Competitive matchmaking's MMR system is VERY good

Competitive Discussion
1

The mmr system is fantastic at finding similarly skilled players.
It’s also absolutely fantastic at sussing out smurfs. I have an alt account that was in QP with diamonds and plats after about 10 games. It’s great at this.

There’s no argument that it’s quite good at figuring out a player’s potential.

The issue with unskilled players attaining a high rank lies in the SR system, and the way matches are made.
There’s a real reward and incentive for reaching 3000SR, and that is to participate in contenders.

Blizzard should want players that deserve that rank to reach that rank, and they do take major action against boosted players: as evidenced by suspensions in overwatch league.

The problem of boosted players is huge, and blizzard thinks so too. Poorly skilled players are being kept in their boosted ranks near or close to diamond, and in other ranks, thanks to highly skilled players duking it out at every single rank during prime-time and not advancing, since matchmaker can find similarly innacurate sr opponents for those outliers. We need to stop this, somehow.

tl;dr:

  • mmr system is fairly accurate, but that does not mean that any given player has obtained an accurate sr due to previous seasons or boosting.
  • mmr system gives you opponents based on similar mmr, not on their sr tier.
  • mmr system does not seem to give you similar mmr teammates, conversely.
  • differences in team comp and luck of the draw player hero pools at the start of a match cause strange unpredictable results or slow the process of sorting players into appropriate ranks.
  • therefore: Matchmaker attempting to create fair 50/50 matches actually prevents players from being accurately placed either higher or lower, because they have nobody to beat or lose to at anything other then a coin flip, since you’ve been put in a 50/50 match.

There’s only one issue with the matchmaker, in my opinion. It doesn’t match your teammates based on MMR, and even if it does, it also matches your opponents based on MMR. Since SR isn’t taken into account at all when finding opponents, there are players with a high mmr fighting amongst themselves to leave what might be a lower or higher rank then they should be.

If the very good matchmaking system instead matched solo queue teammates, and THEN simply placed a matched team against another team based on average sr as the ONLY METRIC, the cream would rise to the top much faster and boosted players would fall immediately. It’s as simple as that, and in fact it would correct the issue that pre-mades have currently. Blizzard could even leave the mmr matchmaking system in place for premades only.

However, since the system will currently place you against equally skilled players at whatever sr tier you happen to be in, your experience at your true potential rank will be the same in any rank, unless the matchmaker has trouble matching you.

It also seems the longer you play compared to other players, or, if you improve significantly from one season to the next, the more the matchmaker gives you boosted or variable players as opponents, likely because their mmr being a hockey stick seems similar.

5 Likes
2

If one of the teammates is incorrectly placed, then the match isn’t actually 100% uncertain (50% chance) it’s actually tilted in the direction of the person placed incorrectly. It’s only if people are correctly at the MMR that the matches are truly 100% uncertain before they begin.

You are right though. The system is very good. It just has some assumptions the people tend to break, in that you aren’t boosted, throwing, or learning a new hero. Since these things happen, it causes noise in the system.

The match based on MMR, not SR btw. If they didn’t decayed GM’s would occasionally play with low Diamonds. They don’t. They just get placed where they’re MMR says they belong and the SR eventually catches up.

3 Likes
3

im having an awful time with the competitive so I have to disagree that its very good.

to me its very bad and involves too much luck with overshadows the skill of individual players. Placing you and keeping you in tiers you know you are above in terms of skill and ability.

3 Likes
4

It’s only goal is to find 11 people of similar skill. Skill being defined as how well they performed in the past.

If you’re asking it to do more than find 11 people that have in the past performed as well as you, you’re asking it to do more than it was designed to do.

1 Like
5

I know, and that’s the CORE of my argument and the problem itself… It might nudge your lost SR or gained SR SLIGHTLY based on performance, but if you still lose due to being poorly matched against another “ringer” there is a problem.

A player’s performance is directly related to how enabled their teammates make them. I’m a -godlike- Zen because I play hitscan dps primarily and I have 16+ years experience. However, if the other team gets smart and dives me I go from an average of 4-6 deaths per game to 10-11.

