Quote:
Originally Posted by feelings
Seems epistemologically equivalent to suppose a range of hands in set {a} and to estimate p(b|a) over set {a}, "without tons of reads and history."
I guess my point was that just choosing p(b|a) = 1 may have benefits in the form of robustness and ease of calculation that make it preferable. I am not sure that most players are consciously making this choice, and they should think about specific situations where the prior should be shifted meaningfully one way or another from uniformity. Your post certainly got me thinking about this. This might be one of them.
In theory, every hand except ours and the board cards exists in our villains range. The hand ranging above is making a shortcut by assigning some likelihood cutoff and discretizing to {0, 1}. In that likelihood, we are considering p(b|a) for sure. So the question is what cutoff are we using. Then, given that cutoff, what is the worst case equity based on this range. This analysis would be sick.
How badly this estimation can be is clearly situational. I'm not sure if you had some insight into how bad this approximation is or if you were just pointing out a gap (the latter is interesting, the former would be really telling and would probably sell a ton of e-books).
Tldr: I think your point is valid for sure. I think epistemologically valid too
Sent from my SM-G900V using 2+2 Forums