Author: Christophe Theron
Date: 23:27:24 10/18/99
Go up one level in this thread
On October 18, 1999 at 18:46:46, Heiko Mikala wrote:
>Hi Christoph!
>
>We have heard so many extraordinarily positive results from Chess Tiger lately,
>that I can't wait to have it on my hard disk! :)
Yes! Sorry for the delay, we wanted to provide the best possible engine in the
best interface, and all these things take time. I would not like our customers
to have the kind of problem CM7000 customers have.
>But I have another question to you:
>
>About one year ago, there was a very interesting discussion on CCC, between you,
>Thorsten Czub and some others, about how Richard Langs program Genius works. I
>remember that you had some ideas which you wanted to try out, to try to simulate
>Genius search behaviour. Unfortunately this thread died, when it was most
>interesting.
I think It was in the beginning of 1998 (or late in 1997?). Don Dailey also
posted on this topic, I'm sorry we don't see him here anymore BTW.
>Now, my question is, did these experiments help you to enhance the strength of
>Chess Tiger? Would you say, that Chess Tiger is a program in the tradition of
>published algorithms and techniques, or do you think that you are using unique
>techniques, which are not present in other programs? Do you think, that Genius
>uses unique techniques not found anywhere else? Do you think, you are using some
>techniques which are similar to Genius ones?
This is difficult to say. I guess no programmer here knows really what Lang does
exactly. Or maybe each programmer thinks he knows, but each is in fact thinking
about something different.
So of course I have my own opinion about what Lang does. I would say that most
of what he is doing is also done by most other commercial programs, and by me
too. So 90% of what he does is also done by other good programs.
A little bit of what he does is done by nobody else. This little bit is what
produces the spectacular long lines that you see even when Genius has to move
instantly.
Unfortunately, this is just spectacular, but not really effective. This year,
just to have fun, I have modified Tiger to do the same. I had instantly the
incredibly long lines Genius produces, and I admit that the effect is
spectacular. I could have renamed this version the "Genius Tiger".
Unfortunately, when I did a serious test I discovered that this version was
slightly weaker than the regular (I would say classical) algorithm.
I'm not trying to write a spectacular program, but just a strong one. So I
dropped this.
What is really effective in Lang's program is its selectivity (the program's
ability to prune efficiently irrelevant moves). I think that's why this program
has been at the top for so long. Unfortunately (again) this selectivity is the
best when the ply depth is low. If you play at really fast time controls or on a
slow computer, and reach only 6 or 7 plies deep, that makes a difference. But at
longer time controls or on fast computers, it doesn't give you more than other
known (public) pruning techniques.
>And what about your work with Ed Schröder?
It's really just the beginning. Only a little bit of Rebel has been integrated
in Chess Tiger. This takes time, and we still have a whole treasure of ideas to
try.
This is also true for Rebel. There is a number of ideas in Chess Tiger that
Rebel will integrate little by little.
Christophe
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.