Author: Dann Corbit
Date: 11:33:46 02/09/99
Go up one level in this thread
On February 09, 1999 at 14:09:34, Marcus Kaestner wrote: >Hello, > >I´m not a fool, I know the testing procedure and I know that it´s good to >play versus H6 because he has played so many games before. Then really, you have answered your own question about why. >But I, and I´m sure very much other people, are interested in seeing actual >games (fast hardware and good programs as Fritz 3, H7, J5 and so on). This is >interesting. Not boring matches versus P90 or Fritz 3 or next time maybe Fritz 1 >or Mephisto I? Iteresting to whom and for what reason is a good question to ask. The games being executed have a high theoretical value for deciding the strength of the program. That is the goal of the SSDF. On the other hand, you want to see two titans slug it out. If both have only 5 games under their belts, you don't care because it is Foreman/Ali you want to watch and not Alfredo Evangelista. But the purpose of your interest is different than the goals of the SSDF. Eventually, those matches will happen. With the CM series, we should be very glad (and perhaps amazed) that they are testing at all. The moves must be entered one by one by hand instead of at night over a serial line while you are sleeping. You can always run your own tournament, or scour the net for one with fisticuffs of the nature you seek. There are some. >By the way, also a match new program against predecessor says nothing even if >the predecessor has 10.000 rated games, because most programs are tuned against >itself. So a 20-3 win for example of H7-H6 says nothing about the overall >strength. As you see, Shredder3 beats the predecessor Shredder 2 clearly and it >must be much stronger. Unfortunately Shredder 2 gains much more points as the >new version against H7!! What does this say to us...? All of these results do have value. The provide statistical evidence we can use together with probability theory to figure out how strong a given program is. One short run of results means very little. But a large number of short experiments will average out. >So I think a match CM6000-H6 on P200 is much more interesting than a match on >P90. And H6 on P200 has enough games for a clear rating. And what does a little >bit of rating oscillation during the testing procedure matter? Depends on what your goals are. Do you want to find out how strong a program is? *That* is what the SSDF does. It is their purpose. >But I´m sure the swedes will test all on P200. Maybe in 5 years... If you want to run CM6000 games, I bet they will accept them from you, if you follow all the terms of being an SSDF member. >I don´t want to hurt anybody, I´m only tired to boring results. Nobody will force you to read them. One man's poison....
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.