by dracula on Wed Feb 12, 2014 9:59 pm
I don't know that adding new semantic classifications for this is really that useful, personnally. Joseki, I.E. "locally equitable" or "locally ideal" sequences, cannot be seen as context independent in the first place, and then the choice of a non-joseki variation is always going to be highly sensitive to the global position, whether it's considered normally a "good" but non-joseki sequence or normally a blunder. The biggest issue in my opinion is gettting good commentary associated with the positions. Semantic labelling can get arbitrarily sophisticated, but will always be limited by the underlying science (game theory, semantics itself, etc.), and finer grain is not always a good thing.