| |
BGonline.org Forums
Philosophical question on higher level rollouts.
Posted By: Daniel Murphy In Response To: Philosophical question on higher level rollouts. (KenB)
Date: Friday, 12 February 2010, at 8:56 p.m.
Tangential to Ken's question is how well ply-induced bot errors correspond to actual differences in skill among real players. How comparable are the categories and magnitudes of errors that bots and humans make? Ignoring that, and knowing nothing about the particular failings of either Ken or his opponent, I'd say that the reason best play is best play is that it is more likely to lead to sequences where you're on roll with opportunities to increase your equity, regardless of whether your opponent has erred during his turns. And, for the imperfect player, I think of best play as generally leading to easier decisions, even though I'm sure everyone can think of exceptions, since nothing's easier than choosing plays so badly as to guarantee that you play out the rest of a game on the bar.
In Ken's hypothetical, he's rolled out the position several times and, with typical respect for bot rollouts, he most likely believes he knows what the best play is. That's play "C," the high-ply play. That's the play he should make. He should make some other play only if he knows something about his opponent's imperfections or his own that leads him to believe that another play will be more profitable because he would be more likely not to misplay, or because his opponent would be more likely to err.
By the way, though, how many positions are there where 2-ply vs. 2-ply, 3-ply vs. 3-ply, and 4-ply vs. ply rollouts prefer three different plays?
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.