Algorithmic Handicapping (MMR) is Wrong for Overwatch

Competitive Discussion
220

I’d bet they licensed MS TrueSkill. When to read up on it and compare what they said it’s either TruSkill or the spent the resources to build something just like it from scratch.

Honestly, I’m not sure how much more transparency people want and I honestly can’t understand why.

Are you really gonna break out the slide rule after gathering the SR of all the players (which isn’t possible anymore anyways)? Why?

Are you somehow going to get more kills just because now you know that will give you 2 extra SR?

Like, honestly, what useful questions does anyone have that they haven’t already explained?

Unless you think they’re being deceitful in the whole process, in which case there is no information that will satisfy you.

221

If I told you why I wanted it, would you fight for my cause. If I convinced you through empirical evidence that my wants were justified, would you take up my banner?

I don’t think you would, so why answer your question :slight_smile:

1 Like
222

I’d like it to be like chess: where you could calculate each SR change with a calculator or simple program, given only publicly available numbers.

If more information allows ratings to be exploited, then the system should be changed so it can’t be exploited (No security through obscurity).

This would make conspiracy theories even sillier, allow us to do bug checks, simplify and make more accurate my summary of how the system works, allow people to know what the ideal behavior is to rank up, etc.

2 Likes
223

It would be super nice to understand why this happens…

Win a game, total stomp. 3 SR
Loss a game, really close. -25 SR
Win a game, really close, play terrible. +20 SR

Hard to have faith in the match making when the SR/MMR system is so confusing.

2 Likes
224

Tact, the questions weren’t really directed personally at you, mostly it was more a general, rhetorical question.

I’m just saying I don’t understand the fuss. I’m curious, but I don’t really have any more questions on the matter that they haven’t explained, other than how groups are handled, but that’s not really anyone else’s concern.

Detailed math might be nice, but it wouldn’t actually matter gameplay wise, though I do agree with your concerns, @kaawumba, those are not the common concerns.

225

That 3 SR thing is a bug. Been around for more than year.

“Really close / total stomp” is not taken into account. In theory, whether you were predicted to win will be taken into account. But be it that you are in gold, almost all of your matches will be predicted to be 50/50.

See How Competitive Skill Rating Works (Season 11) for what is taken into account.

226

The discussion in this thread has been more substantive and focused lately. I just want to say thanks to the contributors. Let’s keep being clear with each other, even when we disagree.

4 Likes
227

it’s just like the question of balance. I think the game is pretty well balanced, except for Mercy and Widow on the top of OWL and torb/sym on the bottom, but blizz will sort everything out and the community’s outrage baffles me.

228

Okay. Took a walk. Feel the same state of rational enthusiasm to debate on the internet. Is this a problem? Should I just be ultra-apathetic and resort to passive aggressive tendencies instead?

229

People like transparency. It’s why we demand it in every sector of business and government. Does it change how they play? No. But it opens up room for discussion and improvement. Right now, people are just speculating and yelling at each other based on nothing but made up pseudofacts about chess. I don’t see how that is any more constructive when the fact of the matter is: people don’t like the matchmaking process. Yes, it’s an opinion to not like something. But if a majority of people don’t like something, that’s different territory and brings us to the questions ‘Are they making this game for their users? Or are they making this game for themselves?’

2 Likes
230

Part of my point was that people don’t even believe, choose to misunderstand, or refuse to accept what they have given us. One side of this “debate” is purely speculation. The other side accepts what has been said at face value but asks questions about implementation details.

My point is that the side that thinks everything is speculation tends to also believe what information they have given is, at best, intentionally vague. They seem to want something concrete, but it’s really not at all clear how or why they think what we do have is not enough. I’ve seen mincing, parsing, and rejection of language on a whole 'nother level in these threads. If you won’t be happy until you have the math behind it, fine, but I somehow doubt the same people who confuse “estimated 50% chance of winning” with “forced 50%” win rate are going to be doing anything useful with that math.

I think it’s acceptably clear but could be clearer as an academic issue. MMR systems aren’t really that difficult to understand. It’s not exactly “speculation” to think that it works very similarly to all the other ones when it is described very similarly to all the other ones.

Unfortunately, I also think that the “rigged”, “handicap”, and “forced” reactions that come out of every single explanation of the system actually do more to convince them to keep it hidden than anything else.

