The MMR/SR Handicapping Thread

Okay, this topic keeps getting brought up. People say they’re on a ‘forced losing streak’ or being kept down. Others say ‘git gud and climb’. The truth is, the latter is more accurate. Regardless of whether or not handicapping is in play, if you get better you will climb. This is not up for debate, if you want someone to support the idea that you’re being held back by an invisible hand that wants you and only you to stay your rank, it won’t happen here.

That said, the ideas are not mutually exclusive, and I strongly believe there is a handicapping system in place to extend the duration of your climb/fall. This would benefit Blizzard by keeping you playing longer; as long as you can tell you are outperforming your teammates and lobbies for the most part you’ll know you can climb and be determined to keep trying. If you’re falling, the occasional games where you get carried hard will make your own impact seem better and help motivate you to continue playing or pretend the fall isn’t deserved. Game makers do not want their games to be unfair for the sake of being unfair, but they do want to create an addictive experience and this would accomplish it.

Now, onto the meat of the topic. The system is obviously not tracking specific players and deciding to keep them down, so we need a clear and succinct outline for what is going on before we can try to support it. I propose the following 2 hypotheses:

-A match is created by selecting 12 players of similar SR, without looking at MMR.

-The players are then divided onto the two teams based on their MMR distribution.

Now, I cannot directly prove this without access to Blizzard’s source or an inordinate amount of data. But, we have some information that can be used to support it. If you’ll pull up youtube and look at bronze to T500 climbs, you can notice one thing:

Even when playing in a rank where these players are absolutely stomping, they are always matched in games with relatively equal SR while in the high population ranks(high silver through low plat).

We need to reconcile this with Jeff’s clear statement that matchmaking is not done with SR, but MMR. This leaves a few possibilities:

-Matchmaking is done with only MMR, and MMR follows SR so closely that the SR distribution remains equal.
We can disprove this by examining these player’s SR gains. They are gaining far more than PBSR allows, indicating the system knows their MMR is much above their SR.

-Matchmaking is done with only SR.
This is easy, Jeff said it’s not true, so let’s rule that out.

-Matchmaking is done by selecting the pool based on SR, then distributing based on MMR.
Interpreting jeff’s statement regarding MMR’s use in matchmaking can fit this. We know that something is keeping the SR ranges tight, and in these cases it can’t just be MMR. But, we also know MMR is ‘the only value used for matchmaking’. If we are to interpret the statement by assuming matchmaking is the process of distributing the pool of 12 players across 2 teams, we are not in contrast to what Blizzard has said.

Now, this is not proof. It is a theory. But, I consider it to be strong evidence. A GM or top 500 player would clearly have MMR far in excess of their rank, which would force the matchmaker to get higher ranked opponents if it were selecting from everything for the pool. This is clearly not the case, you don’t see these climbs getting significantly higher ranked opponents despite winning streaks of 20 or more wins by actual GM players in gold. I believe the most likely explanation is my earlier idea, the lobby is selected by SR and the players are distributed onto teams by MMR.

Why does this matter? Simple, if we can accept the prior 2 hypotheses, which as I will remind you are:

-A match is created by selecting 12 players of similar SR, without looking at MMR.

-The players are then divided onto the two teams based on their MMR distribution.

we can prove the existance of handicapping, and explain why it makes games feel bad and streaky. Given any example game, if the system’s goal is to equalize based on MMR, the average of each team will be as close to equal as possible.

If each team is equal in the end, and all players are at roughly the same SR:

-A player who is performing well will have worse teammates on average than a player who is performing poorly.

For any sample range, if the net result is to be even, the lowest value will have to be offset by higher values. Take a lobby with 12 players of approximately 2100SR, with an equal distribution of MMR from ~1900equiv to ~2300equiv. Each team will be approximately 2100MMR. This means that as a 1900equiv player, your remaining 5 teammates need to average 2140equiv to reach the 2100 average. As a 2300equiv player, your remaining 5 teammates need to average 2060equiv to reach the 2100 average.

What this means, in simple terms, is that a player who is at 2100 and falling is more likely to have easier games while a player who is at 2100 and rising is more likely to have difficult games. This is a simple concept, and all that’s needed to prove it is the earlier 2 hypotheses, both of which have evidence in favor of them.

As far as streaks, the game doesn’t have an innate desire to put you in streaks. Again, assuming the prior 2 hypotheses and following the logic, we can outline why they happen pretty easily. When you are performing much better than normal, you’ll raise your MMR. This eventually gets to the point where your MMR exceeds your SR, resulting in more frequent poor teammates or difficult opponents. In these games, you’ll have a harder time maintaining that performance because you won’t have the teamwork and competence you would in fair games. Your stats drop, and eventually it swings back the other way.

