You start simulations from the current position, so at move one it would be from scratch, correct? Then for the next couple moves you still need your 100 simulations even if a couple moves have already been played out. And in each simulation you need to get to a point where you can evaluate the result.
Technically that’s maybe not 500 full games, but I would say pretty close.
That’s what we are saying (see @flovo’s reply for example): this is called reading for human go players, so we do it already. That part has nothing to do with MCTS. There’s even an analysis feature on OGS just for that.
Fair enough So what’s the expected benefit? Just mechanically push up your ranking?