[computer-go] 10k UCT bots

[dhillismail at netscape.net]

> -----Original Message-----
> From: Jacques Basaldúa <jacques at dybot.com>
> To: computer-go at computer-go.org
> Sent: Wed, 14 May 2008 6:38 am
> Subject: Re: [computer-go] 10k UCT bots


> Don Dailey wrote: 
 
> > dhillismail at netscape.net wrote: 
 
> >> For those currently coding this up, I think the most important thing 
>>>  about this playout algorithm is that it is 
>> >  *temporary*. You will  almost certainly be replacing it with something different and better 
>> > just a little bit down the road.  So you probably don't want to worry 
>> > about hair-splitting tweaks except as an academic exercise. 

> Yes, I agree. Also my hair brained scheme of pre-generated tables of
> > list traversal orderings was just an academic   exercise as you say.  

> But the problem is that when you do heavy playouts you have the same 
> problem except that the probabilities of the legal moves are no longer equal. 

The problem doesn't go away but the trade-offs change considerably. This is an interesting and relevant discussion, but if I were trying to code up light MC playouts for the first time, right now, I would be feeling that this dead-simple algorithm was actually very difficult and confusing. 

For someone in that position (and only them), my advice is
1. Implement light playouts first. It's simple; you will find many bugs that way; it's standardized enough that other people will understand what you're talking about; it's a fast way to get a basic bot; it will be a very handy thing to have as a baseline when you test other things.
2. Get it working the standard way before improving it. It's your baseline that you'll be testing improvements against.
3. Make it fast but don't spend excessive effort optimizing. "Better is the enemy of good enough." 

- Dave Hillis

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://computer-go.org/pipermail/computer-go/attachments/20080514/756c184f/attachment.htm