Is it because if White wins the first stage of this two-stage ko, White might choose to fight the second stage at a rather than at the more conventional b? (This got something to do with what happens if White loses the ko during the second stage.)
tapir: Should not a be the standard answer as it is always better when losing?
MrTenuki: Would you mind about showing why White would want to play at a? It seems to me (as an SDK) that b makes more sense unless the left side is very important to White because the a option would require more threats for White:
(- = ko threat and reply, takes back ko at , = tenuki, at )
tapir: It is one stage ko (direct ko) now, white can resolve the ko by playing at a.
unkx80: Me bad. I forgot that connecting at in the b variation still results in a direct ko. Still, I don't get what's rogue about this ko.
Bill: Well, I am the one who dubbed it a rogue ko. The term was introduced by Berlekamp and Wolfe, but they did not define it. I used it because the ko is not just an approach ko, but it is not a direct ko, either. I think that was their intention. This ko does not fit the standard classification scheme.
tapir: All I can see indicates this is an approach ko for white...
Bill: By the standard classification scheme I mean the one that counts the number of ko mouths to take before ultimately winning the ko (see iterated ko) and the number of approach moves to play before one side can win the ko (see approach ko). Since White can make an approach move on the top side or hane on the left side, this ko does not seem to me to fit that scheme.
tapir: But isn't it one approach move at either a or b (maybe that there is a choice of approach moves makes some difference, but a new type?) with only one ko mouth to take?