Desirability of Outcomes in Local Life and Death

Alex Weldon: In poker, two pairs beat a pair, three of a kind beats two pairs, etc. Similarly, in Go, there is a hierarchy of desirability for local life-and-death situations. Of course, it is not set in stone. For one thing, it does not include things like capturing part of a group, while the rest of it survives. For another, it's taken as a given that ko is better for the attacker than seki (based on the assumption that the value of a ko to kill is, on average, going to be half that of an unconditional kill... always greater than the value of a seki, except in exceptional circumstances), whereas in reality, that depends on the ko threat situation. Lastly, depending on the temperature of the board, it may be better to take sente than to live in gote, especially if the group is a small one.

Nonetheless, here is the list of possible outcomes, from most desirable to least. Treat these as a rule of thumb for tactical situations arising in an actual game, but in the case of tsumego, they are a generally accepted convention. Note that it is assumed that killing the opponent's group assures life for one's own stones.

• Kill in sente (that is, kill and threaten to kill another group if the opponent doesn't defend)
• Kill in gote
• Favourable multi-stage ko or approach ko to kill
• Regular ko to kill, taking first
• Regular ko to kill, opponent takes first
• Unfavourable multi-stage ko or approach ko to kill (this in particular may or may not be better than a seki)
• Sente seki
• Gote seki
• Favourable multi-stage ko or approach ko to live (again, might be better than seki)
• Regular ko to live, taking first
• Regular ko to live, opponent takes first
• Unfavourable multi-stage ko or approach ko to live
• Die in sente
• Die in gote
• Kill self in gote
• Nuclear tesuji