I do use alpha-beta+evaluation. The issue isn't a/b, but the high amount of
pruning required by the slow evaluation function. The real debate should be
whether one can make a good enough evaluation function without using local
tactical search inside the evaluation. That's what makes evaluation so
slow. I believe local search is required since if tactical stability of
stones is not established, there is no way to estimate territory.
It seems like a static/semi-static evaluation based mostly on patterns
and heuristics can only work reasonably well after many stones are on
the board. The deeper a search gets the better it should estimate
groups and group status and thus territory, therefore an evaluator
should be conservative in early stages so as not to lead the search
astray with its errors early on, and the search needs to get really
deep. The potential to dig deeper is why I like the multi-level search
ideas.