Author: Christophe Theron
Date: 23:10:32 10/04/01
Go up one level in this thread
On October 04, 2001 at 22:33:55, Dann Corbit wrote:
>On October 04, 2001 at 21:47:46, Christophe Theron wrote:
>
>>On October 04, 2001 at 17:15:18, Francesco Di Tolla wrote:
>>
>>>Neglecting the proablby important (bu not easy to interpret) effect of the
>>>different CPU models, this is actually perfectly logic:
>>>
>>>Junior is a fast searcher and shredder has more knowledge.
>>>
>>>The faster the hardware that less important the extra search depth!
>>>If you can reach say 20% more knodes this are a given amount of plys at a speed
>>>an less plys at higher speed due to the nonlinear growth.
>>>
>>>So what's wrong?
>>
>>
>>
>>What's wrong is that if search tends to be less effective with increasing speed,
>>actually "knowledge" (probably meaning "evaluation" in your mind) has exactly
>>the same problem.
>>
>>In chess, Search <=> Knowledge
>>
>>I believe in dimishing returns from improved search, and I also believe in
>>dimishing returns from improved knowledge ("improved evaluation").
>
>As everyone knows, chess is O(exp(n)).
>Suppose that by very intelligent pruning techniques, an algorithm [which was
>expensive to compute] reduced the branching factor significantly. We might have
>this situation:
>
>Program 1:run time is k0 + k1 * exp(3*depth)
>Program 2:run time is k2 + k3 * exp(2.5*depth)
>
>Now, maybe k2 and k3 are enormous because of the complicated expressions needed
>to accomplish the pruning. So when time control is short, program 1 will
>clobber program 2. But if hardware gets fast enough, or if the time control is
>long enough, then program 2 will eventually become unbeatable.
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.
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.
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.