Match Making Rating (MMR), Handicapping

I suggest using replay features and count all the times you threw the game yourself. I did it and I became quite humble about my rank and kinda improved. Also I’m sure u are insisting in heroes that don’t fit your play style, just like if I was playing genji 24/7 and wondering why I am so bad at the game. Also PM me, I have a deal for you. :wink:

1 Like

Both measures would be considered handicapping! And of course they are related. Like Yin and Yang, as Blizzard would have us believe.

There are varying degrees of intellectual dishonesty and unfounded assumption to all of the counter arguments I’ve received. I’m not ready to call anyone a liar, but some people are being very ignorant. Most notably at Blizzard.

Good read! Thanks for the article and commentary

Wrong

2 Likes

You know, maybe you shouldn’t talk about intellectual dishonesty when your entire argument was solely based on an intellectual dishonest interpretation of Scott Mercer’s post in the first place.

You probably shouldn’t avoid this topic, all the correct answers are in here:

1 Like

What have I said that seems dishonest to you, specifically?

1 Like

a lot of people are wondering if the mmr system is in single digits (which if it is that would explain a lot) which is in of itself a handicap by crappy design

For starters, the entire idea about MMR being a differentiator between “good” and “bad” players at the same SR.

Your quote from Scott Mercer came from a topic where a player was confused about his SR gains and losses where teams were actually stacked in a single direction, one example not in favor that player being 6 solos vs a team of 4 with 2 solos, and another example, this time in favor of that player, being a 3-stack with a 2-stack and him as the solo vs a 2-stack and 4 solos. There was NOTHING in that entire topic at all where anyone, especially Scott Mercer, said anything about differentiating between “good” and “bad” players. Scott never even uses the word MMR or anything with that idea.

Basically, your idea that the system “knows” about “good” and “bad” at the same SR, and that the matchmaker “balances” as such is fabricated. Remember, SR is NOT used at all in matchmaker, only MMR

Jeff Kaplan straight up said that your MMR is a value between -3 and +3 (it can actually be slightly below -3 and slightly above +3). That value refers to standard-deviations from the mean, so all +3s (and above) are the highest ranked players, -3s (and below) are the lowest ranked players, and 0 is the mean (somewhere in mid-Gold). So for example in that video, Seagull, who’s consistently hitting Top 500, is a +3 or above it. (The entire standard deviation stuff isn’t actually stated in the video itself, but when you look up the definition and how it works, you can make this deduction).

1 Like

Okay but the problem with such a minute factor in matches is a problem especially when a low gold and high gold are very different, I always thought it was close to SR in numbers because at least in that I got more balanced matches it was always quickplay and arcade where that wasn’t a factor where I had such a problem. Maybe they should have adjusted MMR prior to making any balancing decisions.

but it isn’t, really early on i think in a dev update they explained the mmr system and described it as a matrix with over hundred variables

There is no such quote. Quarters linked the video (two posts up) with the correct quote. MMR is just a floating-point number, per role. It also has an uncertainty, which is high for inactive or new players.

All i want to see bliz do is remove the mmr system from comp and match people as they do diamond+ …they are matched on sr vs sr no mmr or anything else intervening. Why should diamond+ get treated better then the majority of the population if the game? Just cuz they r better at clicking heads? Thats like saying a rich person is more important then a poor person cuz he has more money…o cant say that there is any handi caping or there is not cuz me nor anyone but bliz knows the trueth and they wont share it.

1 Like

well, maybe i’m wrong and just imagine it, it’s been two or more years… but where do you think that floating number originates?

clearly not by simply adding and subtracting numbers each match, or you could just use SR as matchmaking rating

the matchmaker still calculates the probability of a match being won and makes it as close to 50% as possible

Yes and no. The matchmaker does operate based on statistics. Using higher statistic players as a way to balance matches with lower statistics players. The system isn’t why you are stuck but it is making climbing harder because that’s how it’s designed.

1 Like

Diamond+ all the way to GM are also matched up as MMR vs MMR. SR isn’t used at all anywhere. You’re probably confusing what matchmaker uses with a term the playerbase likes to call “Performance-based SR”, which is very much a misnomer since it’s simply a very small bonus (or penalty) attached to your gain/loss for winning or losing.

See How Competitive Matchmaking and Rating Works (Season 20), especially the summary → “Summarize matchmaking, rating, and progression for me”.

The entire point of 50/50 is that you need to factor in groups. If Competitive was solo-queue only, then it would be dead-easy to create that 50/50.

That’s also why SR still exists as a public face: to make things easy to read and set an easy-to-calculate line for group restrictions (1000 SR spread, 500 SR if highest player in group is at Master, 250 SR if highest player in group is GM plus limit of 2 players in a group).

All these posts about SR, yet when a person on the other team is one tier higher than the team your on, people complain they got a better team. Or when a smurf is on the enemy team, they complain. The fact is, if you played at the smurfs actual SR or above, you’d win anyways.

