And you can expand it to “any estimator is bad if you don’t understand it”.
If KGS’s estimator is good or Fox’s estimator is good it’s most likely because a player using it understands when it’s reasonable and when it’s unreliable.
Of course when it’s unreliable more often than not (like the weak score estimator is), then it comes across as useless. Or you might have to do a lot of work like adding more stones or capturing dead stones to make it reliable. Time is less a problem in correspondence (though maybe tedious) but not great in live (and people do use them in live games in various places and servers.).
I think the clearer we can categorise what we want a score estimator to do and what we don’t want it to do, then the easier it is to fix it. Without a clear expectation, it’s hard to improve something.
Probably we can have that discussion in a dedicated thread, either a new one or continue and old one like
Or this one (linking to suggestions of using a flood fill rather than playouts)
Or a manual score estimator
Useful to note that it was decided to remove a better (Katago) estimator because it was too good, which was probably a good decision. You can see the breakdown of when the old vs new estimator is/was being used.