Algorithmic Handicapping (MMR) is Wrong for Overwatch

It’s not necessarily bad players who I get consistently in my team to handicap me. Most recent experiences are people that go AFK at crucial moments (last fight in overtime) or throwers (they get angry, pick Lucio and speed boost away from the team the rest of the match).

1 Like

Daym!
That’s hell a lotta words you wrote over there, partner. Too bad Im aint readin’ all that!
Just passing by to quote this sick wall of text and look as tough as you!

3 Likes

Bro, don’t quote it. That took a full minute to scroll through.

3 Likes

I have tried to answer this very question in the original post, which I think you have read:

So let me put it differently:

I think we can all agree the ranking system is the most “productive” when it puts players in the correct order of skill, and lets them rank up/rank down in accordance with their development or degeneration as a player. I’m saying that algorithmically balanced (handicapped) matchmaking is bad for that function because it ignores rank while also making it a certainty that not all of the best players or all of the worst players will have a chance to win or lose the match together (i.e., gain or lose rank). The gain and loss of rank is guaranteed to be distributed evenly between players despite their innate skill and also because of their MMR.

I believe that in ranked competition, matchmaking should be 100% rank-based, random, and solo queue only. To form a match, the Matchmaker takes 12 players from the queue with the same or nearly the same SR score. MMR and PBSR do not exist in this system. Teams are assigned randomly, SR gains/losses from the outcome of the match are equal between teams, flat between players, and nearly the same amount from match to match. I think we can all agree that is a fair system. The system we have is an abomination, a Frankenstein’s monster.

I appreciate you for keeping this topic in your head for all these years, and I’m sorry for the burden we share.

2 Likes

I’m actually not sure we can agree on this, but maybe it’s just the way you phrased it. I would agree that a ranking system should produce both accurate ranks and movement in those ranks, sure. I’m not sure if we would agree on what level of accuracy is reasonable, though.

So, it’s true that sometimes the best player in a match will lose, just like it’s true that the best quarterback will lose. I’m not sure what alternate system would prevent this, though. For most of us, this isn’t a big deal, it’s just how life works sometimes?

This is placed in a way that makes me think it’s meant as a different phrasing of the sentence before it, but I honestly can’t make any sense of the “because of their MMR” part. And again, for most of us, this is just how life works sometimes, isn’t it? Even the bench players get a Super Bowl ring, don’t they?

Solo queue one we actually fully agree upon. Don’t blame the MM for that though, there’d be hell to pay with most players if they prevented you from playing with friends. This isn’t that serious of a game, either, that I really worry about it. Technically though, you’re right and I think most people would agree. Groups mess with it and always have. This is neither revelation nor in dispute, it’s merely a priority disagreement. (And also a hint that maybe you shouldn’t be taking “ranked competition in Overwatch” quite as seriously as you do.)

Re: Ranked Based, I think one issue we have talking to each other is that while we both agree that the system can do a pretty good job of determining your skills relative to other players, you don’t make the second mental step and realize that is the same as ranking, i.e. ranking is the result of determining your skills relative to other players. So it becomes hard to talk to you, because you say that the MM is ranking you accurately and putting you in matches based on rank, but that you would like the matches to be rank based. I can even accept your idea that SR and MMR are not the same and still be confused, because you seem to accept that MMR can, in fact, rank you. I’m really not sure how you get to this conclusion. Maybe you can find where I’m wrong and explain. Do you simply want a stronger guarantee that your SR and MMR don’t normally wildly diverge?

Also, I take a bit of issue that matchmaking CAN be 100% rank based and random. Those are mutually exclusive. I see what you’re getting at, but setting a range of SR and then assigning teams randomly will create expected wins odds that vary from 50% unless you happen to have all the same SR. So you have to decide what range you find acceptable, but you have your initial problem that the best player can lose. In fact, this should make it worse, not better, because now you’re putting your best player in a disadvantaged position. You’re doing exactly what you say is currently happening, but this time it is actually by design rather than by chance and precision.

