MMR based team balancing is not for Overwatch

Competitive Discussion
1

I’ve heard a lot of people say that teams are balanced based on MMR, so the system tries to make sure the MMR on both teams is as similar as possible. This is wrong.

First of all, we already have a skill measurement for ranked: SR. As everyone says, “If ur gud enough to climb to a rank, you will.” So why do we need to balance the teams with MMR? This ends up attempting to solve a problem that didn’t really need solving, but also ends up creating 2 new problems: Hardsticking (IDK the present verb), and increased boosting.

Let’s just say ur hardstuck, you are 1 rank below where you should be. This rank is infested with smurfs about 2-3 ranks below where they should be. The MMR based team balancing system will go: there are 10 ppl about where they should be, there’s you, and there’s the smurf. How do we balance that? By placing you against the smurf. And then you get rolled. So yeah, it is possible to get basically only smurfs on the enemy team.
Conversely, throwers, leavers, etc. They will most likely have a worse MMR, so you’ll get placed with them since your slightly better MMR counteracts their worse MMR. So yes, it’s quite possible to get smurfs almost always on the enemy team, and throwers/leavers only on your team.

Now, onto boosting. if you got boosted up 1 rank, then the same thing happens, but opposite. You get all the smurfs in that rank in order to counteract your bad MMR, and you get carried. On the other hand, if that person is hardstuck, meaning they’re better than the ppl in the rank, but not good enough, you’ll drag them down, further making them hardstuck.

If you got rid of this, the chance of getting smurfs on the enemy team would go from like 90%, to 54.5454…% (since there are more enemies) this may be larger than half of the time, but its WAY better than 90% of the time. This goes the same for thrower/leavers on your team.

“But if this system is removed, wouldn’t ppl just get lucky and get really good games?” I hear you cry. Well, eventually they’ll start getting to where they get so out-skilled their rank will just naturally decline. I mean, the ranked system is perfect right? Then again, that’s only the case when there are no outside influences (Smurfs, boosters, leavers, etc.). Also, if the system was kept in place and they got a lucky streak and began to climb, as I said before, the system would place them with better players, who would carry them.

13 Likes
2

Don’t believe them.

This is how things actually work: How Competitive Skill Rating Works (Season 14). See especially the “Summary” ->"Summarize matchmaking, rating, and progression for me, and “Popular Myths” → “Matchmaking takes people at a given SR, and then matches high skill players with low skill players to make a balanced match.”

3 Likes
3

I really did appreciate the effort you put into your post, but the problem is it boils down to “I don’t really know how it works, but wins are good, losses are bad”. Well. Yeah. We know that. Without knowing exactly how MM pairs over / under performers, nobody can really say definitively whether the system inherently works. If we are just going to base data on experience (which is the only thing we can really do, since Matchmaker and SR calculations aren’t remotely transparent), then there is plenty of justification for Matchmaker handicapping games if you are over-performing. There is plenty of “experiences” like that.

I know if I “one-trick” my mains, I over-perform in my given rank. If I flex, I under-perform. That’s a rise and fall of my rank. If I don’t flex, we can end up with 4 DPS mains, and 2 Mercy mains, and everyone under-performs. So did MM do it’s job? Did it create a balanced composition for you team? Blizzard would say yes, because what was on paper should have been 50-50. But in reality, it didn’t because the enemy could be much more cooperative.

That’s the problem with how SR, MMR, and it’s relationship with Matchmaking works - It seemingly doesn’t take into account willingness to cooperate, or ability to mesh as a team. It looks at raw data (as any developer-minded person would think is feasible), and determines an ideal win-chance by that. The problem is, it can’t (or doesn’t) determine things like - who uses their mic to effectively communicate? Who can effectively fill a tank role with the Genji one-trick? Who has decent survivability with a niche one-trick? Who consistently picks hero’s against their direct counters? Who can effectively flex? Can this player perform without a specific meta (IE: Torb / Orisa. Reaper / Brig. Ana / Genjii)? And then pair that against a given win-rate or over-performing percentage. Instead, based on experience, it seems to look at player stats, and pair good players with bad ones within a given SR to create a balanced team. And that’s where so many people have a problem with MM. If you perform badly, you should be paired with players who perform badly. You obviously don’t belong in the same SR as someone who is consistently over-performing. If you are over-performing, you shouldn’t be paired with under-performing accounts to carry them. One of you doesn’t belong in that SR and trying to force a balance isn’t fair to over-performers.

