Author: Amir Ban
Date: 06:59:02 11/24/98
Go up one level in this thread
On November 24, 1998 at 08:40:01, Robert Hyatt wrote: >On November 24, 1998 at 03:14:02, Christophe Theron wrote: > >>On November 23, 1998 at 22:38:17, Robert Hyatt wrote: >> >>>If they take (say) 5 minutes to do a 10 ply search, at 250M+ nodes per second, >>>that is over 300X the number of nodes a full-width search to depth=10 should >>>search. If you factor in a q-search that is the same size as the full-width >>>part, we have a missing factor of 150 to account for. I say that is *all* >>>search extensions. And I say that is *far* more than any of the rest of us do >>>in terms of extensions. How *else* would you characterize this? >> >>I don't want to go into a heated discussion, but I notice: >> >>A program that does no forward pruning has a branching factor of (roughly) 5. >> >>A program that uses null move has a branching factor of (roughly) 3. >> >> (5^10) / (3^10) = 165.38 >> >>Weren't you looking for a factor of 150 or so ? >> >>If the IBM team is interested I can provide some help for their null move >>implementation. This way we could have the power of Deeper Blue with only one of >>their chip stuffed into a simple PC. :) >> >> >> Christophe > > >One minor flaw in this discussion. :) Neither Junior *nor* Deep Blue use >null-move. *now* how would you explain that factor of 1,000? > >Bob It seems you agree with Christophe's math. Strange because 10 minutes earlier in another post you said the opposite: Quote: >>1. I'm not an expert on null-move, but I got the impression that it buys much >>more than a factor of 10 at great depths. The factor should be an exponential >>anyway, no ? > >yes, but so is the tree... the result is linear. It (R=2, recursive) is worth >about 2 plies (maybe less) in the opening/middlegame, more in the endgame. >But >not a factor of 10. At least for my implementation. End quote. But this doesn't concern me. In fact I no longer understand what the problem is here. What factor of 1000 about Junior needs to be explained, and why ? To recap: The argument was if the search trees of PC programs and Junior in particular are deeper and thinner in comparison to the Deep Blue search tree. Bob said that the fact that DB reaches 10-11 ply is proof that this is not so. I believe he said that because he knows that null-movers can match this nominal depth in about equal time (but about 1000 lower NPS). Christophe's math shows why, and incidentally confirms the point for null-movers. Junior doesn't match this nominal depth (would be depth 19-21 for Junior), so there is no factor to explain. Junior is more aggressively extended than most PC programs, and therefore more than DB. Amir
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.