Its not always the exact rank because especialy at lower ranks, the skill balance often goes bad. Its what called MMR Hell. For masters the problem isnt realy there as they generaly dont involve a lot of AFKers, smurfs and trolls. So thats why for them it feels balanced.
MMR hell is the effect that is caused because players with equal MMR vary that much in skill that no match becomes truly balanced, and its caused by rapidly changing MMR values together with a very low playcount
I can imagine the matchmaker counter balancing this by also looking at the direction in which the MMR is changing and slightly balancing it out. So a 1000->1400 player is considered 1200, and a 1000->600 player would be 800. Or a 1300->900 player is considered 1100.
But thats where a problem starts. Quickly changing MMRs show a part thats unreliable. And trying to compensate that is a bad thing to do. A repeatedly losing player should not be granted some safety. He simply should drop as fast as the game tells. If he is better he will just rise again.
And he might bump between 1300 and 900 that way. But so be it. For the 1200-1500 player, he might sometimes see this player and lose in that match and get even lower MMR, but at that point he can carry a lot better and easily rise out of it again.
The problem can also be that the 1200-1500 player is kept down because he often noticed those who bump between 1500-1800, but at least this can show that a player is truly a weak link. Rather than creating some sort of artificial balancing that dumps 1 high and 4 low MMRs towards 5 balanced ones.
And even then, those masters who get 60% win rates do actualy show that there is an issue, since even with their rank… why isnt it 50%? With 500 games, reaching 60% means you won 100 more games than what you lost. And although thats a quite significant number, it can be explained:
Its because they had to fight themselves out of diamond and they clearly won more there than they lost, and 100 games is simply the treshold needed to win enough to get bumped to their actual rank. After 1000 games they should balance out towards 55% again.
The only people who should be able to get out of the 50% rate are those at bronze 5 and top 100 GM. As these dont see any players more excessive than themselve.
Hence, i think only the last 200 games are reliable to be used as win rate if someone had more than 500 games. And i think most masters also face a 50% winrate there.
50% win rate is forced because thats how a good MMR system should work. And off cases are only there because of the season start dumping many skill levels together.
Stating that someone belongs at a certain rank isnt always simply true. there are too many factors going around, and only if you can ensure players are reliable enough to be measured, then you can truly say it works as intended.
Those 10 games per season players are a perfect example of incorrectly measured players. They dont play enough for it.