Author: Robert Hyatt
Date: 07:11:30 01/01/01
Go up one level in this thread
On January 01, 2001 at 04:37:39, Steve Maughan wrote: >Heiner, > >>>Yes. This is why when one raises alpha one must also cut the PV. >> >>I do not understand the "cutting the PV" part. Do you mean such a PV ends >>with a "<HT>" TT-hit instead of more moves? > >Yes! This is easy to test. Put a phoney value into the PV (ie -1) and see if this value _ever_ gets backed up to the root in a PV. The answer is "no". You can't back up a PV just by raising alpha or lowering beta. If you can, you have a serious bug. > >>Maybe I should reread the corresponding part in Ernst's excellent book. >>Maybe you should also do that ;-) > >I didn't know he covered it. I'll certainly give it a read. > >>> What is >>>really happening is the deeper search info is being graphted onto the search. >> >>Graphted? Typo? > >Oops I mean 'grafted' (as in when a plant is grafted onto another) > >>I mean: Do you really not use the value from a TT entry if the depth >>stored in the TT entry is larger than the depth demanded by the search? > >At the moment I'm raising alpha based upon the deeper search but not lowering >beta. > >>>I think I'm finally started to get my head rpound this issue - it's being >>>vexing me for days!! >> >>I know that feeling! Do not give up! >>Draw some pictures. Construct a minimal demo case. etc > >Indeed I will. > >>Most of the time finally I find out that things are quite simple, and the >>complex part was a misunderstanding ;-) >> >>Cheers, and a happy new year! (Here around it is already 2:10am) > >Happy New Year to you too!! > >Regards, > >Steve
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.