Janzert
Forum Guru
Arimaa player #247
Gender: 
Posts: 1016
|
 |
Re: search tree techniques
« Reply #30 on: Aug 6th, 2008, 8:19am » |
 Quote  Modify
|
on Aug 5th, 2008, 11:03pm, aaaa wrote:
But wouldn't you agree with the reasoning behind why having a fixed 4-step reduction then should be problematic? |
|
Theoretically yes I can see that a fixed reduction (of any size) shouldn't be ideal. It doesn't violate my reasoning that led to using a 4 step fixed reduction though. That reasoning went more or less like this. I started with a 3 step fixed. Bumped it up to 4 steps and it seemed to be better. Bumped it up to 5 steps and it seemed worse. Tried half a dozen to a dozen different dynamic schemes and none of them seemed to be an improvement. So I just stuck with the simple fixed 4 step.  Certainly it's not a resolved question and should be explored more in the future and probably changes on the depth of search being done as well.
One thing about early bot development and OpFor is certainly still in this phase*, I believe it's more important to get reasonably working components in place than it is to get a perfect component before moving on to the next component. This philosophy pops up all over OpFor. For another example I used FAME material evaluation, not because I think it's the best possible but because I've implemented it before and knew I could do it quickly and it would provide do decent job in most situations.
Janzert
* Actually I think all Arimaa bots today are still in this phase.
|
Home
Help
Search
Members
Login
Register
Reply
Notify of replies
Send Topic
Print
Author
IP Logged