Game Shape would now be calculated by a players playing time, and his performance stats related to how consistent his stats are.
Each Player has a two part score, x.y (ex. 8.4), x is consistency y is confidence.
X is the normal game time related GS we have now, but Say a player has a great game earning a 11.0 rating at that position coming off the bench(when normally he'd play around 7.0 or 8.0 rating).
So whenever the fan survey is done the players y variable is reassessed between 1-9. So after an amazing game a player could be 7.8 or 7.9.
SO, that player is playing more like a 8 then a 7 and 7.0 or 7.1 is more like a 6 then a 7.
I think it would add a variability to the game particularly at higher levels D1's and NT, and also give that chance that bad players with good game shape could have a great game as we know can always happen.
basing it on rating would be a bad idea, also since the rating isn't that much depending on the player play, and more reflect the ability to play on circumstances. Your example with him coming of the bench with higher rating is one think you will notice on every player, since he can play harder since he isn't on the court so long, also the rating raise in the begin of the season through raising GS himself(with your system a GS down will provoke the next one) and raising enthusiam(and i think the punishment for a Ct is already quite high).
Making it on scoring etc. performance could make sense, but as manon asaid training off position is already hard, but will also lead to weker stats, and beeing star player on a weak team don't mean that you are always in good shape.
My personal wish would be limit the impact of voverplaying player, since this leads to forfeits ... Even when we them need also other changes, that team will still invest into a deep bench(like higher decline in performance with playing time etc)