Shake Appeal
Member
If it's more or less like ELO, I think all Kaplan is saying is that if your team is, let's say, a "50" and goes up against a "70" team, your MMR won't drop much (or at all) if you lose, because the game was expecting that outcome. On the other hand, if your 50 team beats the 70 team, the system has to make a more radical correction to both teams: the 50 team appears to be better than it thought, the 70 less so. On the other hand, if your 70 team beats a 70 team, your MMR will only go up slightly and theirs will only go down slightly, because things are more or less as expected. Over time, with more data, your MMR become less elastic and it takes more positive results for the system to correct upwards.No it probably goes down but not as much (assuming you were high MMR on a low MMR team against a high MMR team). Who knows though.
I think the system probably uses some sort of average of the players in the room to try and make matches, which means if you're an 80 player grouped with a bunch of 20s and lose to a team of mostly 50s, you'll drop more steeply than the 20s on your team will: the system expects you to do better.
This is all speculative, but it's a useful (if strictly inaccurate) way of explaining how these things work.