You cannot get in a state where games are unwinnable, as variance between lobbies will ensure that some games are relatively fair(if everyone in the lobby is close to their appropriate SR, the system is harmless and games will be fun regardless). Where it becomes a problem is in the middle ranks, where smurfs and new players abound. By simply overperforming a little, you can drastically increase your odds of being paired with a new player who’s MMR has not yet fallen to the level it belongs, or a second account who’s MMR has not yet risen to the level it belongs. These inaccuracies will result in stomps, as a player who is legitimately climbing or falling cannot be expected to offset the impact of those players.

‘But Thorny, bronze to 500 climbs prove it isn’t rigged! Krunkenstein gained 80 SR on his tank this week! Whatever anecdote!’
This all makes sense in the context of the system described. A player who is performing well based on PBSR will always go up, they just may take many more games to do so. As long as they are able to still contribute and avoid feeding in the games that are stacked against them, they will eventually regress to the norm and have opportunities to grab large amounts of SR in the more fair games.

A player who is top500 or GM is simply so good that most lobbies will not have anything capable of balancing them. It doesn’t matter if it’s them and the 5 worst players against the 2nd-7th best, the match is still leaning in their direction because they are simply that good.

So, git gud, whatever. The idea is here. It’s been here for years. All I’ve done is make an effort to look at it from the perspective of a software developer and outline how and why it could exist. Personally, I follow the evidence and it appears to be true to me. I understand that not everyone has the same standard of proof, I just ask that you read and apply logic when developing a counterargument.

As a closing note, these topics have been going for years, but Blizzard has yet to come out and say it’s false. They own patents for systems that can do this. Most threads oversimplify this to a simple ‘If you are good, you will climb’, which is absolutely true. That does not mean matches are not handicapped, and it is in the interest of everyone who wants fair gameplay to objectively examine the evidence and consider applying pressure to Blizzard for answers.

12 Likes

A forum user named pyah says mmr is constantly chasing your sr, and the way to avoid having ‘bad teammates’ is to constantly be doing the most work in game, on fire as soon as possible, as many golds as you can get, etc. Which is different from what many on this forum have said regarding ‘medals don’t matter’ which, tbh, I kind of agree with pyah’s take on the matter.

I find in some of my games if my hitscan damage player is making their shots, I’ll position myself somewhere to stop them from getting dove on, crept up on, etc. and while we do win the game, my stats aren’t that impressive. This is usually followed by one good game followed by a string of unfortunate games. Which doesn’t make any sense. Why do you get punished for enabling your teammates?

One game is not likely to change your MMR enough to cause any sort of retaliation. It would need to be a trend. It is not impossible that the system is working as Pyah described, though it offers much less benefit in terms of addiction and sustain and is very difficult to rationalize with your teammates.

For it to work in that manner, it would need to decide certain players are overranked based on their low MMR relative to their SR, and try to make a game where they are more likely to lose. I personally do not see this as anywhere near as viable as the system I described: coordinating a win or loss that will advance everyone toward their target SR is extremely difficult and does not elongate your climb.

It is much more likely that this is observation bias, and the real consequence of playing with lower stats is that you’ll get poor PBSR and climbing will become difficult even with a positive or even winrate.

I see a lot of truth here but you are forgetting things. The system is made by people it’s not perfect and even if it were perfect it still couldn’t predict bad internet, sudden throwers and boosting. There is just too much randomness for a fair match even if 50% was forced.
Also not every game is winnable even the most optimistic top 500 player knows that.

1 Like

Are there any sort of measurable tests I can do to confirm or rule this out?

1 Like

Unfortunately, the bar for meaningful evidence is extremely high. Between personal observation being loose, hidden profiles, hidden teams, it’s not really possible for ingame data to be collected to substantiate or disprove this type of thing. Before private profiles, it may have been possible to do so with sufficient effort and time(though it’s still thousands of games to establish a pattern that can actually support a theory like that).

The best information comes from what Blizzard has to say as well as looking at examples far out of the ordinary to see how the matchmaker behaves in those examples. We can draw loose conclusions from these.

1 Like

It seems like they prefer MMR over SR what is a big problem. If your level is big and can have a skill enough to climb from silver to plat, then you get plat skilled players in silver to “prove”. It shouldnt be the case. Let players play on the rank they are currently. If they carry, they rank up. You dont have to get even matches if your skill is better than the actual rank you are in.

7 Likes

Its basically not letting noobs get stomped like apex legands every single game…

I dont agree skill should rein and smurfs should be punished.

And team q should be a thing.

