I think there is absolutely also a non-cheating group of players that engage in behaviour that the power levelling change is supposed to catch. I think this behaviour does result in very high ratings. An issue is these players are, I believe, not doing it primarily for the rating
.
The basic problem comes down to that the matchmaking systems that BGA has are bad at giving you a fairly equal opponent. You could get close opponents, or somebody several hundred rating points higher than you, or somebody several hundred rating points lower than you if you queue up an arena game or take any simple game you can, or enter an open tournament. I don't fault BGA for this, it's good at getting games and it'd probably be unreasonable to be too picky with opponents you can get and still keep good queue times (they're already very long in some games).
If you're a high rated player it's reasonable that you may get frustrated with huge mismatches against much weaker opponents, because the games themselves are so lopsided, the outcome is predictable so you don't get much elo, and if it goes wrong you lose a lot of elo. This is not to mention multiplayer games where one weak player can randomly kingmake for any of the other players with weak moves. So some high rated players are going to look for a solution to the matchmaking giving them mismatches.
A very easy solution to having mismatches for high rated players is to get in contact with the other high rated players and arrange games between themselves. This will be a small pool of players fairly closely rated to each other playing each other probably repeatedly. Variance in small samples being what it is, somebody is probably going to go on a winstreak in these games. This is true even if none of the players in this pool is better than the others, but very likely one of them are.
The issue with these games comes if they're rated games (and they probably are). You have a pool of players that are closely rated - so if winstreaks happen, that player is going to be getting a lot of elo because they're playing peer opponents (and winstreaks for someone are very likely). They're also already highly rated - so a lot of elo means almost surely they're getting to the #1 spot. Anyone not doing this has to compete with much higher changes per game as they are matched against much lower rated opposition on average - which is very hard to keep up with.
How do the powerlevelling changes affect this behaviour? The answer is basically not at all. Absolutely nothing was done to incentivize players to move back to playing the low rated opponents - it was just a disincentive from playing the same opponents. If the community is sufficiently small enough this change actually disincentivizes playing low rated opponents for high rated players who care about elo. As far as the disincentive to playing the same high rated players repeatedly - as the goal is not elo, but getting better matchmaking than BGA does, it doesn't matter.
As far as just curbing the effect on elo of these behaviours, as eliminating the behaviours seems difficult, I don't think the new system does enough. The rating changes per game will still be much larger in these games with the powerleveling change than against weaker opposition. The entire issue here is that a small winstreak in such pools puts a player at the #1 ranking, and this can happen before the powerlevelling changes detect it and kick in with large penalties (even though such penalties happen absurdly fast in any other context
).
