Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Requested, and done! (better spacing of column titles)

Author: Stephen A. Boak

Date: 22:52:01 04/27/01

Go up one level in this thread


On April 28, 2001 at 01:23:42, Uri Blass wrote:

>On April 28, 2001 at 01:11:16, Stephen A. Boak wrote:
>
>>On April 28, 2001 at 00:58:13, Stephen A. Boak wrote:
>>
>>Uri,
>>
>>Here are the figures you wanted.  Quite interesting!
>>
>>Relative Ratings from Fixed Delta Ply Settings:
>>
>>                PLY
>> T14    DF      DELTA   BOTH    Games   Ave/Ply
>>>79.4	-79.4	0	0.0	600	-
>>>231.5	53.7	1	142.6	500	143
>>>375.8	190.0	2	282.9	400	141
>>>557.3	372.6	3	464.9	300	155
>>>557.7	614.1	4	585.9	200	146
>>>603.9	798.3	5	701.1	100	140
>>
>>The relative rating average, per delta ply, is extremely linear--approx 145 pts
>>ave.
>>
>>Aren't you the mathematician?
>
>Yes but there is no simple fomula to calculate rating correctly and I did not
>know which program to use to calculate it.
>
>Your data suggest that there is no evidence for diminshing return in the nunn2
>match at small depthes.
>
>I believe that there is a diminishing return but we need to wait to games at
>bigger depthes to prove it.
>
>Uri

I used to believe in diminishing returns (based on some thinking & intuitive
notions).  With this remarkably linear data (limited though it is), I am not so
sure.

If chess is as complex as we think it is, every extra ply may provide equal
dividends--up until a program can see far enough ahead that the game is 'solved'
(win, loss or draw).  Since the game is far too exponential for any current
computer to 'solve' it, in the general opening or middle game situation, each
extra ply that plumbs the depths of that complexity may provide approximately
the same return.

This may indicate, for example, that there are relatively as many win, loss &
draw nodes, generally speaking, at each fixed ply depth, no matter how many
plies are calculated--even if those results are not precisely calculatable by
the program!  Therefore the program that calculates x plies more than its
opponent will have approx the same increased chances to steer toward the winning
lines.  [I hope you can understand the concept I am trying to communicate.]

NOTE 1--The number of games you played at large ply deltas is less than the
number of games played at smaller ply deltas.

NOTE 2--the Elo rating scale is based on the following important fact: the win
expectancy (expected win percentage) is the same for any fixed rating delta,
regardless where those ratings are on the overall rating continuum.

For example, a player rated 2000 has a winning expectancy of 64% vs. a 1900
player.  A player rated 2600 also has a winning expectancy of 64% vs. a 2500
player.

In both situations, the higher rated player has a rating advantage of exactly
100 points.

--Steve



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.