However the combination of a good evaluation function with a reasonable
deep search like 10 ply, is really going to kick major butt in go.
I don't think 10 ply is generally deep enough to get reasonable and reliable leaf node evaluation in the early and mid stages of the game. Depending on the stage of the game and how it develops, I believe more like 20 to 40 ply will be necessary. But this depends on the scope of the evaluator. If evaluation includes complex life and death analysis, local searches, ladders/ko fights analysis, patterns/rules for good/bad shape then the evaluation itself will be extending the search depth in hard to quantify ways.