Alright, let’s break down the way the rating system works.
It always tries to get to the average
If you’ve got, let’s say 1800 personal MMR, but you’re teamed up with two players on 1600 personal MMR, then it uses that initial MMR to determine your team’s MMR which in turn decides the opponents’ average team MMR. This is how it works in the normal ladder, we all know that, yeah?
So with it working like that, it keeps matching you up with opponents to increase the certainty of your rating as a team. Meaning if you played a team with 1700 average MMR, and won that, then it’ll try to match you up with something slightly higher, let’s say 1725 average team MMR, but your own average team MMR also increased by 25. If you win that too, then it’ll try to increase the MMR even more, even though the difference in the MMR between your team and the opponents’ didn’t change.
This effect is from the GlickoRD systems, because it tries to place people “where they belong” faster and faster if the win/loss rate shows you really don’t belong where it’s matching you.
In the Elo system this wouldn’t change from streaks, but the Glicko systems are all about stabilized ratings and destabilized ratings, to ultimately increase the theoretical accuracy of the system compared to the Elo system.
Anyway, so this rating deviation function is very unclear about how it works in the shuffle system, since it doesn’t give out wins and losses rating-wise until the shuffle has ended and then it gives 'em out all at once, so it’s kinda weird.
But anyway, so going back to the example, that’s what it normally does in the normal ladder. It’ll keep throwing opponents your way as a team, MMR being affected by your wins and losses, with the ultimate goal of making all of your ratings equal. With every win, the ones lower rated than you win more than you do, and with every loss, you lose more than they do, and eventually you all end up on the same rating.
Fun, right?
Anyway, so that part about winning and losing more, is 100% present in shuffle’s system as well. It’s trying to do the same thing there, the lower rated ones wins more than the higher rated ones, because it’s designed that way, which serves more of a purpose in the normal ladder than in the shuffle ladder.
So how can this be improved for the shuffle?
There is a way to improve it, but Blizzard kinda copypasted way more than they should have from their normal rating system. Also, to be clear, the rating system Blizzard uses is an “original adaption” of the Glicko system, which Blizzard has been using ever since wotlk started (not the classic wotlk that’s up now). They’ve also further developed it ever since then as well, so it’s not entirely the same as Glickman’s systems anymore, but the core functions are still similar.
So the way to improve it, is to slow it down. It may sound counterintuitive, but slowing it down allows for functions like putting in a maximum rating lost & won when you do either, as well as a minimum rating won when you win a round.
Through the function that puts a maximum as well as a minimum, it would let the shuffle ladder reward the win rates more “appropriately” for the game mode. It’d make people reliably win rating from 4+ wins, and lose rating from 4+ losses, but without being punished by the insane MMR matchups the game is forcing players into in shuffle.