YACHT Diary

[dfan]

I have been told by dan-level players that the attach-and-extend joseki in general is only suitable for special circumstances, and that it is better to avoid it until I can understand those circumstances better. The usual reasoning given is that it settles the shapes (especially Black's) prematurely and makes it difficult to carry out any operations against Black later.

[Revilo]

I don't know what it is about the tsuke-nobi, but it doesn't appeal to me either way. Playing it myself, it feels so heavy. Playing against it, it feels so massive. However, I've only just begun looking into the push and cut variation - especially here after 16, it might have been useful to support a black position along the side. Splitting with 17 as played in the game just gives White an extension that felt really large, prompting me to stir up trouble.

By the way, does anybody know why the letters I refer to aren't displayed in the viewer? Without them a lot of my comments remain rather unclear. They are visible offline in Drago.

[Bill Spight]

I would not have thought of 4% as "whopping". Am I off base on that?

Remember that these bots have mainly been tested against other versions of themselves (AKA self-play). That does mean that the rank ordering of the moves has been tested, since if the two versions pick different moves, they order them differently (unless there is some randomness to the choosing, but even then the order matters). The percentage difference between plays has not been tested. AFAIK, not in any way. If the better version prefers move A, while weaker version prefers move B, that says nothing about how much either version prefers one to the other. OC, the percentage difference is not meaningless, but moi, I would put more stock in the rank ordering. If your play is not among the top five picks, the odds are good that it is a mistake.