Author: Sune Fischer
Date: 01:08:31 02/10/02
Go up one level in this thread
On February 09, 2002 at 22:31:20, Robert Hyatt wrote: >>>It is pretty straight thru 15-16 plies... By experiment. Which means at >>>least "for a while" deeper is better. And it _could_ be straight all the >>>way to the depth where we discover the game-theoretic value. Beyond that >>>point searches can't improve things of course... >> >>By experiment we probably can't decide either way. >> >>Say 13-14 gives a 70 point increase, then 14-15 may only give you 69 extra >>points and 15-16 gives you 68 poins. >> >>This is diminishing returns, but 1 elo probably can't be measured by experiment. >> >>The diviation from linearity is probably within the errorbars of the >>experiments, so I think the experiments are inconclusive rather than in support >>of the no DR theory. >> >>-S. > >I simply say there is no evidence to support a diminishing return theory. It >might happen, or it might not. Experimental data from today's programs, >searching to depths they might reach in the next couple of years, suggests that >no such diminishing return occurs for at least the next few plies. Okay, I think you didn't understand what I said. Let me try again: The experiments don't prove there's isn't DR or that there is DR, they show _only_ that if there is DR, then it is less than the errorbars in the experiment. Since you do not know how large a DR effect to look for, how can you say there is no evidence of it? Probably you have not yet designed a test to reveal it, it is like saying you don't believe in atoms because you can't see them through a magnifyingglass :) One can argue that if the effect is so small then it is not practical significant, so "no diminishing returns" is close enough ;) -S.
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.