Author: KarinsDad
Date: 11:59:10 01/25/99
Go up one level in this thread
On January 25, 1999 at 13:34:48, Robert Hyatt wrote: >On January 25, 1999 at 12:22:25, KarinsDad wrote: > >>On January 24, 1999 at 16:42:03, Robert Hyatt wrote: >> >>[snip] >> >>> >>>>Is it OK to start probing the table at the first ply beyond the root? >>> >>>yes, because you now have a move played at the root.. and you can play that >>>if the EXACT score comes up at ply=2 >> >>This confuses me (probably since I'm still new to this). >> >>This implies that you do not clear out your hash table immediately after the >>program makes a move (i.e. during your opponent's time). For example: >> > > >Correct... I only clear it (and then only the score part) if something changes >so drastically at the root that I alter my piece/square tables. One example is >that after castling, my pawn advance stuff is modified, now that I know where >the king is going to 'live'. It's good to know I wasn't way off base here. > >>D E F G >> \ / \ / >> B C >> \ / >> A >> >>If you make move B, my thought was to remove all nodes C, F, and G assuming no >>transpositions there. If a transposition was found, the count of times found >>would just be decremented. >> >>If you do clear out the hash, then you should not be able to find a >>transposition until ply 4 (i.e. e4 e5 is the same as e3 e6, e4 e5). The >>transposition was not found until a ply 4 search (not a ply 2, that's just the >>first time it was found). >> >>So if you do clear out the hash, then transpositions could not be found until >>ply 4. If you do not clear out the hash, then transpositions could be found at >>ply 1. The strange thing to me here is that if you do not clear out the hash, >>you may find a few transpositions at ply 1 (from 3 ply moves ago), but it would >>seem that this is not worth the effort since you would be saving information in >>your table that is at least 3 ply old (and mostly next to worthless). >> >>I suddenly realized that maybe the phrase "root" meant not the original >>position, but any position passed to the search algorithm. If so, disregard my >>ramblings above. >> > > > >'root' is the initial position... from this position you have to find a move >to play... which is why I don't probe at ply=1, because I need to search each >move to find which is best, not find that they all fail low or whatever... > >But it is a moot point.. probing won't hurt just so you know how to handle the >case where you reach search at ply=1 with alpha and beta set, you do a probe and >get a value < alpha, or > beta before you search anything... Actually, if you remember my posts from a few days ago, probing does hurt me (due to the large size of my nodes and the resulting hash/tree structure that I have decided upon to handle this) for any node that I do not need to probe. However, the entire search must be taken into account. There are a lot more nodes that need to be probed than ones which do not. Not probing before ply 4 will probably hurt me as you said. It's a matter of one if conditional in the search engine for a maximum 15 million positions (per 5 minutes max of searching) versus 7 average probes (of hitting the hash/tree table) searching a mamimum of about one hundred thousand positions (and probably a lot less than that due to the fact that a lot of these positions will be alpha or beta pruned out of the list) for ply 0 through ply 3. KarinsDad :) > > > > >>KarinsDad >> >>[snip]
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.