Matchmaking

General Discussion
I am here to inquire about the specifics of the matchmaking algorithm. I have noticed as a Top 500 player that the teammates assigned to me are widely varying in skill (anywhere from SR 46 to Top 500). The variance in skill among my teammates creates for unpredictable and often non-competitive games. I would like to know why this happens, if at all possible.

For instance, in League of Legends you would never see a team with 4 challenger players and a bronze player to offset the difference in MMR. I understand that would increase queue times but I believe I am in the majority when I say that I would prefer to wait longer for better games. I think the system may be too concerned with finding a game quickly over finding a "good" match.

While I can appreciate dynamic queue for different party sizes, can the disparity in ranks among teammates be attributed entirely to that?

Any insight you may be able to offer on the intricacies of the matchmaking system would be greatly appreciated.
+1

As an SR 2900 player, I have seen people who have SR 40 in season 1 have a higher rank than I do, and people with SR 70 have a lower rank than I do.

Overall, this season's ratings have been very inaccurate in comparison to season 1's, and even in comparison the old PTR.
09/22/2016 02:12 PMPosted by Ruwin
anywhere from SR 46


Are you really getting Bronze players in your game? That's a bit hard to believe, but shocking if true.
09/22/2016 02:16 PMPosted by GrumpySquid
That's a bit hard to believe

The people that I've been getting have been around SR 46 season 1. I'm seeing these people when queuing with players around 3400 SR this season. I'm sure that its not a stretch to see lower level players who had godlike placements or were carried in their first season 2 matches.
09/22/2016 02:16 PMPosted by GrumpySquid
09/22/2016 02:12 PMPosted by Ruwin
anywhere from SR 46


Are you really getting Bronze players in your game? That's a bit hard to believe, but shocking if true.
He's not saying that he's getting Bronze players in this game, just using an example of League. You'd be hard pressed to find many Bronze players in this game in comparison as everyone seems to level out around gold at lowest.
09/22/2016 02:19 PMPosted by drew
The people that I've been getting have been around SR 46 season 1. I'm seeing these people when queuing with players around 3400 SR this season. I'm sure that its not a stretch to see lower level players who had godlike placements or were carried in their first season 2 matches.


09/22/2016 02:27 PMPosted by konr
He's not saying that he's getting Bronze players in this game, just using an example of League. You'd be hard pressed to find many Bronze players in this game in comparison as everyone seems to level out around gold at lowest.


Okay sorry, using Season 1 SR was throwing me off since I usually see it abbreviated "SR1 XX" if we're not referring to the new SR system.

I'm on the same page now.
They will never give us exact details on their matchmaking algorithms

why?

plausible deniability
09/22/2016 02:37 PMPosted by CultofZAL
They will never give us exact details on their matchmaking algorithms

why?

plausible deniability


I'm not asking for specifics.

All I'm asking for here are higher quality games. If that means longer queue times, so be it.
https://twitter.com/CodexaGG/status/779128151397982209

What's with the sudden spike of people at 3000~ SR?

Screencapped from http://masteroverwatch.com/leaderboards/pc/global/mode/ranked/category/skillrating
09/22/2016 02:12 PMPosted by Ruwin
I am here to inquire about the specifics of the matchmaking algorithm. I have noticed as a Top 500 player that the teammates assigned to me are widely varying in skill (anywhere from SR 46 to Top 500). The variance in skill among my teammates creates for unpredictable and often non-competitive games. I would like to know why this happens, if at all possible.

For instance, in League of Legends you would never see a team with 4 challenger players and a bronze player to offset the difference in MMR. I understand that would increase queue times but I believe I am in the majority when I say that I would prefer to wait longer for better games. I think the system may be too concerned with finding a game quickly over finding a "good" match.

While I can appreciate dynamic queue for different party sizes, can the disparity in ranks among teammates be attributed entirely to that?

Any insight you may be able to offer on the intricacies of the matchmaking system would be greatly appreciated.


Stop trying to confuse the matchmaking arguments with your facts and logic.
09/22/2016 02:16 PMPosted by VoRteX
+1

As an SR 2900 player, I have seen people who have SR 40 in season 1 have a higher rank than I do, and people with SR 70 have a lower rank than I do.