132 feels so much better with carry potential and thats why dps qs were so long bc ppl want to prove skill which is impossible with only playing the game to press q

2 Likes

Completely agree with this. I’ve been saying this in fewer words (and not explained as well as you) for years.

this is roughly what I have believed for the last year or so. I have never thought that with skill and good performance you wouldn’t climb because I have already gone to diamond with too many characters on too many accounts. The main thing that has changed for me in the last year is looking at it from the developer’s perspective I think of it less as “handicapping” and more as “making fair games”.

I have a long history in competitive gaming and before the development of these systems you just had smurfs completely crushing the worst players every game and then bad players quit. Making fairer games is harder on the better players, but it helps retain the player base.

Also the better players still climb. It just takes time. But understanding the system is also partly why many of the better players just give up and troll/throw in games that they feel are unwinnable. They look at their teammates and say “this game isn’t winnable.” Whatever I’ll drop back down and try next the one.

My own experience is that you can’t let your performance drop much if you want to minimize mmr loss in losing games so I just play my best regardless most of the time.

Anyway keeping bad players playing is also in the interest of the developers.

It also means that when I hit diamond as lucio last season and don’t play lucio for a season and come back, I can’t quite keep up but i don’t immediately fall a huge ways. The better teammates act as a cushion on my rusty play

This post sums it up pretty darn well, and maybe back in S3 or so started looking at it as “Forced Losing Streaks” but, yeah, it’s not quite that.

First off, Blizzard is looking to create even matches above almost everything else. After much deliberation, i’ve come to think that this is a GOOD thing.

I don’t mind even matches any more. I am actually glad that Jeff and the team sat down and said “You know all those online games where you get spawn camped? That will not be Overwatch!”

I like that. But it means that wins and losses are not everything, and you’re being watched.

Papa Jeff sits in his underground fort in the Burning Steppes, drinking from a large tankard of salty tears, watching every single game played to figure out where YOU rank. (half that drink is made from my own tears, btw) If you do win, he yells “THAT MATCH WASN’T EVEN ENOUGH! (that match wasn’t even enough!, the echo yells)” and he sends his peons to press some buttons, and bam! Tougher match next time. Then he strokes his fiery Molten Core hound, doesn’t even burn his hand.

Actually, that’s what he used to do. Then, he realized it was too much work. So, he built a computer to do it for him.

But the computer is mostly trying to make an even match for YOU. Not someone on the forums. YOU. It’s YOUR match, built to be even with YOU in it, by looking at YOUR MMR. I feel MMR is made up of a million things we might not even suspect. Every little thing you do effects what kind of matches you will receive.

That leads me to believe that OW can be quite a different experience for any given player, even in the same general rank. It’s subtle, but can be felt by players over a long period of time. I think this is due to their stats, their play styles, their habits.

Take me, for instance. I play with the same duo partner for 3 years straight, on mostly the same heroes. So, the matchmaker feels extremely confident while creating and even match with us in it. We often see Overtime games. It’s almost always a struggle, where you feel you have to carry if you have any hope of winning. If either of us go to an off hero, we have almost no shot of winning at all. That match was made with our usual stats in mind, and it’s going to be as even as a computer can force it to be.

The game seems to have an iron grip on how to make me work my butt off just to get to Overtime. A lot. I win some, the game seems to overcompensate a bit, then i lose some. It starts to feel a bit rigged. Forced. “Stinky.” Whatever. But like the OP said, it’s just trying to create even matches. You win, you’re counted a bit higher and the next game’s tougher. You lose a few and it does ease up on you, so you can win some more, and repeat.

So, how the heck can you rank up?

I think the key is to just find your own better teammates. Group with some people. I tried it last night with a user here on the forums! Very good player. The matches felt totally different, and more than just him being good. Filled with different sort of players than i usually see. I got to play with a TEAM, and didn’t feel like i had to carry or lose. Going into a match with two pre-made supports was huge, and made it so we didn’t get that random goofball off trying to solo kill things rather than helping the team. Just… take away the matchmaker’s ability to make it hard on you.

I know people don’t often recommend making a 5 man, but thinking back… grouping up were the times i actually ranked up quite a bit. Just some random players who were stuck in the same rank i was. We stayed as team one random night, and went 14-0 over the next couple of days. Because we were all decent players sitting in gold, finally getting to play with, well, non-throwers? I had this happen a couple of times, and then grouping last night felt like different matches than i usually see.

Guys, i think we’re supposed to be grouping in this game! Took me this long to figure out this radical concept. I now believe it’s a huge step towards “breaking” your monotonous matches where you win some, lose some, forever.

