Computer programs sometimes explore some wild variations. Maybe it is fun to collect some of these and try to understand them?
I guess I don't follow actual computer play but instead variations suggested during analysis.
Here is one...
It is a game between pros and white is ahead and it is time for black to do something. In the game black played "a" and never had a chance. I was watching the game and thought maybe black should play "b" and try to take advantage of the fact that black can attack in the center if he needs stones on the outside. KataGo actually agrees but wants to play some forcing moves first at "c" or "d". Now, I change the rules to "no komi" to make an even situation and KataGo actually wants to play any of "a" to "d".
Another idea is to attack first in the center and that is where we get to the wild variation. It is only 0.5 a point or so less than the best moves and it is not that unexpected until after the first diagram. White's first move could be different in practice but throwing away the tail is cool.
at marked stoneNow the next move is wild
I think it is undoubtedly correct and surprisingly KataGo sees this from afar at even only a few playouts, even though it is not the only move in practice. Probably some L19ers have the imagination to find this move
Hint 1
Hint 2
The variation
at the marked stone
and onward is that it quickly fixes the shape in the center and on the whole board while leaving white ahead. This is somewhat appealing in the komi case, white is ahead and there isn't much left to contest. However, if the estimate of white being ahead (only by 0.5 in the no komi case) was incorrect then white could in a way be resigning the game by fixing the shapes everywhere.