I understand that rank and hiddem mmr are two seperate things. But how does the game determine your hidden mmr? What rank it thinks you belong in? Are there algorithms in place that determine the positioning and game awareness of a player and correlates this information to a rank? Does it go by average accuracy or average deaths, KDA, etc? I’m really curious to know.
It’s probably hero specific stats and how you get those stats and who you get those stats against (i.e. some correction factors)
Algorithms with lots of different factors. I wouldn’t be surprised if AI is part of it. I’m not a programmer though. I really don’t know.
1 Like
Assuming it’s all fair and some hero’s not weighted more than others, bugs or any other illogical actions. It’s mostly your performance with a hero vs every one else on the same hero of that rank.
In terms of QP and arcade, I don’t believe it really judges anything any more. But I believe it is supposed to look at your overall stats across all modes.
The main problem with this is it really cant’ tell from a goofy match with friends and a hard core “try” mission. Once you do poorly it takes a lot of good matches to try and over come that.
But in all honestly we really don’t know because Blizzard will never tell us as they are afraid of people gaming the system if they know the exact details. The main issue is of course winning isn’t a big factor.
If you push that payload and win the match, good on you. However your healing, dmg, eliminations and other such stats be rather low if you just push a payload… making the system rank you as a worse player than that lone wolf who never got within 30m of the objective but had a lot of kills.
1 Like
According to the formula, win rank go up, lose rank go down. The more sR difference the more the rank changes. That is literally it.
2 Likes
for the most part it’s some mix of
- how you performed vs what was the expected result (you + teams mmr vs enemy teams mmr … the diff would be the baseline for the +/- change before any bonuses)
- certainty, meaning long periods of no play, fresh accounts, or accounts that have wildly changed their MMR recently become uncertain and more likely to change. one hard fought win against better opponents alone isn’t enough to rocket you up in MMR, if there’s a mountain of evidence to suggest you’re not of that higher rating. people have varying opinions on how much work it takes to become “uncertain” but i think ~5 games of severely overperforming and winning is enough.
- statistics (performance based mmr, but only below 3k sr) meaning if you overperform in certain unknown metrics relative to people at the same rank, you get a bonus. personally i find just a winstreak and an abnormally high k/d is enough to trigger huge bonuses.
- whether you are grouped or not, because the game does try to compensate for groups. it was notoriously aggressive with it a few years back
if you want some good in depth explanations for how mmr works in games and why look up posts in the competitive section from taleswapper 
This game uses sR you are mistaken greatly.
Think of it like a sports analogy:
Team A wins 7 games out of 10.
Team B wins 6 games out of 10.
Team A should be ranked higher than Team B.
Now add in MMR.
Team A gets 2 points per win, for a total of 14 points.
Team B gets 2.5 points per win, for a total of 15 points.
Team B ranks higher than Team A.
And this is why hidden MMR makes no sense - what it does is reduce the rewards for succeeding for some while increasing them for others. Instead of your actual win/loss results deciding where you rank over time, it’s decided by the hidden metric that Blizzard has chosen to use.
What’s worse is that it’s a perpetual loop - the game places you > your MMR is set to keep you at that place > you wind up at that place because of your mmr > the developer thinks their system is good, because it ended up with the only possible result it could’ve.
2 Likes
there’s a snuck premise there involving how you’re awarding your imaginary points that doesn’t quite work
you will find it if you think about it enough i believe in you, even if you are 13 days late 
2 Likes
It’s an example - basically, past performances shouldn’t have any bearing on your reward for current performance. If your favourite team is in a 40-year long drought and finally has a year where they win more games than any other team in the league, they end the league with the top points. In Overwatch, that isn’t true… they get held back by their past performance.
1 Like
KD ratio plays a very important role in both SR and MMR
2 Likes
They won’t say. But Blizzard assures us, it’s very accurate. And always put people where they belong. And they would never lie to us.
1 Like
Carefully guarded secrets.
Well if they won’t lie, why the secrecy lmao
How they do it doesn’t matter nearly as much as the fact that they do it. It’s never made any sense… it makes the ‘ladder’ not at all a ladder - just reward points based on the performance only in that match, like every other form of competition does.
its also a pretty disingenuous example
i will give you a hint
many assumptions lie here
it wouldnt even be hard for me to add specificity that justifies team b ranking higher in your matchmaker 
1 Like
It’s not disingenuous… I suspect you think that word means something that it doesn’t.
I am simply making the point that if you have a competition where Team A wins 7 out of 10 games, and Team B wins 6 out of 10 games, then Team A should get more points than Team B does, not less. That’s it. It’s a very simple, very pure concept. I’m not talking about their total rank on the ladder, or anything outside of those 10 matches. You can even add in an assumption that those 10 matches were all played against each other if it makes the picture clearer for you - but it’s not necessary for the point to stand.
Also a misleading (and disengenious) example because you brand it as simple but hide the details that specify your example to be reasonable. Because there are many that might make it absurd =)
Add specifics my friend
Are they playing of equal skill?
Did they start at the same MMR?
Did they perform exactly as they should have relative to their peers?
Was it a win-loss-win-loss or winstreak situation?
Was the matchmaker equally certain about all of their mmrs? (time since playing, recent abnormalities in how they play - you know, standard matchmaking stuff even outside of MMR systems)
Even then you still have to explain exactly how you’re awarding the mmr and why, and why it’s a relevant enough match to overwatch’s system for it to even be worth bringing up. Can you?
Otherwise it’s as useful as me saying
“No actually the other team would have more MMR at the end of it. Good night” 
See , now you’re actually being disingenuous - you understood my point, you’re just pretending not to. I am pointing out that all those specifics SHOULDN’T matter for how many points you get. What your mmr was before the match? irrelevant. What their mmr was? Irrelevant. How old you are, how many years you’ve played video games, who your third cousin is on your mother’s side… all irrelevant. All that matters is the matches themselves. I’m not trying to explain how the mmr works. I’m saying that it’s a completely needless system that shouldn’t exist. If you want to weight it by the kdr during that match, or objective time during that match, or how close the times were, or any of a hundread other things from in that match - you can make a case for that. But loading factors for the entire history of the players participating in the game? That’s just silly.
Think of it like this… a basketball player scores 40 points in a game and wins. The next game he scores 25 and still wins. But because he scored 40 points last game, they add 40 points to his score total, instead of 25. That’s what the MMR does. It takes past performance, and adds a loading factor to future matches, instead of letting the results of the match stand for themselves.
I’m not pretending not to. I get your point I just think you maybe don’t like the phrase “the devil is in the details” lol
Relevant in an MMR system, irrelevant in a mmr-less system (because it wouldn’t be there) but to that note I’m sure you’re also aware that in an mmr-less system which rates people relative to their peers, there’s usually a system that looks at the ratings going in and the difference
i agree i hope they don’t know about my family tree
But you give an example with a hidden assumption of how it works to deliver a score you deem unfair to use as proof that it’s bad
Bro are you trolling me 
I get your point that you don’t like history being used to try and aid certainty, you don’t have to double down on an obviously bad argument to communicate that. You realize that, right? You’re gonna end up arguing an opinion, anyway. No need to hide it =)