It fixes the AW matchmaking because alliances with people many 3/70 5 star heroes has an OVER INFLATED team power number so if they get matched against an alliance that has mostly max 4 stars at 70, the alliance of 4 stars will be much more effective because 4 star 70 are more effective than 3/70 5 stars.
Think about each hero having a superficial rating and an actual effectiveness rating. 5 stars at 3/70 have a higher superficial rating because of bogus “team power” formula than max 4s but a lower effective rating. So when you add that up aggregate across an alliance, when one alliance has an abundance of non-max 5 stars in their top 30 of each player, it gives them a higher superficial rating (i.e. team power) which is what matching is done on.
It’s sort of like how in sports betting, people always bet a lot of money on the “public” teams like the Yankees, Dallas Cowboys, LA Lakers, and Green Bay Packers. Those teams’ odds are generally less favorable to the bettor than they should be since sports books know people will bet on them even with crappier odds than the game calls for. There’s a built-in penalty for betting on those teams because of how much action they will get.
non-max 5 star heroes are the same way - there’s a “team power” penalty (i.e. power rating is higher than it should be) on them.
here’s an example, it’s not 100% accurate but it illustrates the concept
let’s say a 4 star max 70 has matching rating of 80 but actual effectiveness of 75. Now take a 5-star 3/70 who has a matching rating of 100 but an actual effectiveness of 70. If you have Alliance A of 30 players with 30 4 star heroes as their top heroes, you have a matching rating of 30x30x80 = 72,000 and an actual effectivness rating of 30x30x75 = 67,500. Now take alliance B who has all 5 star 3/70 as their top heroes. Their matching rating is 30x30x 100 = 90,000 but an actual effectivness of 30x30x70 = 63,000. So alliance A is rated as only 80% as powerful as Alliance B in the matching (72,000/90,000 = 0.8) BUT they are actually the stronger alliance. In actuality, those alliances probably won’t get matched and Alliance B will get matched potentially against an even STRONGER alliance, depending on that alliance’s mix of non-maxed 5 stars.
Now that was a bit extreme of an example since no alliance has 30 players whose 30 heroes each are all the same like that, but it does illustrate that for non-max 5 stars, the team power metric thinks those heroes are much stronger than they actually are and alliances with those heroes get matched vs stronger alliances than they otherwise should.
The change will simply be to bring the team power (i.e. AW matching rating) down closer to the actual effectiveness rating of the hero.