Author: Dann Corbit
Date: 08:00:18 02/07/02
Go up one level in this thread
On February 07, 2002 at 10:36:51, Sune Fischer wrote: >On February 07, 2002 at 10:20:41, Dann Corbit wrote: > >>On February 07, 2002 at 10:15:59, Sune Fischer wrote: >>[snip] >>>>The previous studies also showed that. The real question is whether 16-18 plies >>>>has less benefit compared to 18-20 plies. The reason I say that would be >>>>obvious if you look at the graphs of the studies. >>> >>>I think any effect out there will be hard to measure, the curve is just too >>>straight. >>>Compare 2-3 to 15-16 and I'm sure you will see the effect clearly. >> >>I don't think that has ever been in question. The notion of diminishing returns >>is only interesting if they continue to diminish. So (for instance) if after 20 >>plys, you always get exactly 10% improvements, then the returns are not >>diminishing from that point forward. >> >>The only unanswered question is "do the returns of an additional ply continue to >>become less and less?" >> >>Every study I have ever seen shows that the initial plies are move valuable. In >>fact, it's nothing but common sense. > >Well then you can almost prove it by induction. > >I think you will be hard pressed to measure any thing that deep. >If 15-16 plies is 45-55% lose-win, then 16-17 plies may result in 45.3-54.7% and >17-18 in 45.5-54.5% etc. converging to 50-50% at plies 338-339 :) > >Those are very small effects that you are _expected_ to find, they will simply >not be visible because of the uncertainties, how do you expect to find 0.2% >difference? > >You need to zoom in on this effect by comparing 6-7 plies to 16-17 plies. That comparison everyone already knows the answer to, so it isn't an interesting question. The early plies are more important than the later ones. The only unsolved question is "Are distant plies less and less valuable or does it level out at some point." Because chess is exponential, in 10 years we will see 5-10 plies deeper (if a branching factor of 2 could be achieved it would be 10).
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.