Stats are unpredictable. Yes, maybe I get 57+% accuracy on mcree most days, but I might play him for one life and get instantly popped after missing a couple times because someone switched off to widow while I was respawning too. How does that factor into my hero stats? How is that going to have an effect on my MMR? I bet it isn’t good, since it massively modifies the visible stats on your hero page. That’s not right at all, and there’s way too much noise in the system.

I believe it’s entirely possible to completely game the matchmaker on a new account into believing you’re whatever SR you want by abusing this. Junk and old data needs to be thrown out. Previous performance after a certain point should absolutely NOT be a factor on a hero, ever.

There are players that absolutely should not be where there are and because you are matched on MMR and not SR, the challenge is the same at EVERY RANK unless the matchmaker is convinced you’re terribad when you’re not.

I know. this is my entire argument and it’s a pretty big problem for everything below diamond. Above diamond it’s fine.

6

I don’t get to play OW much and when I do I don’t generally play comp because I’m generally too tired (or want to drink a beer…or both) to play in a game that people are taking seriously (wish others had that same level of respect, right?).

I did, however, in the last few seasons really try to increase my game knowledge and mechanical skill (notably dropped my sens dramatically).

I’m one of a few people that has actually increased a BUNCH through placement matches alone. I’ve pretty much gone from bronze to gold via placements and a few games.

I used to think that past performance has a lot to do with it, but now I’m not so sure how far back it really goes. I think not as far as most people believe.

What I do think, though, is that I’m not convinced at this (prevalent) idea that matching based on MMR is any worse than matching based on SR. I’m not sure how else it would realistically work. No matter how you measure the skill, if it’s accurate, and you put teams of the same skill together, the chance of either winning is going to be 50%, right?

Now, if you’re saying that MMR using PBSR measures skill wrong, that’s a fine argument to have. But that’s not a problem with attempting to find a 50/50 game, that’s a problem with measuring what that 50/50 game would look like.

I truly, really don’t understand how “Matchmaker attempting to create fair 50/50 matches actually prevents players from being accurately placed either higher or lower” unless you believe that they are matched both on MMR and SR.

SR is a facade. It will go up as MMR goes up, it just lags behind it to look pretty and not move too much or in the “wrong” direction. Otherwise it would be obviously broken. If I was at 4900 SR and still matched with MMR around 2000 range, my SR would bounce around 4900. This both doesn’t happen AND is obvious to math nerds, which the people who develop this system likely are (as am I).

I’m fairly certain that attempted 50/50 matches is the ONLY way to measure skill.

I’m a bit uncomfortable with the PBSR aspect of it, but I personally doubt it’s really that big a deal at the lower ranks. I could be wrong though, I’m pretty agnostic about it. I think they keep it because the learning curve to go from, say, 1000 to 2000 is easier than going from 3000-4000 and they want to make the MMR value move faster by using PBSR.

7

This is why: your TEAMMATES are not picked based on YOUR skill. They’re placed with you based on SR, and their OPPONENTS are chosen to match THEIR skill. You also have ONE opponent that is equal to your skill.

The fact that you can’t see this bothers me fundamentally. It’s really simple. Since you’re into math like me:

If a match chance to win is presumed 50/50 and the only way to gain SR is by winning, you can and will have truly diamond and above mechanically skilled players being matched in MMR ACROSS ALL RANKS.

A diamond player should not have to face a diamond player if they are ranked in bronze. Ever. They should DECIMATE and the system should give them win after win until it’s corrected. If anything, the SR someone like me should gain for a win should be 60-100 -PER MATCH.-.

This is not my only account. I was ranked diamond in season 6, when this account has NEVER broken silver. I participate in local tournaments against grandmasters and hold my own. My town hosts a collegiate overwatch league.

I started 6 stacking to break the trend and soloing in with off meta heroes of the rank and just pooping all over people instead of flexing for the good of the team, and I’m gaining on average of 100-200SR a night, but I still encounter the rare counter, and games are just as hard in gold as they are in diamond, and sometimes moreso, because teammates can’t play to my level, but at least ONE opponent can, every time. I have to suss him out, switch to his counter, and focus him so he can’t do what I’m trying to do to his team.

You know something is up when the majority of enemies you encounter have no stars, all smurfs.

