Author: Vincent Diepeveen
Date: 11:09:40 03/27/01
Go up one level in this thread
On March 26, 2001 at 22:36:45, Robert Hyatt wrote:
>On March 26, 2001 at 18:32:53, Ed Schröder wrote:
>
>>On March 26, 2001 at 18:05:01, Bertil Eklund wrote:
>>
>>>On March 26, 2001 at 17:50:30, Thorsten Czub wrote:
>>>
>>>
>>>
>>>>right.
>>>>and the problem is that this all influences the results.
>>>>especially when doing long matches.
>>>>
>>>>you can imagine what i want to say with this.
>>>
>>>Yes, exactly, it must have something to do with the devil and capitalism.
>>>
>>>Bertil
>>
>>
>>I think this is not a good answer.
>>
>>I think that most programs have the "learning" independent of the "save
>>game" command. Rebel for instance does not depend its learning based on
>>the end of a game (save game) but during the game. But in earlier versions
>>the learning WAS dependent on the "save opponent game" flag.
>>
>>So if other programs (Diep)(Crafty?) depend their learning code during
>>the end of the game (save game) Vincent has a point saying the box "save
>>opponents" should be marked as default.
>>
>>Ed
>
>
>I don't need _any_ external trigger to set off my learning code. I did this
>for this very specific case. That I don't want to depend on the other end to
>tell me when something important should be done. I don't want to depend on
>the operator to adjust the book so my program won't play a bad opening. Etc...
>I try to do _everything_ automatically...
This is not the biggest problem Bob. The biggest problem is
the human error which happens if crafty cannot save the games but
a program X can.
Program X MIGHT edit the PGN set just like it wants to.
I'm not saying that it doesn't happen already, but i know i
get sued directly if i replace X by names.
More or less they all do it to some extend. Shredder only seems to
manipulate, of course Stefan has a logical explanation for all this
too which i btw do not accept as fair, the first game only. Just
manipulating, whatever logical cause it has, the outcome of the first
game obviously means that you manipulate very little games.
However if i play 10 games versus crafty, i do not allow crafty to
save the games, then i can represent of course in my display that as
5 wins for X and 5 aborted games.
So that's 5-0 for X then against crafty.
That's what exactly happens.
The small problem is that if you have a logfile, that you can see
the problems AS A PROGRAMMER, as you can checkout the logfile.
As a tester seeing the big shit happening is already quite a bit harder
as 99% of all testers have big problems editting a file. Especially
a non-word file is impossible for them to open.
So you can create zillions of log files, it all doesn't matter.
The only thing what matters is what program X shows on the display.
Some time ago i had a famous auto232 run DIEP-Rebel.
The tester reported that DIEP drew a few, and lost the rest.
Also he reported that rebel had a lot of problems playing games
as many games were aborted.
Aborted games is not uncommon at the auto232 player with certain programs.
It is uncommon in matches versus Tiger and Shredder.
Against Deep Fritz, Fritz6, Deep Junior, Nimzo2000, Nimzo8
and some other chessbase products i get loads of aborted games against.
In the average auto232 run about half of the games are aborted!!!!!!!!!!!!!!!
If a product X gets bad out of book at a certain point, directly aborts
the game.
Is the program playing unfair then? Or is it just another small problem
happening?
Look the real problem here is NOT the auto232 protocol. The auto232 protocol
is ok.
The winboard protocol is impossible to use because you must all run under
the same interface then, which is not going to happen.
the number of testers that can setup winboard, buy a dual, and install
programs is so limited, that i would get out of testers real soon.
But i don't mind if other programs load as winboard into my program of
course.
Because
a) i already know who i play
b) i can do whatever i want to in my interface when playing
this program. So my own interface is completely in control then.
I don't have problems there.
Actually the auto232 protocol is more than ok to play 40 in 2 games
against other programs (for blitz it's a bit harder to say it's ok).
There are just very few commands to implement in an interface for it.
Just link a chrilly/stefan .c and .h file and it already works more or
less.
BUT STILL THERE ARE MANY ABORTED GAMES DESPITE THE EASY PROTOCOL!!!
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.