Overall, this season's ratings have been very inaccurate in comparison to season 1's, and even in comparison the old PTR.


TBH, I don't think you can compare it to season 1 and say it's better or worse from there. Season 1 was a complete mess. Individual performance calculations made DPSers rank highest, tanks in the middle\high, and supports low. It was so bad I don't think we should really draw comparisons to it.
Well firstly, they complety revamped the system, so season 1 rank isn't reliable data to use comparatively to season 2. They also made major changes to the test server system before it went live so that data cannot be used either.

So all we have to work with is season 2 data, and you did not give us the ranks of the players you are referring to, or any type of metric at all.

So at this point, all we have is just another "why am I always better than my teamates" thread, posted by yet another Dps main.

Not saying that this is the case, but without meaningful data to validate what you are implying, that's how it appears.
09/22/2016 11:53 PMPosted by PhotonFlingr
Well firstly, they complety revamped the system, so season 1 rank isn't reliable data to use comparatively to season 2. They also made major changes to the test server system before it went live so that data cannot be used either.

So all we have to work with is season 2 data, and you did not give us the ranks of the players you are referring to, or any type of metric at all.

So at this point, all we have is just another "why am I always better than my teamates" thread, posted by yet another Dps main.

Not saying that this is the case, but without meaningful data to validate what you are implying, that's how it appears.


I suppose I can touch on this.

I am detailing the season one rank of these players because the season two rank is not an accurate reflection of skill. Because the system has shifted from a 'performance based' point win/loss to a 'streak based' point win/loss you will see players who have artificially risen due to luck.

The previous system was much better at determining skill. When you base your MMR gain on factors only you control (your performance as compared to others) the system can be much more certain when guessing your skill. When your performance can be negligible and you still gain ~100 points per win because of exceptional teammates, the system starts to think you're better than you really are. If you're playing out of your league but your teammates are good enough to carry you, you'll gain just as many points as them.

It is clear this new system was designed for climbing. So that people could reach their goals and feel satisfied in themselves. "I'm a Master ranked player! I did it!" That means a lot less this season because reaching Master can be as easy as going on a fortunate win-streak within a matter of hours.

Was the previous system flawed? Yes. Was it better at determining skill regardless of those flaws? Yes. I will admit, the system did not determine performance properly for support and tank players. Perhaps they did not have enough raw data at the time to determine what fantastic performance looked like based on those statistics. However, they were on the right track with performance based MMR win/loss. There should be a flat gained amount (+-15MMR) for wins and losses. In addition, should be the system's consideration of performance (0-25MMR). This way people who are clearly performing among the top .05% of the player base will rise more than twice as fast as those whose performance is average for their current rank. I will not bother detailing all facets of the system, but I trust you will understand why this is better than the current one.
I do sometimes wonder about the matchmaking system. I can be in a very balanced game (in my opinion) where multiple rounds are played to the bitter end. The next game which might have an SR difference of 20 points ends up being a 3-0 stomp one way or the other. One game our team was badly stomped and it was 5 vs 6 with the 6 on our team.

Why the difference? Is it the type of map, the picks of the players, one team full of hard counters for the heroes of the other team?

Is the SR value the only metric used to group players together? To me it seems the SR value is not enough to capture the "skill" or performance of a player in a team. For instance, it helps no one to throw 6 players of the same SR into a team when all 6 have only played Lucio. What else is used to match a team? I personally feel that there is no balance consistency between matches at all.
09/22/2016 02:12 PMPosted by Ruwin
I am here to inquire about the specifics of the matchmaking algorithm. I have noticed as a Top 500 player that the teammates assigned to me are widely varying in skill (anywhere from SR 46 to Top 500). The variance in skill among my teammates creates for unpredictable and often non-competitive games. I would like to know why this happens, if at all possible.

For instance, in League of Legends you would never see a team with 4 challenger players and a bronze player to offset the difference in MMR. I understand that would increase queue times but I believe I am in the majority when I say that I would prefer to wait longer for better games. I think the system may be too concerned with finding a game quickly over finding a "good" match.

