| |
BGonline.org Forums
An interesting hypergammon programming project
Posted By: Timothy Chow In Response To: An interesting hypergammon programming project (Tom Keith)
Date: Thursday, 4 September 2014, at 10:01 p.m.
Tom Keith wrote:
But if there is no obvious fish strategy, then you'd just be picking a fish strategy at random. And the answer you get is going to be random too, isn't it?
I'm still not sure what you're getting at.
If the non-canonical choice of the fish strategy bothers you, let's make the following canonical choice of a fish strategy: The fish chooses among all possible legal moves uniformly at random. Does that address your concern? I'd still find it interesting to know how much better the shark does against Canonical Fish compared to "perfect play" versus Canonical Fish. Do you think it would be a lot or a little? I don't know. Either way, I think we'd learn something.
I don't really understand what you mean by the word "random" in this context, but let's suppose we did literally generate a large number of fish strategies at random somehow. I still think it would be interesting to compare, for each fish, shark-vs-fish with perfect-vs-fish. Is it a big difference? A small difference? How much variance is there across the population of fish? Again, I think we'd learn something from the results.
Even more insightful, in my opinion, would not be to generate fish strategies at random, but to create caricatures of certain styles of weak play. Then it would be interesting to see how different the shark strategy is from perfect play. Perhaps it doesn't deviate much; perhaps it does. Perhaps it deviates in an intelligible way; that would be the most interesting case, because then you would get some insight into how to exploit specific weaknesses.
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.