Author: Robert Hyatt
Date: 18:57:49 11/21/00
Go up one level in this thread
On November 21, 2000 at 11:33:34, Pat King wrote: >On November 21, 2000 at 09:45:18, Robert Hyatt wrote: > >>On November 21, 2000 at 08:57:07, Bob Durrett wrote: >> >>>On November 20, 2000 at 22:58:10, Robert Hyatt wrote: >>> >>><snip> >>> >>>>> >>>>>Incidentally, the name "Shannon" could be mentioned. Any scheme which throws >>>>>away, repeatedly, large amounts of information is inferior in the sense that it >>>>>is theoretically better to use most or all of the information available, >>>>>assuming that a practical way to do that is found. >>>>> >>>>> >>>>>Bob D. >>>> >>>> >>>> >>>>Unfortunately alpha/beta throws away 99.999% of the information passed over in >>>>a normal tree search. >>> >>>That doesn't sound good. Can anything be done about that? >> >> >>Yes. It is called pure minimax. You get to visit every node in the tree, >>so that you can remember _any_ scores you want. But you will search exactly >>1/2 as deep, which is the killer. > >Why? That 99.999% of information is irrelevant anyway! Isn't the whole point of >AB that it delivers the same result as minimax WITHOUT traversing all that >"information"? Correct. But it does mean that some questions can't be answered. IE which is the second-best move? What is its score?
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.