you say that you are calculating exactly what the proper bet size is to elicit a call when villain has a 35% chance of winning because it is rare that he will have more than that by the turn. so, essentially, overbetting by too much will fold off the hands you want to call your push, right?
No no no otherway around. The 35% is *us* winning (sucking out) not him. Obviously if we shove with 50%+ equity we're +EV by definition whether he calls 0% of the time or 100% of the time. The question is when we shove with LESS than 50% equity -- how often does our semibluff shove need to work? That is, how many times do we need him to fold and we take whats in the pot compared to how often he calls and we lose the pot. Does the former pay for the latter?

I see your point. You're saying if we shove xBBs and get called y% of the time, then when we shove MORE than xBBs we should expect to get called LESS than y% of the time. In a vacuum this is probably true, but I think after villain sees us do it once, it won't be true anymore, and we can easily start taking the worst of it when villain tightens up his range and nutcamps us.

if you size your stack/bets to where your push "should" bump out 45% you not only pick up some more pots when you were behind, but you also force him, if he decides to, to call with even worse odds....making it an even bigger mistake by the FToP.
The problem is where do you draw that line. Extreme example for the sake of well, example. You buy in for 1000BB and shove 996bb into a 4bb pot. He'll fold an enormously huge % of the time, and when he calls when he's behind he'll indeed be making a HUGE mistake as dictated by the FToP. In that aspect the above quote is dead on.

But your logic breaks down when you realize villain will only call a bet that size when you are drawing dead dead dead. Yes he folds a lot and we take the pot, but we don't take it enough to compensate for the one time he calls with the nuts. The amount of times he folds just won't pay for the times he calls and busts us.

Thats what the EV chart shows: How often do we need to take the pot (or suck out) to compensate for the fact that when he calls and wins we lose our stack? So if we estimate our villain will call our shove 40% of the time then we need to suck out about 20% of the time in order to break even. Obviously if we suck out more than 20% of the time (we have greather than 20% equity) we make money, and obviously if villain calls less often we make money. However the converse is that if villain calls MORE than 40% of the time and/or we suckout LESS than 20% of the time, we lose money. That simple.

Hope that cleared it up.