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

If there are indeed additional variables outside of simply balancing SRs, then people aren’t crazy and the matchmaker can be giving the appearance of being broken. Which is it? Anecdotally speaking, since they’ve changed SR gains for streaks (season four or five), I’ve endured these weird hot / cold streaks exactly like they’re being described out here. You’re either steamrolling, or being steamrolled. The slog has become unfun because you feel an utter lack of control in your games, despite putting your best foot forward.

1 Like

There is a hidden MMR, and SR is just the easiest way to interpret it as a single value, since actual MMR is a bunch of constantly changing and scrambled variables. All MMR does is determine matches based on stats and averages (all mp matchmakers have this BTW), but its existence gives the reckless theory of forced loss streaks a single shred of possibility.

And so all the people that can’t improve fast enough to get out of their ranks latch onto it as the only possible reason they could be stuck in whatever rank they’re in, because they would rather blame a system than themselves.

1 Like

Ping is also a variable. It tries to group people with similar ping.

1 Like

My incredulity has nothing to do with my current Elo. I’m an old gamer and happy to be ranked platinum. However, isn’t it a bit curious that the only consistency we see with win streaks are the inevitable losing streaks afterward? I haven’t taken to plotting data like some of these guys have, but the yo-yo effect is very real.

Can you source this?

The majority of people lack game sense to make better counter picks and or changes in playstyle to climb and loss streaks happen when they only know 1 way to play and the enemy knows more ways. More games are winnable than not and people just want to blame the system. Example is a “smurf” widow carrying, she keeps getting picks but your team isnt running a single shield and then the hog on your team decides to go widow to counter the widow because dps isnt killing her when you have 2 hitscan, which the benefit of getting a shield for them would probably outweigh the benefit of hog going widow. (Yes there are many factors to this and the example was just showing where the mind set probably should be going instead of where it does for character choices)

1 Like

The real reason for this “yo-yo” effect is simply you climbing to a point where you can’t win anymore, then losing, its a very simple and common phenomena with competitive ladders. The problem I’m referring to is people using this as “evidence” that the company is trying to keep them in the rank they are in.

Also there is no real source for this, ping is just a common factor used in matching any amount of people in a system where latency can effect the outcome of a game. If it did not check for ping, then you would sometimes get matches upwards of 300ms and not even be able to land a single hit.

Keep in mind, you most likely play a specific role. If you are queued up with a team full of people playing their main role type, you are more likely to steam role an equal SR team who is trying to flex.

I play with 1 main tank, I’m an off tank, and a main healer. All 3 of us can flex to a supporting hero in that role for each required map type. That means, we need a good off support, and 2 good dps to queue with us. Usually, we get people who want to play DPS but are not actually good at the role. It’s a rough game and hard to carry, but I can do it. When we get true DPS who earned their level, we straight steam roll teams. If we get DPS that suck, it’s hit or miss if we can carry, especially if the other team is good on all rolls.

If our main tank chooses to play off support or dps, our randoms almost never play MT. They always choose off tank (my role). The other off tank chooses an off tank that isn’t supportive of that map, or the comp type. Those games I lose more often than I win. Ironically, our MT player who does DPS always gets gold damage competing against a DPS main.

In any rate, over simplification of “forced wins/losses” doesn’t make much sense due to this. There are far too many variables.

In my opinion, where there are in fact games you cannot win, the larger your SR swings, the more inconsistent you are playing. When you become more consistent, you’ll have less SR swing gaps, and eventually move up. It comes down to a mistake you know you make, and mistakes you don’t realize you make.

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 http://overwatch.blizzplanet.com/blog/comments/blizzcon-2016-overwatch-qa-panel-transcript/3

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

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 ( Overwatch Forums).

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 https://www.reddit.com/r/OverwatchUniversity/comments/aatezy/why_match_quality_is_frequently_poor/.

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 Overwatch Forums

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.