What's the New Deal with MMR?

So the second option? You think they’re deliberately trying to confuse the matter?

My understanding is that they are actually talking about both. This is also a stance that could be supported by the “closely linked” quote I’m sure you’re aware of.

This is another quote from the same thread just to illustrate MMR is undeniably being discussed:

So then why do points for losses and wins seem so random? Well, the amount of MMR (and SR) you go up or down isn’t simply a matter of whether you won or lost, and what was your predicted chance of winning. There’s a couple of other things at work. One is the matchmaker’s confidence in what your MMR should be. Play a lot of games, it gets more certain. Don’t play Overwatch for a while, it gets less certain. You go on a large win or loss streak, it gets less certain. The more certain the matchmaker is about your MMR, the less your MMR will change in either direction based on a win or loss.

If you put all of their sources together from the past, we would largely be led to believe that for most of the population that are operating under normal conditions (active accounts, no large streaks in recent history, etc.), their MMR and SR would be close enough to one another that they would essentially exist as being the same value.

As for how that is in the current game with everything that’s changed or how the latest post on matchmaking reads… I honestly don’t know what they’ve changed or have any worthwhile amount of confidence to argue too much on matchmaking topics. :frowning:

2 Likes

This is contradicted by Overwatch Forums in which MMR is described as a single number. Not composed of something and SR.

“Anyways 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.”

It is contradicted by Overwatch Forums in which SR is being described as being pulled in the direction of MMR. If MMR depends on SR and SR depends on MMR, that would be circular.

“When you do come back and actively play matches, you’ll also typically gain more SR from a win until your displayed skill rating and internal matchmaking rating have again reached ‘equilibrium’.”

That isn’t “very closely” (Overwatch Forums) which is the main post that your theory contradicts.

This is my old theory, and reflected in my current summaries. But I may change it. We’ll see.

I’m going to disagree with you on this one.

For decay, yes. Other modes, use a match making rating, in the generic sense of the term, but they are not linked to competitive, so outside the scope of this discussion.

But it’s getting harder for me to justify, as Scott persists in acting like MMR is no longer a thing (or uses the terms MMR/SR interchangeably, when, strictly speaking, they are not).

Sheevah and OzoneOOO have given you good answers.

As has been said countless times (but I’m not sure if to you) is that they don’t “force” 50%. They use MMR to predict a 50% match and use the result to update the MMR of the players. Which makes all the difference. The longer version is at How Competitive Skill Rating Works (Season 10). See especially Summary → Summarize matchmaking, rating, and progression for me and Popular Myths → Matchmaking pushes a 50% win percentage using broken criteria.

Void argument as SR can be part of a formula that results in a single number MMR.

Alternatively, if you do want to believe there is a two tiered process for match making:

“All the system does when it comes to matching on skill is attempt to match you with people of a similar number”

Can also be interpreted that the non-SR skill part of the matchmaker system matches purely on that “similar number” (which could be considered the true MMR) which again would support handicapping.

You appear to be interpreting that sentence as if it actually reads:

“All the system does is attempt to match you with people of a similar number”

That qualifier in the middle you appear to skip is actually pretty important.

There are many ways to avoid it being circular, again this argument is null.

Decay isn’t a valid part of this argument anymore - there are other ways that it can be implemented and match the current outcome exactly. It’s as relevant as discussing the MMR of quickplay. For example as decay is only for above a certain SR level, a completely different matchmaking system could be in place.

My theory only contradicts your assumption that MMR=SR. I used the example of a streak to prove that “very closely” as you want it is impossible. And this game is super streaky.

If these are your only problems with the “SR is part of MMR” argument, then you really need to swap sides.

Any idea if the system does this when calculating PBSR and MMR (if it exists)? If it doesn’t, it really should.

When your performance is compared to the population (your rank, same hero), you should only be compared to stats from winning/losing teams only.

Naturally, you’re stats will be better compared to the population if you’re on a winning team, rather than a losing team.

Example using OverSumo as D.Va:

  • Winning (good) teammates stats better than 75-80% (sometimes even higher) of other D.Va players my rank.
  • Losing (poor) teammates stats are only better than 50-55% (sometimes lower) of other D.Va players my rank.

Obviously, my real skill is somewhere in between, but it always seems like I consistently lose a few more SR than I gain when I go 50/50 for a session. Such SR losses are even more pronounced when I play Rein, even if I do my tank job well.

Ah…good old “handicapping”. Another word for “balanced matches”, which the system most certainly does, doesn’t need two MMR/SR values to do it, and is preferable to the “desired” alternative, which is for everyone that is a below average player have a <50% win rate, all games of random skill levels on each team, and most of the games being stomps, and boosting galore.

The “two value MMR” theory doesn’t lead to the idea of handicapping, the theory supports people’s belief that their SR is being held back by their MMR.

Don’t get me wrong, it’s not really an unreasonable question here. If your MMR puts you in matches where you have a 50% chance to win, how do you increase your SR?

I’d explain it, but I know you won’t listen because we’ve explained it hundreds of times already.

Just don’t think that it’s complicated or convoluted or cryptic. It’s really easy to understand if you are willing to learn.

The hallmark of a conspiracy theory is twofold.

First, it must be un-falsifiable. Second, it’s malignant.

You’ve given the first condition in your statements like “there are other ways that it can be implemented and match the current outcome exactly…a completely different matchmaking system could be in place.”

Of course, there are any number of ways we could see the outcome we see if we assume there are multiple matchmaking systems in place. Heck, maybe gnomes draw numbers out of hats to give us our SR.

If you’re looking for little bits of secret words to show what you want to be true could be true…you will ALWAYS find them. I trust it’s clear to everyone that’s what you’ve done here. “Matching on skill”…really? Scott Mercer isn’t a malignant AI. He’s not using over-precise language to trick you. Generally, his language isn’t precise at all when talking to the community.

