Fastest score estimate algorithm

[Javaness2]

Beyond just counting the number of stones on the board, does anyone know what the fastest algorithm is to compute a rough score estimate.
I'm aware of a few methods, but I've never seen any benchmarks.

[Schachus]

If you are looking for a fast algorithm, you should specify exactly what it sould be able to do.

Else estimating any poisition as equal(i.e jigo(perhaps checking before whether komi actually allows jigo)) would be fastest...

Obviously you dont want that, but you want to take the position into account somehow(so what does an estimation have to fullfill, so that you would say "ok, thats good as a rough estimate"?). But I think very naive estimations might easily be worse than not looking at the position at all, saying "no clue, probably equal" and if they are more involved, they are probably not "fastest".

If you allow for preprocessing then the good choice is probably training a shallow value natwork to do it(they are very fast if you dont count the time needed for training before), I would guess

[Kirby]

Neural network, probably. You can gather input features that are quick to compute, then train on real games.