Does the matchmaker use SR averages only, or are there unspoken variables?

Shouldn’t it be a gradual climb to reach that point where I’m incapable of winning? Most of these posters are commenting that these loss streaks occur after a winning streak. By my same logic, shouldn’t it also be a gradual progression down the ladder? How come there is so much volatility getting you to your proper Elo?

Also, you keep using terms like “common” and “simple”. Is there documentation written in common and simple language that we can read to better understand how all of this works? To a layman such as myself, it’s all very interesting and I wish I had your knowledge of its intricacies.

There might be a new topic about this in the new forums but im not sure. This is from the old forums. Some of this has been changed concerning groups and % difference allowed before a match is made but most of it is still in place.

1 Like

The volatility of this “yo-yo” effect is most likely caused by the randomness of online play, and game with 12 people in it is going to have 1 or 2 people that are having a bad day, are tired, or just mad at the game. People make things random, and randomness induces volatility.

I can’t link specific sources as most of this I learned years ago in college (I graduated in game development). But to put it simply:

A matchmaker check a whole bunch of different things when making a match, from ping differences, to specific stats of all considered players.

One of the biggest influencing factors in matchmaking is ping, or the time in milliseconds it takes for a signal to go from the player, to the server and back to the player. If 2 people have similar pings, they will be a preferred match as large ping differences will lead to shots not connecting when they look like they should and so on.

Now the specific stats that make up MMR in Overwatch are not public, but generally they try to match teams together with 50/50 odds of either team winning. Sometimes they have to stretch this if the queue time goes to long, which can result in bigger SR differences between players.

That’s about all I can theorize outside of what Blizzard has told us, keep in mind that I do not work for them and this is just general knowledge of how matchmakers work in these kind of games from a game programmers point of view.

Thank you for your response, but my contention is the lack of randomness myself and others are experiencing. It’s all very clockwork; you stomp for an evening, then you get stomped the next. The goal of the matchmaker is to prevent that, yes? The post that Mizzerella shared had a response by Jeff Kaplan explaining that was the goal. There was some good information in his post, but none of the minutia I was seeking. In fact, the ambiguity of how to manage five different scenarios of games lends some credence to our complaints. How exactly do you pair someone up with 11 other people when it’s not known how you play with all “21” heroes?

Regardless, all that matters to the end user is their perception. I actively seek quality games when queuing in competitive mode. The frequency of one tricks, throwers, smurfs, on both teams, multiplied by the veneer of toxicity over this community, adds to a perceived disproportionate amount of lopsided wins / losses. It’s bad enough to compel me to voice my concerns on the forums. Something is broken. And if it’s not, it sure feels broken enough that I won’t play competitive again until something changes.

I get it. None of us really know how the match maker works. We all go by our own experience and trivialize each others’ complaints. I just want Blizzard to know my thoughts. Thanks for reading!

No. MMR is a single number per player.

“So we are looking at their MMR, and skill rating is very closely attached to MMR. MMR means matchmaking rating. It is a hidden number.” – Jeff Kaplan

"A second factor we take into account is ping. " – Jeff Kaplan

There is nothing particularly weird about streaks. You’d see similar behavior if you were just flipping coins.

Win probability changes slowly with rank because there are so many random factors in each individual match. Unfortunately, it follows from this that frequent and long streaks will occur, and a player’s rank will oscillate widely. Essentially, a player will tend to bounce between the range of where he is nearly guaranteed to win and where he is nearly guaranteed to lose. The range varies from player to player, but +/- 250 SR/MMR is common and +/- 500 is possible. This problem can be analyzed in depth, mathematically (

See How Competitive Skill Rating Works (Season 15) for more detail.

i am not sure about the ping thing to be honest i have had people who had a ping difference of 60 or 70 to mine. i think it is not a variable or at least its not highly prioritized in the search system

Unless they do MMR different than any other game in existence, its not. Notice how he didn’t say anywhere in there that it was a “single” number, funny how a quote taken out of context can support a confirmation bias right?

Jeff did however continue on in the next sentence to say “It looks weird and mathematical” which would suggest that it is not as simple as “a single number per player”. :thinking:

Sure, but how hard can you flip a coin? Can you slam heads or slam tails? The point is the quality of the games of these streaks we’re describing. They’re never close.

1 Like

That’s just the nature of a game like Overwatch, something as small as 1 person not performing as good as they can lead to a stomp in most situations if said person does not fix their mistakes during the match. A game that relies heavily on teamwork will always have these kind of games when played with random people online, the only way to get around this is to play in online tournaments like Open Division or setup some custom games.

To talk about quality instead of result, check out

1 Like

Weird and mathematical presumably means that it is a floating point number.

You say that I have confirmation bias, but your confirmation bias is to the point where you cannot understand the plain meaning of the text.

But here is another quote for you: “Anyway, this leads me to matchmaking rating…All the system does when it comes to matching on skill is attempt to match you with people of a similar number.” – Jeff Kaplan

But if you want all the quotes, see How Competitive Skill Rating Works (Season 15), references section.

To add to this, Stylosa mentioned internal conversations with Blizzard employees that had MMR as a “horrendously long number”

Nothing concrete to say if it’s a floating point value or an integer, but the point is, it’s not meant to be human-interpreted.

The system is probably a bit more complex than just trying to average out SR’s.

That said, there isn’t a variable that says “this person has been winning a lot, let’s screw them over”. The matchmaker has no control over leavers, throwers, smurfs, or idiots. It can not predict human behavior or interactions, all it can do is place players with similar stats together.

No matter what you think, the game is not rigged against you and specifically you.

Yup, that’s my entire point. MMR being directly shown to the player would confuse 90% of the players and lead to more outrage as to why it changes the way it does. SR only exists to define the in game ranks. They should honestly just get rid of SR and only show the ranks, plenty of other competitive games do this to great success and none of them have these crazy “forced loss-streaks” theories popping up in their forums.

I prefer to call them observations. heh.

I so wish they did this.

Even if players were to believe the 50/50 conspiracy, it wouldn’t matter if their values were hidden, because they’d be free to believe they still net a gain due to the performance factor. It makes the entire experience much more enjoyable than being metaphorically kicker in the balls over and over again as you’re forced to watch yourself stagnate, or even drop.

Unfortunately, I sorta doubt it’d happen. People freaked out when they dropped winrates from Quick Play stats. Imagine the outrage if they were to just say “Hey, remember that SR thing? Uh… forget about it for Season 16. Just trust us with this one, especially you, T500 players!”

When they include baseless accusations, assumptions and insults, towards a company that only benefits when people enjoy their game, I would call them conspiracy theories. Especially when they directly ignore any logical fallacies in their argument.

Yup, this is exactly why the whole “internet outrage culture” is slowly killing the games industry (at least AAA games), people are more excited about finding something they can insult and be outraged at, than actually playing anything.

Here are two players that even went as far as recording their “observations” on video.

All this proves is that people are self-biased (and let’s be honest, who isn’t?) and thus, bad at evaluating themselves and other people. Especially the first dude. He renamed his YouTube videos after everyone rightfully criticized his “perfect play” as he first labeled them.

1 Like