If you’re over-performing, you should be challenged with over-performing enemy teams - not paired with potatoes to balance out the given SR. Just because you’re over / under performing within a given role, doesn’t mean you’ll be able to effectively perform at the same level with say… a bad attack Sym. Or a DPS Lucio. Or two off-tanks. Or no Support mains. The fact that these all exist within a given SR instead of a clear separation of those skill-sets is a glaring issue with how MM works. One player shouldn’t be able to decide that he’s going to try attack Sym for the first time over the course of 5 games, and know the system will correct his behavior with carries. That’s the experience people are having.

The problem with trying to predict or understand what MM is doing without any transparency besides vague statements that leave room for wild speculation from the community, is that you end up with “Win = Good” and “Lose = Bad” with some pepper flakes of “perform well with the hero you choose”.

Well yeah. We know that winning is good, and performing well is good. But then if you’re performing well, you should be winning, and thus climbing, right? No it doesn’t. And it’s that grey area of “Why am I not climbing if I’m performing well?” that simply leaves the community unfulfilled. And that’s a really bad place to leave Competitive in a popular game.

7 Likes
4

Ranking systems are inherently reductive in nature. Something as ineffable as willingness to work with a team can’t really be quantified. Win rate and stats can be, so that is what the ranking system works with.

And I imagine your “experience” is that you are the good player, and your teammates are the potatoes, yes? This is called the Dunning-Krueger effect.

The reverse (where you think you are the bad player, and you are getting carried), is called imposter syndrome, but that is less common.

People are extremely poor at evaluating their own ability. Well more than half the population thinks that they are above average. A significant number of those people are wrong and will reject quantifiable evidence (such as SR) to the contrary.

I don’t have an exact specification for the system, but I can conclusively reject handicapping as you and the OP describe. How Competitive Skill Rating Works (Season 14) → “Popular Myths” → “Matchmaking takes people at a given SR, and then matches high skill players with low skill players to make a balanced match”.

1 Like
5

The ranking system is probably fine, or about as good as it gets.

The real downfall of this game, is that in many ways it is really easy to pick up, and become average at, but extremely hard to become great at. What this means though is that you have very caste based system that can’t be evaluated by any numbers you put into the system.

You get hard-core, serious gamers who work and play hard to attain high rank in order to enter the e-sports spectrum (achiever), to the casual competitive gamers that try hard, but will simply never have the mechanical skill to make due in high ranks (explorer), the very casual gamer who simply is trying out the more “structured” competitive but has no real aspirations to succeed but like the social aspect (socializer), and finally the griefer, or killer archetype that are simply there to cause different types of mayem because they find it’s fun to work against the expected behaviors. These fall into the the four Bartle Types.

Within each rank there are higher percentages of these four types. For instance bronze and silver are predominantly the the latter 2, and Master and Grandmaster are predominantly the former 2.

This makes it really hard in the middle and lower ranks for players to feel like they “belong” in those groups, because they are consistently paired up with players with very different play types.

The achievers get frustrated that the killers keep choosing a character that doesn’t reflect the best team build, the socializer hates that the achiever keeps trying to tell them what to do on voice chat, and so on.

These are things that a ranking system won’t change. It is a result of Blizzard’s original intent to make Overwatch as open to everyone as possible, and their current effort to make is a fun to watch as possible for e-sport. It’s an impossible balance to maintain unfortunately, and unless you’re in one extreme or the other, it will be difficult to find continued “fun” by the majority of Overwatch players.

2 Likes
6

There is much more that goes into “high skill” and “Low Skill” and if it’s not taking that into account, that’s where the gripes come from. A genji one trick could be flagged as “high skill” where the majority of the community would unanimously agree it’s more of a hindrance to the team, then an added benefit unless the ENTIRE team plays around a one-trick Genji, especially when he’s countered - which is inherently counter-competitive, and allows for a single player to hijack a game.

Does the MMR flag him as high skill? Or low skill? How are one-tricks measured? Likewise, throwers / smurfs? How are they measured? If the underlying algorithm sways towards a significant amount of “ineffable” data, how can you claim it’s an accurate system. It’s not difficult to hop into a silver / gold / plat game and see wildly different skillsets, game-sense, awareness, aim, roles, and abilities, yet it exists all within the realm of the same SR. And it’s blatantly obvious.

Thank you, I’ve almost completed my Master’s in Psychology actually.

