Sounds like you want something like Trueskill 2? https://www.microsoft.com/en-us/research/project/trueskill-ranking-system/
Or if not? There are issues with manipulating teams inside MMR ranges. Your MMR rating loses accuracy. If MM is not related to your MMR values, if the steps used to generate your MMR value are flawed you throw off the curve and your distribution is broken. Is why I am constantly frustrated by "forced 50%" Arguments. Forced 50% wants to move bad up and sink good players. MMR wants to do the exact opposite. MMR values and expectations are built from the natural trends of bad players to lose games and good ones to win. They get more accurate with time and data. but if you remove all value from the data (its supposed to naturally get built off wins and losses) Then you lose all meaning to a rating, then your system has nothing to use to accurately rig matches.
Then usually I get Riot is using some statistical system to judge player skill and use that to Rigg matches. You already know how I feel on the **massive **difficulty of making a subjective skill rating system.
https://imgur.com/a/AtZMR2W
I am not great with MS paint. So left, That's what happens if your not doing a good job. It can also be a good representation of a high uncertainty MMR system. When functional it will turn into the graph on the far right after being allowed to run where everything collapses into a curve. X Axis would be "skill" Or win %s. Built from the distribution theory your using. Logistical Typically. Y would be your assigned rating.
MM is about taking players in a range to build your curve. You want to make reasonably fair games so that every time system is correct OR incorrect on its assumptions its moving players the proper directions. You want your system to have the option of being wrong because there's no point to using a self correcting method if you don't.
Elo and MMR are more then Wins. Its wins, losses, and who your playing with. There is probability theory involved in this, which is why adding changes that damage probability theory damage the ratings.
Stat MM. this is still stat rating. Your judging the player and assigned a value to them. under, average, above.
Performance has to be taken into consideration or else you'll end up with a team of 5 mid laners
I played seasons 1 and 2. That used to be a thing. My opinion is biased on the topic so take this with a grain of salt. If you get stuck in an MMR and are an OTP? Your a less skilled player then everybody in that MMR who can play multiple champions and roles. Your a handicap to your teammates if you don't get to play the 1 thing your remotely decent on and you deserve the losses you get when you fail in other positions.
Riot did the exact opposite of what I woulda liked. They catered to more casual players by trying to fix issues created by them with positional Que positional MMR. But casual players don't seem to care about competitive integrity so the whole positional Que was a bust.
Riots still playing with positional MM but they hid it again like most things because the player base cant be trusted to not abuse systems. Call me old fashioned but having a shallow champion pool is a weakness on a player imo.
First you learn how to get kills and not die as often, then you learn that if you can make plays with no kills and all the deaths, stats won't matter.
Stats and Winning is a wild system of checks and balance lol
So... I'm thinking your likely wanting something similar to True skill 2. Is that correct?
Or do you want to see Stats only used on a portion of the Ladder?
Sort out some chaff low skill players in lower MMRs?