As for the rest, while I can appreciate that is one way to do it, it seems like your attachment to having flat SR gains for all players is, well, a bit overzealous. I can appreciate that what you want would make a fair system, but I think you’re missing out on some benefits that PBSR and MMR provide, without any real clear gains in return other than a simpler system that most people actually like less in practice. Especially when you consider how “serious” this ranking is meant to be. Sounds like you really just want it to be super super serious, but that’s just, like, your opinion man. :slight_smile: And you’re entitled to it, sure, but the impression you give in this post makes people think very different things, you know. People think you mean that the system is preventing them from ranking up, where from what I can tell, you just don’t think the system is as good as it can be.

It also does nothing to solve your issue that sometimes the best player in the match will lose. Which, again, I don’t really see as a problem but you do, so I’d think that you’d either have a solution for that or just accept it as an inevitability of team play like the rest of us do.

I also don’t see how this would be more productive than they way they currently do it. So, I agree that what you want would basically work, I’m failing to see how it’s any better than what we currently have (or had, remember I haven’t looked at this for 4 years.)

I’m left wondering what your actual problem with the current system is. What you say is the problem would indicate that you want stronger PBSR (so that the strongest player always “wins”) and to be matched purely on MMR (so that the games are 100% rank based), but then you say that you want to get rid of both completely?

3 Likes

Oh, and this portion. Why do you think this is true? If we’re talking hypothetically here, why wouldn’t a system that works like this (my version) or this (taleswappers version) work better to produce an accurate ranking?

Not that OW necessarily functions like either of these, but, hypothetically speaking, what makes your “6 of the strongest players are on one team and all 6 of the weakest players are on the other” a better way of determining “the correct order of skill, and lets them rank up/rank down”?

2 Likes

Let’s remember that matches have 12 players. The outcome of the match reflects on all of them, to varying degrees and in relation to things like what role they are playing.

But that is a false equivalency where you are talking about a sport with established teams who compete in a league tournament. That’s completely different from Overwatch’s competitive design of players being ranked individual but being allowed to compete with all combinations and composition of group types…which is whack in itself.

You are underestimating the importance of this detail, and here’s why:

Team and group play is a huge part of Overwatch’s appeal. How they are handled in ranked competition is massively important, insofar as the game is important. And I think that everything people do is important, especially when it’s tens of millions of people collectively spending billions of hours of their time and energy. Activision/Blizzard and Microsoft clearly think it’s important, or they wouldn’t have gone about this enterprise. Consumers should pay attention to what they are doing, because much of it is nefarious.

You are talking about two different processes taking place in different systems. MMR and SR and different things. Stop trying to conflate them as the same thing.

Yes we have to talk more about this, because clearly it’s a point of disagreement. I do not believe but I know that SR and MMR “diverge” because I have seen in happen in my own Overwatch career. In the time that I played, my SR ranged from Platinum all the way to low Bronze and back several times. I observed that my career trajectory was random, despite sure knowledge that I had improved at the game over time.

Anyway, that is only the first-hand experience that made me suspicious. It is the words of Scott Mercer, Jeff Kaplan, and other Blizzard representatives which confirm my assertions.

SR and MMR are completely related yet completely different things. Comparing them is not apples to apples. It’s like comparing an apple to the tree it grew from.

MMR does not rank anyone. Rank is not its function, that is the function of SR. But the function of MMR is relative to the function of SR. MMR’s frame of reference is constantly changing because a player’s competition is always changing.

SR and MMR cannot diverge because they are not at all the same thing. MMR is generated in the short term, dynamic and constantly changing. SR is generated in the long term and is relatively static. PBSR relates to both systems, and we can get into that whenever you want.

They are not mutually exclusive.

The range would be tiny, nothing like the 1000+ SR window that exists today. That is something to keep in mind. Queue times would practically instantaneous as well.

Vary from 50% as predicted by whom, or by what? And yes we are talking about matches where everyone is practically the same rank/SR. This would be easy to arrange with solo queue participation rules. Groups are the main reason why matchmaking is so convoluted.

The acceptable range, when you think about it, should be very small. If we suppose that the new system we are describing would numerically resemble the existing SR system, then you might say the acceptable range is around 1-5 points of SR.

That is something that can happen naturally, not a “problem” that I have stated.

That’s wrong. The best player is typically disadvantaged by the current system, however.

We are talking about randomized teams, so how can you say this?

You have no idea, you’re just assuming that. A simpler system would be much better and is 100% necessary for ranked competition.