Having teammates that don’t communicate - detriment. Having DPS who get directly countered and refuse to swap - detriment. Having first time role-players - detriment. One-tricking - mostly detriment. Smurfing - detriment. Throwing - Detriment. These sorts of behaviors are blatantly obvious within a small set of games. Why do they all exist within the same SR? Even if you go down the “well maybe people were having bad games / days” route, you can’t reasonably justify the wild skill gap in any given SR sub-Diamond. And MMR certainly doesn’t do an adaquate job in measuring where one-tricks, smurfs, or throwers stand. It’s pretty obvious when someone doesn’t care about winning, and refuses to swap over lengths of games. Just because someone has a 51% win-rate on Sym, doesn’t mean that she’s a “high skill” player. If the MMR and SR system is trying to create balance, then those simple data driven statistics should be reflected, and they’re mostly not, and left up to speculation.

With these wild skill gaps within an SR, there are two ways the system is handling this - It’s either accounting for it and balancing the teams, or it’s not. And if it’s not, it needs to be eliminated altogether, because wins / losses within an SR range should be all that matters. MicBeMac is 100% right. There are only a few types of players, and in the middle ranks, it’s a crap shoot. What I don’t agree with, is the idea that the system is fine, and can accommodate the different kinds of players.

I don’t think I’m better than I am. I’ve been consistently plat the last couple seasons. This season has seen a sharp decline in the community, starting with my placements (two leavers in a row on my team for my first two, a blatant and admitted thrower on my last two) followed by a slew of team detriments that’ve left me in low gold. It’s incredibly frustrating when you are willing to flex, and the MM system pairs you with a Sym or Genji one trick, who gets directly countered and refuses to swap. Same with getting 5 DPS mains, and you. It happens all the time, and if the MMR system can’t detect that kind of stuff, then you’re absolutely right, and the post makes total sense with the data we are given - Win / Loss is all that really matters.

1 Like
7

Then why do you keep saying how horrible everyone else at your rank is?

If you think you are average for your rank, but everyone else has widely varying skill (both higher and lower) you should make that clear.

If someone has a ~50% win rate, they are near the correct rank. Skill in overwatch means one thing, and one thing alone: the ability to win ( > 50% against lower skilled, ~ 50% against equally skilled, and < 50% against higher skilled).

Different people are bad for different reasons. That doesn’t mean that they belong at different ranks. If the DPS who refuses to swap isn’t countered, he does just fine. If a person throws every other game (and hard carries the rest) he is at the correct rank, even though he decreases match quality for everybody.

If your team play is good, you are held back in some other way, perhaps mechanics.

1 Like
8

The OP is right. MMR is wrong for competitive play.

3 Likes
9

yup, seen this in action several times.

Had just won 4 games as a 4 stack. Got some not so great team mates for next few games. Nothing untoward in terms of skill. Maybe even us playing tired or tilted. We lost 3 games. 3rd had a Pharah that died several times to enemy Reaper and didn’t appear to know how her jumpjets worked. Hey ho…

Then the 4th game. We get a Widow that hits one head shot all game and doesn’t swap vs. a Winston dive and Rein hybrid team. Fair enough. Maybe they were having a bad day?

But, piece de la resistance. The instalock Moira that managed 1000 healing in 8 minutes.

Yep, matchmaker really wanted us to lose, so it dragged this potato out of somewhere to make darn sure that we lost.

I had similar experiences playing with a GM on his gold smurf. Some “special” team mates, such as a Zen not using any orbs (Or doing much of anything). After that game my GM friend said that he was going back to GM as trying to carry in gold was too stressful for him.

If you have some highly skilled players in a team, you’re sure as apples are apples going to get some potato team mates to balance it out thanks to MMR matchmaker.

Unless you 6 stack, of course…

1 Like
10

im curious if its an intended effect of making your own team? MMR wouldnt be an influence if you had a full 6 stack no?

11

Kaawumba, you speak the truth and it makes so much sense. People don’t want to listen because, like so many things, it is easier to put the blame somewhere else. Sure, MM in overwatch is not perfect, but it is a lot better than people make it out to be.

3 Likes
12

Nobody knows how MMR works except for Blizz and they keep a tight lid on it.

That said, though it is just my opinion, I believe OP is right.

However, no matter how MMR works, it is rigging matches in some way or another and should be abandoned for a simple SR system that randomly groups people of like SR. Then the ladder could settle itself organically.

