A function to make stockfish play the best move (more precision, less elo)

[humble jasper]

Beware that "increase perfect play at the expense of losing some Elo" doesn't mean anything

Stockfish search parameters are already massively tuned, and the games played for tuning begin from positions whose evaluation is about 1, so I doubt doing less pruning there would work

And the idea 2), simply isn't compatible with an alpha beta pruning framework

Stockfish already does its best at sorting moves from best to worst, and adjusting depth depending on how good the move looks

Assuming I got what you wanted to say, because point 2 is really really ambiguous.
(eval difference between what?)

[golden iron]

for you, perfect play means fastest mate?

in this case sf is not optimal indeed. The idea of an engine that converts a winning position as fast as possible seems interesting

but stockfish's aim is to be as strong as possible, so i don't think that work will be done to improve "conversion speed"

i wonder if there are engines that do this though