Author: Uri Blass
Date: 04:11:20 11/22/02
Go up one level in this thread
On November 22, 2002 at 07:00:19, Gian-Carlo Pascutto wrote: >On November 22, 2002 at 06:52:57, Omid David Tabibi wrote: > >>Many things are not that obvious. Please read the "Conclusions" section for >>other algorithms I tried but were inferior to the presented algorithm. >> >>One interesting point is that at depth 8, the size of the tree constructed by >>vrfd R=2 was slightly larger than std R=2; at depth 9, vrfd constructed a >>smaller tree, and the gap widens as we search deeper (see Figure 4). So, I >>believe than on every program, starting from a certain depth, vrfd R=3 will >>construct much smaller trees in comparison to std R=2. And the benefit will >>increase as we search deeper. >> > >You didn't expect this? It's fairly logical. vrfd R=3 is the same as R=3 >below the first fail high. If you search deeper, you get bigger and bigger >parts of the tree that are done with R=3 instead of R=2, so you'll get >smaller trees at some point. > >-- >GCP It may be interesting if you can post results of deep sjeng at 300 seconds per position and not only at 10 seconds per position. The difference for movei is smaller at 300 seconds per position and I did not use research(I suspect that with research reults may be better). My hardware is AMD1000Mhz so you can use smaller time to get eqvivalent results but clearly more than 10 seconds. You can also see that there is a problem in comparing correct results because some problems that are solved by movei with R=3 are not solved for the right reason and movei may change it's mind later(I showed one case when it did it). Uri
This page took 0.01 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.