The second condition isn’t so obvious in this post, but “handicapping” is a code word for “keeping you at your SR so you buy another account”.

I’m not quite sure what you mean by “does this”, but the gist of your post seems to be a concern with how much PBSR affects you.

Regarding SR gains/losses ONLY, like you said it averages out and also it pales in comparison to the main factors which are wins and the difficulty of those wins. Basically, just play to win and you will rise.

I emphasize “only” in the last paragraph because there is some evidence that PBSR affects behavior negatively which is why they got rid of it for Diamond and above.

I’m agnostic on the matter as I don’t really think it matters. You can gain rank easy in the lower ranks by learning the game and earning wins (or so I’m told, I haven’t tried). I suspect they kept PBSR in place because the learning curve is so easy (comparatively). They want people who suddenly “get it” to rise quickly. In my experience this is how it works, too…people plateau then something clicks and they skyrocket. The upper ranks they just plateau, I believe, unless you really put some work in to go from top 10% to top 3%.

If you are maintaining a 50% win rate you will hover around a particular SR whether PBSR is involved or not. You just may move a bit faster depending on how well you played.

Sorry, if I wasn’t clear.

I’m not against PBSR or a performance-based MMR. My concern is that the system compares your performance against everyone your rank on that hero. This skews the results in two ways.

  • People with good teammates tend to perform better.
  • People with poor teammates tend to perform worse.

Think of all the times you’ve had a thrower or just a person playing very poorly. Your stats for elims, deaths and objective time will be much lower than what you are capable of. If you have really good teammates, those stats might even be falsely elevated.

What I think the system should do for PBSR is:

  • When you win, only compare your stats to other players’ winning team stats
  • When you lose, only compare your stats to to other players’ losing team stats

When you win, you’re compared only to winners.
When you lose you’re compared only to losers.

Your idea would be more accurate, sure.

But remember that the stats you’re being compared to involve bad games too. There isn’t some high standard out there of 100% win rates. The average includes your losses and theirs.

Plus…like I said, getting rid of it affects behavior and speed of a rise or fall, but probably doesn’t really affect your actual SR.

Missing the point in a long winded post that didn’t actually put forward an argument (as usual).

You along with a few others are now actually the ones grasping at theories when what I’ve said reconciles everything.

SR is part of MMR.

And yes, if true, then handicapping is real. Also it means games are not balanced at all how you believe.

Bliz should stop being convoluted with their explanations.

They should give it to us straight or not give us any information, not this middle-ground nonsense.

This is the part I’m interested in. It’s a pretty clear difference in description, a change that’s been asked for, and would be a fix of a broken design theory. It’s possible they didn’t get rid of the synergy bonus so much as nerf it into oblivion. It’s a distinction without a difference in practice but may matter to the point of not communicating the change.

It seems like something they’d want to make sure people knew about if they had concerns with people not grouping.

There’s also anecdotal evidence that SR based punishments are essentially treated as decay and bonus SR is given afterwards. They clearly have an interest here in not letting that cat out of the bag.

But ultimately, is there really any functional difference between MMR not existing and having the same value as SR? Either way is going to LOOK to an outside observer that matchmaking is done on SR alone, even if it was based on MMR alone.

I think it’s pretty clear the community can’t handle the thought of being matched based on a hidden value, why stir the pot when the difference only matters to people who get weird matches all the time anyway (top %) and people being punished…

come to think of it…

yep. MMR is gone. Y’all celebrate, YOU DID IT! :wink:

Matches built on SR ONLY ™.

MMR is not gone. It’s too valuable a tool for Blizzard for them to get rid of it. More likely they realized that the concept of MMR (a hidden rating used to manipulate player progress) doesn’t sit well with most players so they stopped talking about it.

When an announcement goes into the level of detail this recent one did without mentioning MMR, or it’s removal, it can only mean one thing – MMR is still there but they intentionally decided not to talk about it.

1 Like

I don’t have anything to say that isn’t repeating myself, so I’ll stop now.

1 Like

The game, as a whole, takes this into account. If it did not, average SR (for everyone) would steadily drift downward.

If this is true, then the game is indicating that you are performing poorly, in a statistical sense. However, tanks are pretty hard to measure with statistics, so they could be wrong in your case, especially if the penalty is small. If you set your profile to public, and share your SR gained/lost and hero played for each game this season I may be able to say more. But then again, I may not be able to say more.

Do you have any data here? I’ve just seen some random forum assertions. And we all know what those are worth.

Yes. Another possibility is that Scott just looked at the forums, facepalmed, and decided that we couldn’t handle the truth, and decided to relegate MMR to an implementation detail. That is, something that exists in fact, but is not relevant and useful information to the public. Similar, say, to how they do matching based on ping. Saying this would just feed conspiracy theorists, though.

What do you think, overall, of the new LFG and grouping rules? I know you wrote that long discussion on the topic, so must have an opinion on what Blizzard actually implemented.

Overall, if I play D.Va, then I’ll lose around 4 SR in a 50/50 situation. Which I guess isn’t that much. I go Rein or Orisa with gold objective time and 50k damage blocked, and it’s 16 SR for a win and -26 for a loss. The thing is I don’t get a ton of kills with Rein or Orisa. Instead, I do my job to make space, protect and have very few deaths.

Maybe not getting medals for elims is the culprit here. If so, seems counter-intuitive for a tank’s performance to be measured by # of elims for SR gain. If I focused more on elims, then I might get more SR for a win. On the downside, it would mean less wins overall.

I’m not blaming the SR calculation for my rank, but it does seem that main tanks and healers are penalized by PBSR when they perform their role as a team player instead of stat chasing.