I mean…if people take general concepts like attempting to have equally skill matched people on every team and turn them into multi-thousand post diatribes about how teams shouldn’t be balanced at all because it’s against some “competitive spirit”, why on Earth would they give us the detailed math behind each and every player and why they are in that particular match? HOW would they even do that and keep the game…a game…which is all it really is.

They’re not hiding it from people that are curious…they’re hiding it from people that twist words and intentions into something villainous.

Also, people tend to get things confused. MMR and matchmaking are related, but complaints about one isn’t necessarily a complaint about the other. For instance, it’s frustrating to get 5 Mercy mains on a team, but that has nothing at all to do with MMR. Related to that, people seem to want unrealistic perfection from the system. If you can figure out how to perfectly match 12 humans together every time you’ll be very, very rich in no time. There’s a discussion to be had there, sure, but people don’t even try to have a basic understanding before they start the discussion.

1 Like
231

There is a middle ground between being ultra-apathetic and raging about someone bringing up chess rating systems in a discussion of Overwatch’s rating system (which is not off topic, by the way). Stay in that middle ground. Take a break whenever you get angry. Feel free to come back and debate when you are calmer.

When I asked you a question, above, you just gave me a snippy response. That isn’t how debates work.

1 Like
232

So there is a blue post saying that SR and MMR are very close. They have also said that SR can go up or down while MMR doesn’t move much. These two points lead to an issue on what “very close” means.

In this game, it is very easy to go on a 10 game loss streak, even when you are playing the best Overwatch you ever have. Please don’t argue about the ability for this streak to happen, that would be a moot point as there are many ways it can happen (and I’ve experienced it along with many other people) - ie maybe it’s a top 500 GM smurf account trying to win but with a streak of leavers that just stopped them from being able to 1v6 for 10 games in a row. Focus instead on disputing the outcome and how it is reached. So, in one of those instances then SR will move down while MMR would barely move:

Start:
Game 1: SR: 2500 ESR: 2500
Game 2: SR: 2480 ESR: 2495
Game 3: SR: 2460 ESR: 2490
Game 4: SR: 2440 ESR: 2485
Game 5: SR: 2420 ESR: 2480
Game 6: SR: 2400 ESR: 2475
Game 7: SR: 2380 ESR: 2470
Game 8: SR: 2360 ESR: 2465
Game 9: SR: 2340 ESR: 2460
Game 10: SR: 2320 ESR: 2455

So after 10 games, this player now has a difference of over 130 between his SR and his ESR (ESR being the Effective SR if MMR was made into the “same” number for comparison purposes). So Blizzard must mean a greater difference is still “close”, and that “chase” could mean upwards of 40 games (based on how long they had people with abnormal SR games straight after placement before Season 5) for MMR and SR to realign.

But to realign then no further streak would need to happen. This game is very streaky, so every time a person has a streak like this, their MMR and SR would split further apart.

The only argument against this being possible is that MMR and SR are the same value always. Which contradicts the quote that SR always moves yet MMR doesn’t. And thus is speculation on what “very close” and “chases” mean.

As they now need to give that account a 50% chance to win, while having a mismatched SR/MMR, how MM works to reach that chance to win is still unsure. The recent post helped understand that SR definitely creates a limit on that predicted chance, but if they have 2 players that have that same mismatch then it still holds that they will always be put against each other, not on the same team. Otherwise that 50% chance to win would be skewed past 60%.

See, this is the reason I want to see MMR as well as SR (and I don’t care how MMR is made) is that it would within 1 day show whether the above is happening, that handicapping is happening.

If there is an obvious MMR vs SR issue leading to handicapping, then we could enter the “why” and “how to fix it” issues. And like Kaawumba said, if revealing MMR allows MMR to be easily manipulated by stat chasing then it was never good enough to begin with.

