Author: Robert Hyatt
Date: 07:24:30 02/11/02
Go up one level in this thread
On February 11, 2002 at 04:25:46, Sune Fischer wrote: >On February 10, 2002 at 23:33:14, Robert Hyatt wrote: > >>Note that I qualified my answer. "there is no apparent DR for searches that >>we do today, or searches we might do in 2 years." > >Sure, but to me, it sound like you believe this is a case of assuming no-DR >untill proven otherwise. If anything I would probably say the opposite: assume >DR untill proven otherwise. > I try to avoid assuming _anything_ until I see something that suggests it will happen. Obviously DR happens somewhere along the way, and I have no doubt about that. But for practical discussion, will it show up in a way that programs will hit a wall and not play better with deeper searches, within a few years? I believe "no" myself. You only have to watch a lot of games on ICC to see why I believe this. When someone gets a faster machine, they simply win more games against the "stock computers" that have been there for a while. >>I really don't care if there >>is a .01 DR per ply for the next 5 plies. > >It could be much larger than that, in all the rating lists I've seen the >errorbars are +-30 rating points or something like that. Of cause we don't >expect to see a DR of 2-3 points on stats like that. >If there is DR it is most likely not 20 points per ply, it is probably more like >a few points, maybe even fractional points. I agree. And if the gain per ply goes 100 (70 per doubling and doubling doesn't get a ply) 97, 93, 90, 87 to go from 12 to 13 to 14 to 15 to 16, then that is going to keep me going for a good while. At 2 years per ply roughly (two doublings) worrying about DR beyond 10 years is just worrying... > >>I only care that there is a definite >>and pretty constant gain from 12 to 15 plies, which will take me several more >>years to pass. IE thru depth 15, I am _certain_ that the gains will be both >>measurable and significant. That will hold me for a while. I'll let those >>that are 2 years old worry about what happens 100 years out from now. :) > >Even if there is DR, going deeper is still a good thing:) > >-S. I agree... :)
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.