4 Likes
13

Blizzard didn’t keep a tight lid on it. EVERYTHING they ever said about matchmaking, and they’ve said a LOT, is compiled in Kaawumba’s topic. Check the references section.

The only reason why it appears they don’t post here is because there’s so many posts of no real substance (basically amounting to childish tantrums) yelling about rigged matches, being intentionally and systematically placed bad teammates to keep them down, etc., that no reasonable Blizzard employee would or should waste their time reading and responding to them.

1 Like
14

Did anyone else use the pursuit app? You could literally see how bad the matchmaker could be. It would show everyone’s contribution and you could see the players who clearly didn’t belong in a 50/50 match. It’s probably why they banned it.

Defenders try and gaslight anyone who criticises saying they think they think they are better than their team mates. But I think once you play enough hours you can tell when the teams are balanced and you can tell when one side is stacked.

I don’t think it’s on purpose I just think the matchmaking can be really bad. Add in hidden mmr and you are just fueling conspiracy theories.

3 Likes
15

That’s exactly why they banned it and also why they added hidden profiles under the guise of protecting players.

The matchmaker creators is comprised of people who suffer severely from the dunning krueger effect, they are awful at creating matchmakers and actually believe they are far better at it than they really are. They blame the thousands, if not millions of people complaining about the poor matchmaking instead of clearly realizing the problem is themselves.

3 Likes
16

Not true, at least according to this Reddit post:

TLDR: It’s actually quite hard to figure out someone’s skill from win/loss percentages. There are too many confounding variables and most players do not play enough games to give an accurate rating.

2 Likes
17

It doesn’t work like this, or the matchmaker isn’t good at it for a significant portion of the matches. If the matchmaker consistently created matches where each team had a 50% chance of winning, then no one’s SR would change much. Everyone, regardless of skill, would have ~50% win rate.

The 50/50 balance is used by the matchmaker, but not 100% of the time. It has to allow for people to climb and fall based on their skill and number of matches played.

18

First off, I didn’t. I said there are wildly different skill levels within any given SR Sub-Diamond, and it’s blatantly obvious. Anyone who doesn’t see is, is simply not looking.

With a SINGLE hero. That’s the difference. It’s a huge gaping hole in the logic of a “team based” game utilizing this kind of MMR, allowing for this flexibility. If you already have 3 DPS and are flexing around that, you don’t need a 4th Sym one-trick, as it’s an inherent deficit in your team-comp. My question is whether the system accounts for this, and “balances” or if it doesn’t, in which case, one-tricking is invisible to MMR so long as you have average stats. Regardless, it’s hijacking a game with an off-meta hero that requires an entire team to flex around it - and that’s inherently non-competitive.

Except no. Because if someone is throwing every other game, they don’t belong in competitive. There isn’t chances for that. In any sport, if the Goalie has a consistent history of just standing to the side and letting the other team score, he would be benched immediately. He would be allowed to do it for 28 games, then get a slap on the wrist, and allowed to do it again for another 28 games just because he decided to hard carry a few games in between. That’s ridiculous.

This doesn’t matter. There is still 5 other people on his team that he’s holding back. You guys keep forgetting the context of the idea that there is still a rest of your team that’s being affected by selfish and non-team play. Sure he may have a 50% win-rate on HIS account. But he’s causing loses for 5 other players, 50% of the time. Regardless of “balance” or the argument that “Then enemy has the same problem”, is irrelevant, because it shouldn’t be allowed on EITHER side. Just passing it off as balanced and normal is a large part of why nothing gets done about toxicity, throwing, smurfing, or selfishness. It’s not normal, and shouldn’t be allowed.

2 Likes
19

The system doesn’t model one-tricking or flexing. It only compares a player to other players of the same rank/hero/map on a game by game basis (below diamond) for performance metrics. Winning is more important. In diamond and above, only winning matters.

It looks to me like your problem is less with the system, and more with other players. Blizzard only has blunt instruments to modify player behavior and tweaking the rating and banning systems is of limited effectiveness except against the most egregious behavior. If your enjoyment of the game requires a certain kind of teammate, you can’t rely on solo-queue. Either LFG, or friends lists, or discords, etc. is a necessity.

20

Why are the game designers so sure skill distribution is a Bell curve. I’ve seen arguments that the ‘power law’ curve is much better at representing skill distribution.

If the designers are using bad assumptions I.e skill distribution is a Bell curve they end up creating systems that just confirm their own bias , and garbage in, garbage out.

1 Like