Author: Christophe Theron
Date: 13:55:10 10/05/01
Go up one level in this thread
On October 05, 2001 at 02:28:59, Dann Corbit wrote:
>On October 05, 2001 at 02:10:32, Christophe Theron wrote:
>[snip]
>>Yes it's the classical explanation of this theory of "search vs knowledge", but
>>as far as I know you do not need an enormous k2 and k3 to get the best branching
>>factor that can be achieved nowadays.
>>
>>And if you have a huge k2, I can predict that your program is going to be
>>criticized for being a root processor! :)
>>
>>It might happen some day that we find some expensive way of achieving better
>>branching factors, but at this time we do not need to slow down the nps rate
>>(even in a very fast searcher) in order to get state-of-the-art branching
>>factors.
>
>How do you know that Stefan Meyer-Kahlen has not achieved this in Shredder (IOW:
>maybe he _has_ achieved a branching factor better than anyone else by ludicrous
>gobs of knowledge).
It is quite easy to tell that at this time nobody has got this. If somebody had,
you would see a tremendous difference in relative elo between the program
playing blitz and the same program playing at long time controls (relative to a
set of known opponents).
Sarah has recently published a blitz rating list. The ranking were very close to
the rankings of the SSDF list...
>>I would also like to point out that if you find some extremely expensive way of
>>getting a better branching factor, you can keep your nps very high and still get
>>the benefits of your better BF by applying your system on the first plies only
>>and letting the last plies running at full speed. You do not have to have a run
>>time of k2+k3*exp(k4*depth).
>>
>>Further, the formula has nothing to do with my idea of dimishing returns from
>>knowledge (exactly like dimishing returns from search depth), which was what I
>>was trying to stress out.
>
>Actually, I think I did understand you. All chess programs are going to "hit
>the wall" because of the exponential nature of chess. Once you get to (for
>instance) 17 plies in ten minutes, ply 18 is going to be hecka-hard (maybe
>taking days).
>
>But a smaller branching factor (of say 20%) can accomplish something phenomenal
>in that regard. Sort of like what NULL move accomplished. Shave the tree way
>down so you only search the fat branches and suddenly you're zooming. Of
>course, it's *still* an exponenatial search. But I would hate to compete with
>someone who did trim the tree and I achieved no trimming at all. Pure
>alpha-beta will get slaughtered by a null-move alpha beta. If a similar
>reduction is gained through other means [e.g. I believe that Rebel does not use
>null-move but achieves the reduction in other ways] then you will have to be
>millions of times faster to achieve the same depth.
>
>Reducing the branching factor is where it's at. As soon as we reduce it to 1,
>chess will be solved.[1]
>;-)
Yes, you are right.
I'm working in this direction all the time since several years...
In this regard I think I'm state of the art or almost (like the other authors of
the top programs), and at this time I believe that if a better BF can be
achieved, it will not require the programs to slow down significantly.
We will see...
Christophe
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.