I think this is actualy the case. I dont think MMR can be too far from SR. Seasonal placement was always good for corecting your SR so it can be placed close to your MMR. I dont know about others but my highest jump ever in placemet was +373 SR. I had very high stats before placement for the rank I was in and you have to count that there was 10 placement games, so probably every single game was correcting, amping up the MMR even more. So yeah it hard to tell when we dont see MMR, but I dont think SR can be far from MMR. I cant probably prove it but as a player who played on multiple alt accounts doing different kind of experiement,s I can honestly say its that.

Thats why you wont see for example streamer tesla on bronze to gm only moira challenge to play with higher ranked people than he is. He can be like 1000 SR with insane stats but he is still playing with and against people in same SR range. You can see them on his stream, they are all 1000 skilled players. He is performing amazingly well and because as I said MMR cant be too far from SR, his very high SR gains are trying to push him higher enough to catch his MMR which cannot go higher on itself.

What do you mean by that? Highest PBSR you can get it +150 or -150, I experienced both in past. Or not exactly PBSR but simply maximum SR loss or gain. It was like this even before role q btw. And I had these number way past the initial placement SR bonus.

Why should 1900 MMR player have more easy game? Just because his other team mates are slighty better? He will not have more easy game at all. I dont understand your logic why he should have more easy game. It will not give him more easy game, it will give him only slightly more skilled team mates, but how does it matter? He will still compete with enemy team who has the same MMR composition. You are kind of forgetting its not you vs your team mates, its your team vs enemy team.

Of yourse you can get into unwinnable games, top players losing games in bronze are proof of that. With system which can be abused by player its completely possible. People can have certain mood, they can be one tricks playing bad hero vs enemyteam comp, they are toxic, there is a lot of things and if too many of these thing will affect your team, tought luck, game will be pretty much unwinnable.

Yes, of course but its not only just Gms, ever plat will have same effect in 90% bronze and silver games.

What do I get from all this? I think system is doing poor job into placing people where they belong. Or more likely doing that at too slow pace. Quickplay and other modes are actualy much better at guessing your place in ranks than competitive is. Competitive mode is doing that way too slow and even if you belong higher, you will be slowed down by system. But maybe its on purpose because with faster climbing, there would be faster boosting and deranking too, so idk which is better.

1 Like

So it’s like WoW’s ranked system? Your SR is trying to be equal with your MMR. So if you have for example 2000SR and have an MMR of 2500, when you win, you win a lot of SR. When you lose, you’ll lose just a few SR. But when your MMR is equal with your SR, or even lower (idk if possible) then comes the thing called “elo hell”. When you win, you’ll win a few SR, and when you lose, you’ll lose a lot. Hmm.

This pretty much sums up your first bit. We need more data for it to be anything other than my thoughts vs your thoughts.

Maximum gains or losses require a MMR difference, as far as I can tell. I’ve had single games where I racked up stats on the level of these climbers(I’m sure most people have had single games where they did amazing with few or no deaths). However, I don’t gain more than 30 SR.

I’d be interested to see you pull even 50 SR on someone’s bronze account that has low MMR, it would lend a lot of credibility to your argument.

It is a matter of perspective. Sure, the game will be fair in both cases. But, if both players are 2100SR and one is being given teammates that are intended to make up for their underperformance and the other is being given teammates that are intended to hinder their overperformance… that’s exactly the definition of handicapping. Whether it’s good for the game or not is subjective, but it is handicapping.

This is just a misunderstanding, of course some games are basically unwinnable. I said you cannot get into a state where games are unwinnable, meaning no matter what happens with your SR/MMR you can get winnable games. The variance in lobbies ensures that some games will be winnable no matter what your MMR/SR are currently at.

Seems like some progress is being made here.

I play against Masters and GMs in quickplay and Arcade but I’m stuck in gold on my DPS. If you legitimately believe that then there is serious handicapping happening. Hell I even played against a lot of Masters in Experimental. It was super fun.

PS: I in no way think I’m a Master dps. I think I’m platinum maybe very low diamond.

I agree, well I think I have more than enough data but I still agree more would be better.

+77 SR

+72 SR

+92 SR

I did write it wrong tho. Quickplay and some others modes are usualy good at placing yaou where you belong only based on your performace without other thing which would affect your rank like communcation skill, being under preassure in comp and others stuff.

2 Likes

Those games are same context as bronze to t500, you place in bronze on a low level account and straight climb so same debate about whether MMR is high applies.

I’m saying that I do not believe you can gain that kind of SR on an account that is already in bronze by just performing well for one game, MMR has to be component. If that is not the case, one game on an established bronze account with at least 50 SR gained would be very indicative of such.

1 Like