Author: Christophe Theron
Date: 08:30:14 11/01/03
Go up one level in this thread
On October 31, 2003 at 18:47:29, Mike S. wrote:
>On October 31, 2003 at 11:53:15, Christophe Theron wrote:
>
>>(...)
>>There's a real world out there, you know.
>>
>>If you don't make any effort, you get slaughtered in the real world.
>
>Really? So what do the users do in the real world, when they use The King in
>Fritz? Do you think they don't provide the King engine with a book? They just
>simply take another book to test and play (often one book for all engines), or
>starting postions from an opening database for egnine matches.
>
>So where's the slaughtering??
>
>When the majority of Chessmaster customers are no specific computerchess fans,
>they don't have the problem either, because they'll use the CM-GUI anyway
>(only). No slaughtering either.
>
>So, what real world were you thinking of? I don't find a realistic example for a
>situation where King is slaughtered because of that.
>
>As I have said, the SSDF setup for that test is *typical for the King's usage
>among computer chess fans*. This is the point. They test in a realistic way,
>because they know how it's used in the real world.
>
>That you have actually acknowldged these unfair things, is a pity and certainly
>nothing to be proud of.
>
>Btw. in the recent past, when Tiger was used with time controls with big bonus
>time, like 10m+10s in the CSS rating list, I often pointed to the fact that this
>may be unfair for Tiger which - as you wrote here once or twice - can't use the
>increment time. (Meanwhile it has turned out, that CB-Tiger indeed *does* use
>some of the increment time it collects, obviously due to the engine adaption to
>the Fritz GUI, but not as good as other engines.) So, my concern was, that this
>could be somewhat unfair for Tiger. Recently I have proposed to use a time
>control like 40/10,40/10 instead, for the next version of the CSS rating list
>after an upcoming hardware change. These ratings get much attention among german
>fans.
>
>In view of your philosophy you express here, maybe I shouldn't care for that
>engine specific issue either. Tiger can't do Fischer time controls? Oops, oops,
>bad luck, so I guess it will get slaughtered in the real world.
>
>So you could be the victim yourself, of your much too strict testing philosophy.
>Did you consider that, too?
>
>In the real world, compromises are necessary to achieve the best possible
>results (like i.e. really good information value of test ratings aimed for
>practise, and not just artificial testlab figures).
>
>Regards,
>Mike
I have been slaughtered myself for these kind of reasons.
Anyway, my point is that the whole mess comes from Johan not providing a book in
a usable format for the SSDF to test.
Using the CB GUI is not a big deal. Actually I don't care about the GUI.
The issues are:
* the book
* the book learning algorithms.
Both give elo points, none have been written by Johan. That's unfair, period.
It is extremely easy to solve. Johan just has to spit his CM9000 book in PGN
format so it can be converted to some CB-useable format. Or maybe the CM8000
book could be used, as I have heard.
Then just turn the book learning off in the CB GUI and let the party begin.
What I don't get is that the lack of cooperation of the engine's author is
rewarded by the right to use EXTERNAL HELP.
It's a clear insult to all the developpers who have cooperated with the SSDF and
had to abide to strict rules like availability of what is tested to the
customers.
It's an insult to all the guys that work really HARD (and it's a tough work) to
build strong opening books. Jeroen, Alexander and Sandro come to my mind at this
time.
It is turning an excellent work (the SSDF list) into a fuzzy list where almost
anything is going to be allowed.
The responsability of the mess is both in the hand of the SSDF guys and in
Johan's.
When you think about it, it would just require one day of work for Johan.
Christophe
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.