Author: Sune Fischer
Date: 00:38:31 04/02/04
Go up one level in this thread
On April 01, 2004 at 19:05:09, Dann Corbit wrote: >>Or, if you want the search to eventually return mate scores, probe EGTBs when >>bitbases say it is won and beta>mate_bound or bitbases says it lost and >>alpha<-mate_bound. >>Perhaps probing directly into EGTBs when window allows it would be faster, >>matter of tuning of course. > >I guess I had not thought about it carefully enough. I imagined using bitbases >to get a won/lost/drawn opinion (at all nodes). But unless you know the exact >value of the leaves, I don't see how you can choose the best move. Most of the time the search window is not in the mate scores. For example if the window is (336,547) and you get a bitbase hit telling you it is won (or lost), then you know the value is at least mate_bound (mate_bound = mate-1000 or something like that which would indicating mate in 500). We can see it is going to fail-high (or fail-low) and whether it is mate in 500 or 21 won't matter. >I imagined something like this: >If the best evaluation is drawn or lost, who cares. Note btw, that draws are scored exact for bitbases too, since draws do not need DTM. >Do whatever move is among >the suggested list. >If the best evaluation is won, then: >Examine the bottom leaves that are won and perk the correct values back up. > >How will we otherwise find the true value? I am afraid I don't understand how >it can work. You don't need true values at most of the nodes, eg. you never know if a FH is a true value only that it is a lower bound. You could look at alpha and beta before you decide if a fast bitbase hit is good enough. If you need exact scores and are not too close to the leaves then try EGTB. Of course you can always try a bitbase probe, it might return draw :) It's also not very likely that the window is so large that you're looking for exact mate scores for both you and your opponent, usually we have either beta<=mate_bound or alpha>=-mate_bound. -S.
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.