Author: Bob Durrett
Date: 15:24:47 12/10/02
Go up one level in this thread
On December 10, 2002 at 17:55:51, Dann Corbit wrote: <snip> >To solve chess you must store at least the square root of nodes of the solution >tree. Considering the half move clock and castle rights, it easily exhausts any >possibility of solution. <snip> I'm not sure what it is you are calling a "solution tree," but some [Uri, etc.] have expressed the expectation that new undiscovered search [& position evaluation] algorithms of the future will greatly decrease the number of nodes that must be examined. The practical limitations for such improvements are unclear. In the limit, the only nodes that would be examined would be in a single branch, assuming all nodes have only one "best move." The number of nodes in such a branch would be very manageable. In cases where two or moves are equally good from a single node, then maybe the pruned tree would be slightly more complex, but still manageable. Maybe that's what you are calling a "solution tree"? Is it possible to establish the absolute limits for pruning and similar "tricks"? Could it then be proven that the resulting tree would be too big to analyze? Perhaps the real problem is uncertainty. There may be some sort of "Uncertainty Principle" at work here. Maybe one can never be absolutely sure that the nodes examined were properly analyzed to find the best move(s). Whether or not a move is the best move may be something that must be expressed as a probability and not an absolute certainty. Bob D.
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.