Author: David Dory
Date: 02:54:34 02/17/03
Go up one level in this thread
On February 16, 2003 at 10:38:01, Albert Silver wrote: >>This is the way Rolf believes the SSDF list should be ordered: >>(From a previous post in this thread) >>" >>it would help a lot if you would take a look into some stats readers. I am not >>inventing opinions, I am talking about simple basics, laymen often forget. That >>is also why it is not possible to adopt a certain routine out of the context of >>a whole method and then do some calculations and whoopie preseting the winner. I >>repeat the actual results - given in the list itself - does NOT allow to >>present Shredder as number one. Also: if SSDF understood what they were doing >>they had done a presentation of the sort 1.-3. are... Then Shredder could well >>be named in the first place. 1.-3. Shredder, Fritz, Fritz >> >>Rolf Tueschen >>" >>Of course, when the number and result of games could statistically separate the >>programs strength reasonably beyond the margin of error, then a #1, etc., could >>be and should be dutifully reported. >> >>Dave > >I had hoped he had something more interesting in mind. All the above says is >that the SSDF shouldn't publish their results until 10,000 games have been >played OR present the results as Tony Hedlund jokingly replied. > >I think it's nonsense. The SSDF is simply an organization that has programs play >hundreds of times against each other and they present the latest results >periodically. They provide all data so that people know how reliable these >comp-comp ratings are: match results, number of games, and the error margin. For >some reason people constantly go gung-ho if they don't like the results. Rolf >claimed they aren't scientific as they aren't absolute, and can change in time. >As far as I know, the SSDF *never* claimed these results were immutable and set >in stone. They are merely the latest results. No more, nor less. As always. > >Take the FIDE ratings for example. Do they never change? Should we attack them >for being non-absolute? There is even the alternate rating list with different >rankings for the top players. What about that? Attack it to? If the data wasn't >deliberately modified and the parameters and methodology are transparent I don't >see what the problem is. > > Albert Have you ever known Rolf NOT to have a problem with the way something is? :) Seriously, his stated problem is the (over) emphasis on marketing, ("we're number 1 on SSDF!), which he feels is used to persuade people to buy new chess programs, unnecessarily. I see what he's talking about - it does somewhat distort the perception of strength in a new chess program, when in fact, said program may NOT be the strongest program after further testing. I think the average consumer is not going to take the margin of error into account, or understand how the ratings may change for a new program on SSDF. I'm sure it's no problem for folks on CCC, especially. Dave
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.