8

Wut? I’m tempted to ask who the “they” pronoun in this statement is referring to, but either way, how would someone be a diamond player if they are in bronze short of throwing, buying a bronze account or otherwise bad behavior?

This is where your mistake lies…

true, you only “gain” SR by winning, but your MMR can stay the same on a loss. The next time you win, which you will at some point, your SR gain will be bigger. Alternatively, your MMR could go up more on your wins than lose on your losses.

That’s the whole “SR chases MMR” thing and why they have a separate number in the first place…

the reference for this fact is here: https://us.battle.net/forums/en/overwatch/topic/20758686566#post-16

Like I said, SR is a facade. It’s not real. It’s just a pretty representation of your MMR. There’s not such thing as a “Diamond MMR player in Bronze”. A Diamond MMR player is in Diamond BY DEFINITION. Now, a Diamond SKILL player could be in Bronze, but only through bad behavior or perhaps copious amounts of alcohol.

Also,

is not necessarily true and it directly contradicts what you have already said.

I’ll assume you meant “They’re placed with you based on MMR” there’s every reason to think that the first attempt is made to find 12 players of similar MMR. Yes, when the MM can’t find those people you can get unfair matches, but there’s no reason to think that happens in peak hours for players of a common skill tier.

Maybe you would be willing to wait longer for a better group to come up, that’s a legitimate debate to have.

But it certainly doesn’t show that “Matchmaker attempting to create fair 50/50 matches actually prevents players from being accurately placed either higher or lower”.

9

What MMR? matches are still one sided, played a good 5 hours last night and matches were nothing but (except for one) ether not getting to point A or being that team that stomps them at point A.

2 Likes
10

I think you might be making a leap on this one. The post says the following:

For example, it’s possible to win a match and not gain any MMR.

I think that also means you could lose a match and not lose any MMR, but it doesn’t necessarily mean you can gain on a loss or lose on a win. It could be hard capped at 0 change to MMR.

Feel free to point out if I’m picking the wrong line or missing something to the contrary.

Side note: I posted in that thread on the old forums like a cool kid. :sunglasses:

1 Like
11

Fixed with less than 20 characters, thanks.

1 Like
12

You’re number one mistake is forgetting one simple fact about SR that solves this entire problem: It gets bonus gains to rapidly catch up to MMR, if they have diverged. I.e. If my SR is below my MMR. I will gain more SR on wins, and less on losses, until I catch up.

This is most visible for decayed players, but happens to everyone. In previous seasons, they intentionally lowered everyone’s SR only after placements. The entire point was to abuse these “SR gains” to allow EVERYONE in the game climb SR (at first). Obviously, not everyone can win, the average win-rate must be 50%, but this let everyones early SR trend up. Turns out, people were more mad about having their SR lowered in the first place, so the idea was scrapped, but that’s a different topic.

I don’t really believe you can be at some level and, just through playing, shoot your MMR up 500+ equivalent SR or anything, but that doesn’t matter. Even if you did, there would be no way for you to stay at your SR, unless you lost far more often then you won (which would also drop your MMR down again).

Given your premise that MMR is accurate, and the fact that SR always chases it, that implies SR is in the long term accurate. Though it’s certainly possible to have short term fluctuations (I mean, who hasn’t played tilted and had a bad loss streak?)

2 Likes
13

I think you might be making a leap on this one.

Currently complaining isn’t the same as more mad. It’s just the people that want to see progress without actually playing better that are mad now. They should have just kept their mouths shut about it and everything would have been fine.

14

Show me the evidence of that, it simply isn’t true. You can see it clearly when teams of two or three are together. Players who are hitting above their weight get paired with the most random disorganized bunch, but the same is true on both sides. If you’re lucky enough to be the counter to your doppelganger, you’ll win.

There’s no real skill involved except in a full stack, you’re just throwing the dice.

And, even with sr being subtly adjusted from game to game, you WILL hit a rut eventually and undo any of those gains by losing just two or three in a row.

15

Alternate theory.

https://en.wikipedia.org/wiki/Dunning–Kruger_effect
1 Like
16

Unless that group of two or three happens to be nearly the same MMR, then it’s the PLAYERS that are causing the issue, not the MM.

