MMR - The big talking point lately

So there's a lot that's incorrect about this, mostly with how MMR is assumed to be calculated and used. Since you separate your post into sections, I'll try to mirror that format, though there's some overlap where things are intertwined.

MMR promotes speed over quality.

Not incorrect, but I'll comment on this. The devs introduced the current MMR system as a replacement for ranks in September 2021.

Not too long after they implemented this, they got feedback that people didn't care for it and they decided to try and tune the system better for the community. I want to say these tests occurred within the first year of the system being out (I don't remember at this point).

But the tests ran for about a week or two, and were single-blind: meaning the devs changed something, didn't tell the players what they changed, and then asked for feedback every day to gauge how those changes felt for players, even though the community didn't have the slightest clue what they changed.

At the end of those tests, the devs released their data and had dialed in some numbers to tweak things. They played with the spread of MMR, how long it took to start expanding the search (lower quality games, but faster queue times, and vice versa) among other things. The end result was what players said "felt best of those tests at the time" and the devs were more or less happy with the queue times (so, they found a semblance of balance).

One of the tests, once they revealed the results, was going back to the old system of "ranks", and people hated it. It was one of the least popular tests they ran, and keep in mind people didn't realize they were in that system... They just went by how well they felt matches were balanced for that particular test.

So yes, the system promotes speed over quality to a degree. It initially tries to find a "good match", then steadily expands the search radius for less accurate matchmaking as time goes on. Ideally, it finds a match before things get too far out of balance, but things like back fills, lobby dodging, etc, can effectively turn off matchmaking entirely to fill out an already started lobby, so people aren't waiting (since the game doesn't even create a "lobby" until it thinks there's a valid match anyway).

How MMR is calculated.

So, starting (or "new player") values: I will say, though, that players have an MMR for each killer, and if you play a brand new killer it will start you at "roughly your killer MMR, but slightly lower" at first.

The idea being that you should have a general killer MMR, which covers the fundamentals, but you'll probably have to adjust to a new kit on a brand new killer. So you don't start over completely every new chapter release, for example.

The big issue is with the MMR gain and loss. The maximum gain or loss is +/-20 and roughly +5 for a kill, and -5 for an escape on the killer end. But this assumes that every player in that match is exactly the same MMR... something that basically never actually happens in reality

However, in any functional MMR system there's a weighting on MMR gains and losses to keep the system in check. So, if you have a 12k hour blight player at, say 2100 MMR, and they somehow get matched against a brand new team of MMR 650 survivors, the system should never give full +20 to the blight for a 4k and -20 to the survivors. We already know the outcome just by the numbers, the blight is going to win by a landslide. And the MMR system should also know that this is the expected outcome, and not adjust MMR hardly at all for this mismatch.

So, generally speaking, the system will calculate the average MMR for each team (in DbD the killer "team" is pretty straightforward) and match based on that. And if there happens to be an exact match, then you get the +/-20 values you tossed out.

But the norm is going to be that matches have one team with higher MMR than the other. So if the underdog team loses, that's the expected outcome, and if they win it's the unexpected upset. The underdog team should lose less MMR and the higher MMR team earn less MMR when that match plays out as expected. And it should be based on how mismatched the teams are: so similar teams are close to +/-20, but our extreme example of blight vs newbies should be almost no change in MMR at all on either team.

That's how nearly every MMR system in existence works, and I surely hope that's how this one does. Because if its simply a flat +/-20 regardless of the matchup, then just the fact that the game is set up for a 60% kill rate means that (if they play long enough) every survivor has minimum MMR and every killer has maximum MMR.

MMR brackets.

These don't exist, at least not in a meaningful way. The system doesn't care where you are in the MMR system, with only one exception (and I'll get to that).

The way the system tries to match you, is it takes your MMR and looks for a suitable game at that number. If it can't find 4 other people at your MMR within a certain time (like a minute or two) it expands to looking within +/-100 (or so) of your MMR and keeps going. It expands the search range until you find a game, so shorter queues should also mean, roughly, that you have a generally accurate game also.

But that process doesn't care if you're "on the edge of a bracket", and works exactly the same way if you're 500, 1200, or 2100 MMR.

The only exception for this is the soft cap, which I'm going to keep saying is 2100, but that might not be exact (and the real value is kind of irrelevant). Since you could keep winning, theoretically forever, at huge extremes of skill (pros) players would take an extremely long time to find a match. The soft cap is just a limit where, your actual MMR can be above that number (say 2400), but the system will look for matches as though you're 2100 instead.

But: since there are diminishing returns on MMR gains (or should be), then at this extreme level that player is gaining almost nothing for stomping players will below their MMR (see above) and similarly, if and when they lose they'll lose a ton as well (because the underdog won).

So it becomes difficult to maintain that high MMR unless you are genuinely one of the best in the game. It's also possible, and this is much more common, that people can win an awful lot, and struggle in matches, and still aren't "Max MMR".

Final words

Not responding to yours, per se, just sorting out my final thoughts.

Maybe a community manager or Dev can confirm some of this, because I think a lot of what I cover is public knowledge (although the exact numbers I think are intentionally hidden).

I'll also say that I don't think MMR is perfect by any means.

Back fill and lobby dodging basically turn off all matchmaking and prioritize finding anyone with a pulse, and match quality goes out the window.

I'm almost entirely certain that, even if the system had a perfect, infinite supply of every MMR player that the game would match a 1000 MMR killer with two 1000 MMR survivors, a 1200 survivor, and an 800 survivor. I think the matchmaker intentionally makes a weak link on the team (this is a bit tin foil hat, admittedly).

Most importantly, most players are not "Max MMR", or even at the soft cap. Just because you win 40 games in a row, doesn't (or shouldn't) boost your MMR by 800. You'll eventually hit diminishing returns farming the "easy" games, and lose/struggle against opponents who are better than you. Playing a lot will mean you hit an MMR that is your skill level, and that is most likely not at or above cap. That's also ok.

Good point, i did forget about the anti Smurf thing.

Back fill and dodges could definitely be made better, though I feel like the "best" solution is to either rework the lobby entirely or just get rid of it altogether. I'm not sure if the devs agree with either of those approaches (or have the time/resources to prioritize either one).