Author: Uri Blass
Date: 17:31:26 10/13/02
Go up one level in this thread
On October 13, 2002 at 19:03:39, Mike S. wrote: >On October 13, 2002 at 18:24:58, Deepak Nityanandam wrote: > >>During Dr Hsu's chat he said that the DB2 programmers didnt even consider DF >>for benchmarking purposes as they thought it was "way to weak" to play against >>DB2. > >Maybe in 1997. But now as it's clear (or fairly clear...) that Deep Blue's >typical depth of "11(6)" did NOT mean 11 + 6 of nominal depth, but a depth of 11 >and a *maximum* of 6 plies in the hardware for quiesence search (I hope I got >that right, finally), it seems obvious that Fritz has become stronger than Deep >Blue was. And most probably some other programs too, which are of similar >strength like Fritz 7. > >It has been reported that Deep Fritz in Bahrain reaches a depth of 16 or 17 (I >can't remember exactly and where I read it) in the middlegame. > >Also, even on single cpu running at 1,2 GHz, Fritz 7 reaches "base" depths of >15...16 in a few minutes, and selective depths of more than 40! It was said that >Deep Blue II had much more knowledge in the evals (than used today), but can >that compensate for ~4...5 additional plies? In general, I don't think so. > >Regards, >M.Scheidl I also believe that Deep Fritz is better than Deep blue 1997 but the plies are not proof. 1)You cannot compare plies. 12 plies may be more than 17 plies even with the same evaluation. If the 12 plies include a lot of extensions when pruning is used to get the 17 plies then it is possible that the 12 plies include in 90% of the cases all the important lines for 18 plies so they practically help to get the relevant depth(This is theory and in the practical case I guess that the 12 plies of deeper blue were only eqvivalent to 14 or 15 plies of deep Fritz and I talk only about search and not about evaluation) . 2)I believe that a good evaluation can compensate for 4-5 plies if you search deep enough(Not that I believe in the superiority of the evaluation of deep blue of 1997). I believe that the main problem is to find a good evaluation and I do not believe that good players know it. I believe that it is the reason that good players like vincent diepveen fail again and again to win the computer world championship. Uri
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.