tderz: Bill, a question: How did you calculate this?
tderz The probability that Black gets the free & sente exchange -, is not 100%.
Is this somehow recognized in the evaluation?
Bill: No. The principle for evaluation is that stable followers that are sente are assumed to be played with sente. In practical terms this is because of privilege. The player with sente can normally make the play with sente. In theoretical terms it is because the value should be consistent in different environments. See the method of multiples.
Did you take this as alternative reference diagram (also gote for White), take the difference (to the tobitsuke diagram) of 4 points (a, b, c, d) and divide by two?
Should/could the value be something between 1.5 points (= only 50% chance that black gets it) and 2.0 points (= 100%, black gets the exchange for sure).
Bill: No. Remember, the value of a sente is really the value of the reverse sente. To set it as less would suggest that White should play some gote with a value less than 1 instead of playing the reverse sente. That is normally not the case.
And that's implied by evaluating the White territory after as more than 2 points. And that's implied by evaluating - as less than 2.
The lower probability could be justified if there are more 20-pointers around (having no further continuation).
I estimated the total value of that corner as 20 points (could be 18, because black needs 1 move to kill in the tobi-tsuke diagram.
tderz]: or, after realizing her mistake of in the hanami ko diagram (if it was one, i.e. not intentionally, e.g. with a 2nd or double-ko somewhere else), white could also go for the sente-seki with capture .
Could that be expressed as 2.5 point loss over the tobi-tsuke diagram?
Bill: If you don't count points in seki it's a 1 point loss for White.