| |
BGonline.org Forums
Comments on Chuck's blog post (long, picky)
Posted By: Timothy Chow In Response To: Comments on Chuck's blog post (long, picky) (Jeremy Bagai)
Date: Thursday, 5 February 2015, at 4:35 a.m.
I liked Chuck's post too. But as long as we're nitpicking...
So the aversion in question isn't to risk, but instead to perceived criticism just as you mention.
Suppose that the coach's utility is derived primarily not from the outcome of the game, but from what kind of criticism or praise will ensue. Then the conventional play will likely result in muted criticism/praise whereas the unconventional play will likely result in harsh criticism/extreme praise. Choosing the conventional play can then be regarded as risk aversion if the coach is motivated by reducing the variance in the criticism/praise.
Similar comments apply to the backgammon example. Making the bold play might cause me intense elation when it works and intense regret when it doesn't, compared to making the conservative play.
"Game Theory," in its most accurate and limited definition, assumes that all players are perfectly rational actors who never make mistakes.
This sounds out of date to me. Modern game theory often relaxes the assumption of unlimited computational power. Computationally bounded players will make mistakes even if they are rational.
Oh, and one for Chuck...most chess programs aren't based on "neural net" algorithms. They're still at bottom fancy alpha-beta searches with some static evaluation function. The static evaluation could be a neural net but typically isn't, and certainly neural-net algorithms don't have anywhere near the importance in chess as they have in backgammon.
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.