While I can appreciate dynamic queue for different party sizes, can the disparity in ranks among teammates be attributed entirely to that?

Any insight you may be able to offer on the intricacies of the matchmaking system would be greatly appreciated.


Hi Ruwin,

Can you please clarify a little bit? Were you actually matched with a Bronze player or were you simply citing that as an example from League of Legends? Also, it's more helpful if you refer to Season 2 skill ratings as Season 1 skill ratings do not map 1 to 1 with Season 2. The other thing that would be helpful is to know if grouping was involved in any way.

Top 500 players are the hardest players for us to matchmake. They are obviously the most skilled players at the game and there is not a large pool of players for them to play against. Because we matchmake aggressively to get you a good ping, this further reduces the pool of players we have to match you with. There are oftentimes not 5 other Top 500 players in your part of the world sitting in the queue in Competitive Play at the same time as you.
Oh good a top 500 player says Matchmaking is garbage and gets a blue post. The other 100 threads saying matchmaking is garbage completely ignored. Awesome.

Basically it reads: We want the top 500 players to have a good time, screw everyone else.
Oh good a top 500 player says Matchmaking is garbage and gets a blue post.


There was absolutely nothing garbage about his post. It was a legit question worded in a polite manner. You could learn a thing or two about how to post by reading over it again.
09/29/2016 11:59 AMPosted by Erras
Oh good a top 500 player says Matchmaking is garbage and gets a blue post.


There was absolutely nothing garbage about his post. It was a legit question worded in a polite manner. You could learn a thing or two about how to post by reading over it again.
He didn't say the post was garbage. He was agreeing that the matchmaking function is garbage.
09/29/2016 11:30 AMPosted by Jeff Kaplan
09/22/2016 02:12 PMPosted by Ruwin
I am here to inquire about the specifics of the matchmaking algorithm. I have noticed as a Top 500 player that the teammates assigned to me are widely varying in skill (anywhere from SR 46 to Top 500). The variance in skill among my teammates creates for unpredictable and often non-competitive games. I would like to know why this happens, if at all possible.

For instance, in League of Legends you would never see a team with 4 challenger players and a bronze player to offset the difference in MMR. I understand that would increase queue times but I believe I am in the majority when I say that I would prefer to wait longer for better games. I think the system may be too concerned with finding a game quickly over finding a "good" match.

While I can appreciate dynamic queue for different party sizes, can the disparity in ranks among teammates be attributed entirely to that?

Any insight you may be able to offer on the intricacies of the matchmaking system would be greatly appreciated.


Hi Ruwin,

Can you please clarify a little bit? Were you actually matched with a Bronze player or were you simply citing that as an example from League of Legends? Also, it's more helpful if you refer to Season 2 skill ratings as Season 1 skill ratings do not map 1 to 1 with Season 2. The other thing that would be helpful is to know if grouping was involved in any way.

Top 500 players are the hardest players for us to matchmake. They are obviously the most skilled players at the game and there is not a large pool of players for them to play against. Because we matchmake aggressively to get you a good ping, this further reduces the pool of players we have to match you with. There are oftentimes not 5 other Top 500 players in your part of the world sitting in the queue in Competitive Play at the same time as you.


Everytime I have a terrible Diamond player on my team and I look up their Season 1 ranking it is always around 45. The Season 1 rankings are undoubtedly more accurate than Season 2 rankings, except for players who started S1 just when they got the game and have since greatly improved.

On a separate note, I noticed that one of my friends managed to break into the Top 500 and Master with under a 50% win rate, when his placement was about 2800. Something seems very wrong with the algorithm.
09/29/2016 11:51 AMPosted by FunkyBunch
Oh good a top 500 player says Matchmaking is garbage and gets a blue post. The other 100 threads saying matchmaking is garbage completely ignored. Awesome.

Basically it reads: We want the top 500 players to have a good time, screw everyone else.


What does and doesn't get a blue post isn't really clear. A week ago somebody with 1 post made a post about how they wanted a feature to be added into the game and got a blue post in response.

Join the Conversation

Return to Forum