You either out of touch with the experience of Overwatch players, or you are being dishonest about it. Anyone who’s played competitive online games should be familiar with how invested players get in the outcome of their matches, and how seriously they take the practice of learning these games strategies. That goes double in a ranked environment like Competitive Play.

When the outcome of matches is effectively randomized with algorithmic handicapping, that confuses people about everything from battlefield tactics to character meta and their own development as players. Algorithmic handicapping is toxic to the gaming public, especially when it is done covertly in ranked competition.

The system does prevent players (mostly experienced and skilled players) from ranking up. By the same token, it prevents other players (mostly inexperienced and unskilled players) from ranking down. The MMR system is very good at what it does. I do not want the system improved, I want it abolished.

Where are you getting this from? I did not say this.

No, PBSR should not exist. Did you not just agree that SR/rank gain/loss should be flat per match? That is what we’re talking about.

No, and again MMR is not rank. The function of rank belongs to SR.

Correct, and again I did not say that “I want stronger PBSR” or anything like that. PBSR and MMR should not exist. Matchmaking should be solo queue and based purely on rank (SR).

1 Like

You -could- have small PBSR component on a per-match basis that has nothing to do with anchor stats or rank averages broken by alts and rigging and 6 years of scott’s no-reset laddering ™. The small PBSR, properly done in context with ladder redux, would help move people to where they belong faster and rigging-free.

As for MMR yes ofc it needs to go, all SBMM, EOMM, and DDA needs to go. Imagine having hitboxes change on you mid-game based on your real-time performance - that’s where they’re planning to take the industry → consumers are the product and their rights be damned. Typical for the ATVI and MSFT playbooks.

With your system of just grabbing 12 people of the same SR, the game is just equally randomized, is it not?

1 Like

Can you remind me, when was Chess ELO reset the last time?

2 Likes

Chess ELO doesn’t invade your data (mmr). It doesn’t care about how you play only if you win/lose against others who win/lose. And it doesn’t rig the lobby it finds players of similar SR (in this case chess elo) and ships. It’s also a solo queue experience.

Not comparable based on those facts.

i dont see any math…

1 Like

after win streaks it puts you into matches that are insanely hard to carry and I jumped the boat as soon as I noticed like having multiple afk… drunks or launching ult’s off the map on purpose just “because” “CHILLLL” like legit you will run into the worst side of the community after you win games and that’s what makes me at least leave… I’m sure it does for others as well especially since blizzard is one of the worst companies right now about policies and keeping their community, leaderboards and stuff clean and nice… including how the community acts.

I want to keep playing hard games or run into harder players on the enemy team I don’t want a half afk team refusing to work together because the MMR decided since I won too many now I get them for trying too hard. Back on the first forums they actually had talked more openly about the match maker and how it work but then retracted their statement and deleted the threads from back then.

math isn’t limited to numbers. it’s an area of knowledge, reasoning, logic, discourse, structures and relations, etc.

maybe you’re thinking about arithmetic?

reasoning logic and discourse are parameters of critical thinking, which is the basis of AI. maybe you have the concept of math wrong.

1 Like

You are so sad and pathetic.

1 Like

maybe you failed to grasp what i said:

See? I beat you to it. Not surprising since you can’t math.

So, I want to clear something up first. If this whole thing was about groups, and you wanted a solo-queue, you could have saved yourself 4 years of heartache and a bunch of neo-namecalling by saying that outright. Most people would be in agreement with you. Those who weren’t would be making a totally different claim, not disputing terminology or tactics. I’m not saying you would have won, but you would have a very different group of people with you or against you and some change may have actually come about. Same with PBSR, though this one I think has a bit more nuance and no really correct answer, just various compromises.

I’m not going to bother arguing with you on your points above. I don’t play the game anymore and I won’t. I just want to make sure you understand two concepts that I really don’t think you understand, but maybe you do. I’m going to tell a couple of stories that have nothing to do with Overwatch. They’re just parables of concepts. If you don’t see any connection, don’t worry about it. Just think about the stories.

Story 1:

Two people are hanging out and wanting to play a game. All they have is a coin. They decide that they are going to play a game where they each flip a coin and the first one to 10 heads wins. They play 100 games, and of course the outcome is very close to 50/50 because it’s random chance. They are equally skilled at this game, because the game takes no actual skill.

