As I’ve said before, the way most people describe how they WANT it to work is also how it ACTUALLY works.
First, lets throw out the “hidden” aspect of it. How it works is completely irrelevant to whether you can see it. In other words, they could show us the MMR and do nothing else and it would function exactly the same in terms of ranking. I also ignore this aspect of it because there’s a lot of people on both sides of the debate that agree that having it hidden ranges from unnecessary to counter-productive. It’s not really a point of strong debate and it frequently pops up as a side show in the discussion regarding MMR and SR. See post #260 in this thread.
On to your point; I went over this before around this post (#276) but I’ll gladly expand here: Algorithmic Handicapping (MMR) is Wrong for Overwatch - #304 by OzoneOOO-1681
In a sense, you’re right that matching on SR would be efficient. That’s not the mistake anyone is making here. The mistake is thinking that MMR doesn’t work like you see SR work.
If you haven’t already, read this post (#28): Competitive matchmaking's MMR system is VERY good - #28 by OzoneOOO-1681
I think the trouble some people get into when thinking about this is that they see their SR float up and down and incorrectly think it always goes up a certain amount on a win and a certain amount on a loss. While it is true in practice, that’s not actually the underlying mechanics of it.
It LOOKS that way because you’re almost always matched against opponents that are about as good as you.
But SR isn’t an approximation of your winrate. It’s a description of your MMR (how accurate of a description is irrelevant to my point). That’s why I keep telling Cuthbert that it’s simply factually incorrect to say that SR depends on wins and losses. If your SR was 2000 but you somehow got placed into a series of games at 4500 SR, it wouldn’t really matter how many you lost, you would hardly drop any SR at all, if any. That’s how ALL Elo systems work. Your losses are irrelevant to the system, or at least secondary to WHO the game is played against.
I UNDERSTAND WHY PEOPLE THINK THIS. It’s by far the most common way of ranking teams in sports. If the NY Yankees have a 50% win rate, they’re a solidly average team. How can 2 people in OW have a 50% win rate but have two different rankings? I get it, I really do.
Sports teams play the same number of games and always play games against each other. It’s the same with OWL. Every team plays every other team, is what I mean. The winrate here is an absolute measurement. Someone could win all the games. Anything less is relative to the max number of wins, which is a set, known number.
In OW, or really any multiplayer game, you don’t have the benefit of a set number of games nor the clarity of playing every other player. We’re not going to play a million games each season. There isn’t a max number of wins to compare yourself against. You can’t use an absolute system. You MUST use a relative system.
So you need a DIFFERENT way of ranking people other than the usual win-rate method.
The different way is to rank people based on who they are winning and losing against. Your CURRENT rank is higher than the player you just beat but lower than the player you just lost to. If you have exactly a 50% win rate then your rank will be accurate. If you have less or more than that then your rank is unknown. The further from 50% you are, the less accurate your rank is.
I get that this is wildly different from how people normally think of win rates. Normally, win rate IS the rank in some sense. But that can’t work in a game like this. There are too many variables.
If you matched on SR it would be the same as matching on MMR. Exactly the same. Your SR would be (and is) where your maintain a 50% win rate. The only difference between that system and the current on is the ability to provide punishment, decay, and whatever “feel good” that they may or may not still add (which I’m against, for the record).
Some people (and I believe this includes Cuthbert) think that teams should be randomly assigned regardless of skill and your win-rate (SR, in Cuthbert’s understanding) then becomes your rank.
This would rank people, that’s for sure. I’ll let you be the judge of a system where having a loss rate or a win rate higher than 60% was normal and where you could play 10 games for a 90% win rate and be ranked as high as Jjonak.
I now feel the need to say it every time, so I remind people here that there are several things that could be improved in the system, but you ARE matched based on your rank and always have been, having MMR and balanced matches isn’t what’s broken.
Here's an example I removed from the main text
If you maintain a 50% win rate at 2000 SR, then your SR should be 2000, right? You shouldn’t expect more or less than that. If you maintain a 50% win rate at 4000 SR, then that’s where you belong. What doesn’t happen is that you maintain a winrate of 50% at 4000 SR but have 2000SR. Just watch a bunch of streams or your own games. The average SR of both teams is nearly always +/- 200 SR of yours and generally very close to each other. In fact, before LFG it was generally within 50SR both ways when they had a higher selection of solo individuals to match together.