott, can you 100% clarify if the accumulated matchmaking rating from Role Queue Beta will be or not be passed over to season 18? I have been going off of the news blog post that all stats from the beta will not count, but the community is still concerned about this aspect.

Answered below:


On a personal note, I sincerely apologize for delivering incorrect information to many of you earlier last week across multiple threads. It is not my intention to cause confusion or frustration over this. While it is my unending goal to help wherever and however I can, I am still a player like you. Forum MVPs are not Blizzard team members, but we are selected to be Forum MVPs by Blizzard because of our passion for their games and desire to help others. Click here to earn more about the Forum MVP program. Thank you everyone for your understanding and continued support in my efforts here on the forums.

~ Myst (^^)v


Does role queue beta affect MMR?
Wyoming Myst .... characters and words
#6

This post has so many golden quotes I need to archive for future use


#7

Scott:

that was a lot of writing, and I honestly and sincerely appreciate any input we get from the dev team whatsoever, but FWIW none of that really addressed my main concerns about this change


#8

Wait… Did they just misuse Dunning-Kruger to imply that streamers are inept? Ha ha ha… Like, I already knew from the first 3 paragraphs that nothing relevant was going to be posted. But that was my biggest take away.


#9

Fun read, more of just an update that they’re actively working on it, rather than updates on what exactly they’re doing. Still very appreciated <3 We love and support Dev communication.


#10

i really like role-q but i hope the SR system gets better because many dps players are not on the rank they should be.


#11

I literally lost all but 1 game on my dps placements and got mid diamond. Any rank but your main rank on role que is busted.


#12

I wonder, full and soft resets may not work, but what about weighting the MMR towards the most recent seasons?

I’m thinking something along the lines of the club coefficient that UEFA uses for Champions League and Europa League seeding purposes. UEFA uses a 5 season rolling system, so all results during the previous 5 seasons count, but beyond that they’re scrubbed.

I feel that this would be a way to improve the chances of those that actually get better starting to climb, without actually changing anybody’s current SR at all. It would be the MMR that fluctuates.


#13

No, keeeyt, they are, it is just that they themselves do not realize it. Nevertheless, their SR changes will become more aggressive, so never fear.


#14

How did you handle people whose only DPS competitive playtime consisted of last second switches to Tracer at the end of a round? These people probably had horrible DPS stats without it really meaning anything…


#15

I do always wonder if blizzard devs themselves ever play the game. mmr reset IS and will always be needed. Someone good at one meta will not be good at another. OWL is clear proof of that. MMR resets yearly are needed and will always be needed. Not even starting them off as a fresh player but something to make them gain it back and prove they deserve the rank they are. Someone getting up with a hero that was considered broken and now unable to play anything at a that same level after that hero is nerfed is another clear issue. Blizzard realize 1 year ago and now won’t have same effects. Test out mmr reset because without it game quality is already horrible. I’d rather a few months of poor game quality rather then the rest of overwatch’s existence being horrible.


#16

He said they tested it a few different seasons and they did not like the results. Albeit, for role queue, ironically, we are pretty much reset anyways. At least those of us that played one role. The only difference is that we have to go on loss streaks to get back where we belong unless we could secretly play a role we never learned, which apparently is the case.


#17

no it is not.

i’m glad they don’t listen to people like you.

mmr reset will ruin this game


#18

Imma be honest, it felt weird being at 2600 SR as a support, like I don’t belong there. I haven’t been that high up since season 4, and I got to 2600 these last few seasons because I played damage characters.


#20

May i ask what rank are you and how often do you play comp?


#21

Rolling mmr over a set number of seasons, earliest season scrubbed when a new season starts. No actual reset