The FIRST attempt is made to find 12 players of similar MMR. If the starting shot has 3 people 1000 SR apart, WT* is it supposed to do with that?

They’ve already stated that groups mess with the MM system. It shouldn’t be surprising that groups mess with the MM system.

People that expect it to find a perfect match for their “snowflake group” as Jeff Kaplan himself put it, are asking too much of the MM.

But groups messing with the normal functioning of the MM has an easy solution, Solo Queue only, but of course that has downsides as well.

That is a legitimate discussion to have, but it doesn’t show how “Matchmaker attempting to create fair 50/50 matches actually prevents players from being accurately placed either higher or lower”. It does show how allowing groups may allow players to be inaccurately placed though.

17

Thank you for that concession, at least.

Look, when tools like OverSumo exist this argument that the matchmaker and MMR and SR systems work gets defensitrated. I consistently, regularly (except on weekends), play at or above 54%-75% of all players at all ranks consistently, every single game I play at or above players above my rank by a wide margin.

Without fail, when the game is populated with MANY users and the matchmaker has no trouble finding a game within seconds I will derank. Why? Because the players ON MY TEAM, despite their being awful players on the other team too, are terrible and they can’t counter my counter.

During prime time I get all gold even as a support. During off peak hours I have to work for it - but I win more often than not - , and the players on my team are of a much higher caliber or smurfing without fail.

I have more then enough screenshots and evidence of this. Plus I’ve got oversumo - which tells you your performance for the last 5 games. The SR is a seasaw of climbing during the week, then endless loss streaks on weekends. My consistency of play never changes.

All I’m asking for is teammates who are as good as me or a role queue At least with that, I can feel like the matchmaker is being fair.

18

The matchmaking is accurate. I’m just sure that there are PLENTY of other players within atleast my skill rating (hidden MMR and/or SR) that I could’ve just queue’d with, but matchmaker be like “NO! EVEN IF YOU TAKE A BREAK FOR AN HOUR, YOU PLAY WITH THE SAME PEOPLE AGAIN!”.

Its just not that good at spreading people with other people, IMHO.

2 Likes
19

The matchmaking is accurate

Yes it is. That’s the problem. It’s accurate at EVERY RANK, meaning there are players like me I have to FACE OFF AGAINST at every rank.

This is correct. You get the same people due to ping - the game tries to match you with players with low latency to each other, and similar MMR - this is why, especially on the east coast after 9pm you see the same people over and over. I’ve been tempted to proxy to central USA just to escape it.

I keep screengrabs of consistent players. I get the same people night after night on the enemy team during the week. On weekends it’s a complete tossup.

Let’s face it, I’m in some kind of high mmr low sr island along with a bunch of other poor sobs who are constantly duking it out. Games feel the same at every rank except during off peak hours.

And besides, I have a second account I effortlessly made diamond on in season 6, because there was -no account history- and I only played during low player populations to prove the point.

On this account, I made plat in season 1. I didn’t finish placements in season 2, and after finishing placements in season three, it put me BELOW 500 (I bet it was reset to 0)- if that’s not a bug I don’t know what what is.

I’ve had to earn 2000+ SR just to be where I am now, and if you count all the weekends that put me down 150+ SR, I’ve had to earn thousands just to be where I am.

My CAREER HIGH is 2170ish on this account and I have 3 golden guns, and I still have a bronze portrait. What does that tell you about how many games I’ve won and lost?

Games were HARDER at bronze then they are now. Games in DIAMOND were EASIER than the games I play in silver.

if you want an opinion with a bit more clout watch the latest video by your overwatch.

youtube watch?v=ptSP9MeOVyU

Their motto is literally rank doesn’t matter. Because it doesn’t. Stats matter, and I have them.

20

I looked at your OverSumo.

It says you’re average, a bit above average on occasion, in Gold.

I don’t think you’re lying, I think you’re misinterpreting OverSumo.

If you think this is true you’re lying to yourself. Everyone’s consistency changes. Unless you’re a robot, of course, then I apologize and I welcome my new robot overlords.

This isn’t a concession. This issue is well known and acknowledged by Blizzard. A claim that is not one that you are making.

No one has yet shown how “Matchmaker attempting to create fair 50/50 matches actually prevents players from being accurately placed either higher or lower”