I know exactly how, why, and where their system is wonky and why it will keep putting people in the wrong places.
There is an issue of stability and reliability in a ranking system that is basically something called “Bayesian priors” and they way they created that data was from faulty theory, and they still don’t have nearly the amount of data they need to make their bad theory for ranking (Based in ELO) functional.
Google this: “Elo vs. Regression to the Mean: A Theoretical Comparison” if you want to deep dive
They created permanent instability by refusing to use individual statistics and now it’s showing more than ever, They keep using win/loss primarily and it’s just, purely, scientifically wrong. There IS a way to fix their kludged and wrong-headed system design though.
Two overlapping systems. Individual play focused at low rank but back to team focused (win/loss) at high rank.