Later, they buy a checkers set. Neither have played checkers, but they are both just as smart as each other, so they start playing checkers. Their skill started the same, at zero. They both learn at equal rates! And after 100 games, the outcome is very close to 50/50. They are equally skilled at this game, though the game does take some skill.

Later, they go to bet on the Super Bowl. The first bet they make is on the coin flip. They go to the bookie, they bet against each other, giving $50 each to the bookie. Only one of them can win, person A gets their $50 back plus $48 for winning, person B gets nothing, and the bookie keeps $2.

The bookie asks if they want to bet on the winner of the game. They both know A LOT about the teams in the Super Bowl. They’re experts. Maybe they have some inside knowledge. They know that Team 1 is way, way better than Team 2, but that the popular notion is that they’re about equally matched, for whatever reasons. The bookie, George, gives the same payouts as the coin flip. George can do this because he’s getting bets from other people that are about evenly matched on either side. Everyone looking at the prior results can’t make up their mind about who will win. But since our protagonists somehow know that Team 1 is a pro team and Team 2’s players got way too drunk last night, they empty their savings and put it all on Team 1.

Team 1 wins, of course, and they get their million dollars back and nearly a million more in payouts (bookies’ gotta eat!). To celebrate, they go home and decide to play tennis. One of them is a tennis pro, the other has never picked up a racket in their life. The pro says, “Hey, you wanna bet your half on the game?” The non-pro replies, “Nah, I’m not clairvoyant, but I can do math.”

Story 2:
Sally has a mess. Somehow, some way, Sally ended up with 5000 strips of paper, each of different lengths, and she has to sort them. Sally thinks about it for a second. How is she going to sort these papers? She realizes that she can just start putting the strips up next to each other, comparing them side by side. So she gets to work. She picks up two strips, puts them next to each other, and puts the longer one on the right and the shorter one on the left. Then she picks up another one, compares it to the leftmost one, and finds it’s longer. Then she compares it to the rightmost one and finds it shorter, so it gets placed in the middle.

On and on it goes, picking up a slip, comparing it one by one, putting it in its proper place. Finally, after a lot of hard work, it’s done. She has all 5000 pieces, painstakingly arranged by size, left to right. Sally is happy and satisfied, until disaster strikes!

The door opens and a big gust of wind blows the papers everywhere! George (yes, the bookie, but off duty) walks through the door and sees Sally crying. George tries to console Sally and learns of his part in her troubles and thinks he knows a way to help. George has a ruler, marked in millimeters and centimeters.

Sally is confused for a second, but George explains that Sally can just measure each piece of paper, write the number on it, and then do the same thing that she did before, except instead of comparing them directly one by one, she can sort them by the measured number. So Sally takes the ruler, grabs her pencil, and gets to work. Before too long she has them sorted again, but while it did save her some time, it didn’t save a lot of time. She thanks George, who leaves. Through the door. The same door. With the same result.

Fricking George.

However, since Sally has already measured and marked the strips of paper, she’s not so distraught. She just picks up the papers, looks at the number, and gets them all sorted by size without measurement or comparison. The hard work has already been done.

If you like my stories, hit like and smash that subscribe button, and maybe I’ll write more!

2 Likes

No, the whole thing is not about groups. Groups are only one aspect of it. Overwatch’s algorithmic handicapping applies to solo queue players as well as groups.

Why do you say “of course the outcome is very close to 50/50?” This is not a given, in theory nor in practice. In a set of 100 flips, by ‘random chance’ the split could be as much as 100/0. And probably the outcome would be far from 50/50 in many other cases.

I like your stories, but they don’t relate to Overwatch or algorithmic handicapping in any way that I can tell.

1 Like

It’s most certainly a given in theory. In practice, you’re trying really, really, really hard to argue and not putting any effort into understanding.

Make it 1000.
Make it 1000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000.

Make it whatever it takes for you to understand my point.

And if you really think that an infinite number of hypothetical coin flips can reasonably result in a purely one-sided outcome, that tells me all I need to know about your abilities to comprehend.

In actuality, I just think you’ve never tried to understand. You have feelings and you try to justify them.

I suspected as much.

1 Like