The general gist of this basically asks for more “personal based metrics”, let me explain that Overwatch’s skill rating adjustments did used to use skill-rating adjustments at all ranks (and still does at Platinum and lower ranks). It didn’t work very well.
The biggest noticable trends back in the day is that performance adjustments could only be effective when comparing the heroes that player plays in comparison to other players who played the same hero. Not to that player’s teammates who would play different heroes. This allowed some really undesirable behaviors especially during seasons 4 through 7 where players could play “off-meta heroes” (like Torbjorn or Symmetra which were extremely underpowered at the time) or heroes with a relatively low “skill-ceiling” (like Mercy). Players could play heroes like these and climb with miserable winrates as low as 30% and still effectively climb, this also caused players who played more win-effective heroes that were considered meta to often lose those games when such players one-tricked non-meta heroes. This is why skill rating adjustments was removed for players at Diamond rank or higher starting in season 8.
In the end, I am not convinced your proposal is viable or effective to resolve the perceived problems of players getting boosted or smurfs. I also would point out that Overwatch skill rating and matchmaking system is far more complex than you have claimed it to be. For more details, take a look here: