Shusaku Fuseki / Review With AI

Shusaku opening  

It is not fair, nor probably very useful, to review Shusaku's fuseki with AI, as it dates from the no komi era, where White was encouraged to play actively (slightly overplay) from the beginning, to overcome the disadvantage of going second. Vice versa, Black could play conservatively (passively) into a steady, marginal win. This is the context for , which doesn't take the opposite corner and , which was hailed for its steadiness.

Interestingly, after , KataGo thinks is one of the best moves and its main candidate during the first 2K playouts. Approaching the other corners are obvious other candidates.

But let's go back to the reasoning behind .

Ideal double purpose play  

If takes the last open corner, as is AI and professional gospel, then makes an enclosure, threatening to make another one. White "has to approach" at , which gives Black an ideal move at , a double purpose move. This was the thinking at the time, according to this page about allowing two enclosures.

KataGo, like LeelaZero and other AI, doesn't like pincers in general. is evaluated as a 1 point mistake. In fact, it allows White to play the kind of active game that might have helped in no komi days to catch up.

Active game  

First White pushes Black along the 3rd line. Black gets easy territory, for White influence, which is allegedly "non-AI play", but the point seems to be that White can next attack the Black pincer at a. When that stone jumps out to defend, White can attach at b to reinforce itself. At that point KataGo gives W ~3,3 so about -3,2 on the board. If Black follows standard play, the advantage may even approach komi. None of these moves, not even b is novel, so this flow was available at the time.