A few pre-counter arguments:

  1. But but I can climb still!
  • The above doesn’t mean you can’t climb. I have never said that.
  1. It contradicts blue post xyz
  • It doesn’t.
  1. SR matches MMR
  • There is a blue post describing how they move differently. If they move differently then they do not match. “The more certain the matchmaker is about your MMR, the less your MMR will change in either direction based on a win or loss.”
  1. MMR is described as a single number. Not composed of something and SR.
  • MMR is a number resulting from a formula. SR can be part of that formula. Or, think of it a different way - at the end of each match the adjustment to MMR is based off a formula that includes SR.
  1. If MMR depends on SR and SR depends on MMR, that would be circular.
  • Why does SR depend on MMR?
233

I’m not going to argue with you. It’s pointless. You make up a fictional “loss streak” and ESR numbers and expect someone to argue with you. No.

I don’t think you want to understand how it works. You want to think the system is designed in such a way that you can have an MMR that is permanently higher than your SR. You want to think that your 50% win rate at X SR is keeping you from climbing because you actually have a much higher ESR and of course you can’t climb with a 50% win rate.

In short. You want to think you are better than you are and the system is holding you back.

I posted something awhile back that explains this phenomenon of climbing with a 50% win rate. It’s not a silly question. I wondered the same thing. But once you think about it for a minute it’s easy to grasp.

https://us.forums.blizzard.com/en/overwatch/t/competitive-matchmakings-mmr-system-is-very-good/43184/28?u=ozoneooo-1681

There. It’s long, but it’s about the best explanation I have regarding your question, if it’s an honest question.

I hope I’m wrong about you.

1 Like
234

Because you can’t. Not because the numbers are made up, but because you can’t fault the logic. You’re the one who asked for maths.

Completely incorrect. I have even posted before how I prefer games in Plat. I’m a casual older player who is at a rank that works for me - I have fun games in this area because it seems to have the least amount of MMR/SR BS. In other words, I’m where my SR and MMR match. I’m too old to really improve anymore and play the game for fun.

I’m off to read this now, but can you clarify the question you think I’m asking/you are answering?

1 Like
235

The question is made quite clear in the post.

236

There is nothing in that link that has relevance to anything I posted. Are you sure you posted the right link?

What is my question? I didn’t ask a question, thus why I was confused.

Edit: did you mean this question?

That’s about a formula that determines SR based off your MMR, that would somehow affect your MMR, that would then affect your SR etc as a cyclical pattern.

1 Like
237

Sorry. I wasn’t trying to be coy. I was on my phone.

You only have one question in your post and it doesn’t appear to be more than a rhetorical question, I had to infer your question from context.

The question answered is in the second paragraph:

It’s the most common question I see. It’s perfectly legitimate too, but it’s not generally phrased in that manner. Often it’s couched in this “forced 50%” or “handicapping” language.

1 Like
239

You want to put money on that? You’d lose. Here’s an example of why.

Let’s take a Rein player who’s main goal is helping the team win. This Rein mainly shields, makes space and otherwise helps his dps and supports do their jobs safely. It works, his team has a 60% win rate. He gets 15 SR for a win and -25 for a loss.

This results in a net -10 SR per 10 games on average.
6 wins at 15 SR each = 90 SR
4 losses at -25 SR each = -100 SR

Out of frustration he tries something different. He sees how off tanks, who mainly focus on elims, get more SR per win. He becomes a dps Rein who only shields long enough to get to within hammer distance. Then it’s clobberin’ time.

His team drops from a 60% win rate to 50%, but a weird thing happens. Rein is getting 28 SR for a win and -23 SR for a loss. At 50% win rate he gets a net gain.
5 wins at 28 SR each = 140 SR
5 losses at -23 SR each = 115 SR
That’s a net SR gain of 25 per 10 games.

Now which would you rather do?
Before you get all crazy saying that I just made up these numbers, let me tell you that I have played Rein both ways. These numbers are accurate in my experience. The main reason I can’t regularly play Rein like this is people start tilting due to the losses.

The pain of a loss is immediate, and it takes many games to realize the SR increase from a lower win rate with higher SR gains per win.

So yes, if people knew how the SR calculation and matchmaking worked, they would change how they played in order to maximize their gains.

1 Like
241

Every piece of information is based on blue quotes. Silly forum doesn’t let me do links.

You seem to have failed to understand the post, which is that MMR does not match SR. As shown by your need to somehow make it about:

And I clearly said:

Which again means that your entire direction in this thread is off the rails, and has to be intentionally as I’ve seen you do it multiple times. Thus it’s trolling, and being marked as such.