There are so many avenues to this game that are not understood at lower levels, that people think they are doing great, but are doing terrible. For example, you’re team just lost 4 people, and it’s you (Genji) and your support (Ana) and you both use your ults against the 6 people alive on the other team. If you kill 5 people, and you and your ana die, they still have one person. They won the team fight, you wasted your ults, and you lost that push. The genji would claim that he was carrying, when he was actually causing the loss, and the Ana would complain that the Genji didn’t kill 6 or peel for her. You could be more mechanically skilled than a higher tier, but losing more due to your decision making. Then when you watch the enemy Genji, you’d think your team mates are why you lost.

Playing with my lower skilled friends, they make snap decisions that can be high risk, high reward, and win a team fight. But they are making those decisions when it’s not needed. You could do less than usual and have a high % to win, but instead go for the risk play. A rein hiding for example. Instead of getting the play off, maybe they kill you, and you lose a key point when the enemy has spawn advantage.

Maybe you’re not focusing the same target as your tanks, and always trying to do 1 v 1 duels. Maybe you win that 50%, but a focused target is a dead target.

The point is, higher tier players have a better understanding of how to win (win conditions) and how not to lose. They understand what abilities need to be used by the enemy team to know which target to focus. They have a better understanding of when to ult and how to ult. While you might have better aim or be more mechanical, if you’re not processing all that information, you’re making it hard to win. That is why some games are easier (other team isn’t doing it either or worse) while others are more difficult (the enemy team is managing their abilities better and focusing more).

End result is, your SR is earned, the people you are playing against are more often than not your equals. And when you hit Elo Hell, you’re actually at your correct SR. The only way out is to learn the game better.

grouping almost makes no difference

how does any of that refute my point that matchmaking is more than a simple number?

I don’t think that was the point that Mercer was trying to make in that post.

Now let’s talk about groups and matchmaking a bit. Interestingly enough, there’s actually more grouping going on than you might think! For the next set of data, we’re going to look at all competitive games from February 1st up to May 28th of this year.

  • Only 16% of all games purely consisted of solo players.
  • The most common match composition is 2,1,1,1,1 vs. 2,1,1,1,1, which represents 28% of the all matches.
  • If you are in a full team of solo players, 73% of those matches were against another team of solo players and another 24% were against a 2,1,1,1,1 team composition. Only 3% of the enemy teams were groups of three or more players.
  • If you’re in a duo, 74% of your matches are against other 2,1,1,1,1 teams, 14% against 1,1,1,1,1,1 teams, and 8% against 3,1,1,1 teams. Only 4% of the enemy teams you faced had groups of four players or more.
  • If you’re in a full six player group, 92% of your matches are against the combination of other six player groups; 5,1 teams; 4,2 teams; 3,3 teams; and 2,2,2 teams.

The matchmaker was designed to try and create games with equivalent-sized groups, especially for solo players, and these statistics show that it’s doing a pretty decent job. We believe that games with equivalent groups create the fairest possible experience, and fair games create the best chance of players having a fun experience. When the matchmaker does compromise, it’s usually during off-hours when there aren’t as many available players, as well as at the lowest and highest extremes of SR.

When Mercer made that one post about 50/50 (the one that Cuthbert tries to spin it somehow into an issue about handicapping good-players with bad-players), it was a response to a Masters player posting about an issue about uneven teams that factored grouping in. His two games he had an issue with were something like a 5+1 vs 2+1+1+1+1, and the other one being 3+3 vs 6 solos.

The matchmaker has always tried to and continue to try to keep group-sizes fair enough on both sides as the more recent Mercer post explained, but it can only work with whoever’s in the active queue and only up to a point without making queue-times really bad (and that post was made before Role Queue).

Of course, there was that infamous Brazillian 6-stack way back in Season 1 that took advantage of a relatively-low player count and abused it to reach the highest ranks, albeit it took a while because they NEVER played a 50/50 and thus their gains were slow, but it didn’t matter since they kept winning all the time.

Rewatch the video by Jeff. That MMR is a single number, with uncertainty, per role, cannot be contested. The summary I linked was to address your other questions. But because you either did not read it or do not understand it, I will type some more.

The concept of a player skill being summarized by a single number, plus an uncertainty, goes back to Elo, at least. Read Elo rating system - Wikipedia to get background on the topic.

MMR is more accurate than SR, for the following known reasons:

SR cannot go below 500 or above 5000. MMR has no limit.
There is an SR nerf above 4200. MMR does not have such an issue.

SR is an integer, so subject to rounding. MMR is not.

SR is reduced when leaving. MMR is not.
Formerly SR was reduced by decay.
In the early days SR was reduced at season start.

Therefore, it is better to use MMR to make matches.

Yes. So?