Author: Christophe Theron
Date: 11:32:14 07/17/98
Go up one level in this thread
On July 16, 1998 at 22:17:47, Don Dailey wrote:
>On July 16, 1998 at 13:06:46, Christophe Theron wrote:
>
(snip)
>>What are we going to evaluate in this championship? The programmer's skill to
>>write a good chess program, or the programmer's skill to find a good sponsor?
>
>
>That's a good question. But I don't think the answer is what you
>think it should be. I think the sponsers and the ICCA is interested
>in awarding the title to the strongest machine/software combination
>the entrants can muster. It is NOT a programming contest as you
>believe it should be.
>
>As evidence of this, the WCCC has always produced some finely engineered
>chess playing HARDWARE devices. If the event was viewed as a
>programming contest SEVERAL of our computer chess world champions
>would have been disqualified at the door! But these devices have
>not always won either! Micro's are strong enough to win this
>event, the current world champion is Fritz! Would you deprive
>Fritz or any other micro the opportunity to beat a program like
>Deep Blue and win the championship? Maybe yours will win the
>next one!
>
>I know how you feel however. Every since I have been in these
>tournaments I have been among the weakest in terms of hardware.
>It's disheartening to put all the work into it and then face
>a program like HiTech or Deep Thought or Cray Blitz. But
>simply excluding these programs is not a very good answer.
>Recently I have been much luckier with hardware and have had
>support from MIT. And yet still, I am rarely running with
>the fastest hardware at the event.
>
>I like your idea a lot, as a separate event called the "World
>computer chess programmers contest" as you suggest. However
>I don't think it should REPLACE the WMCCC. To make YOUR
>suggested event fair (to test the algorithms and quality of
>the ideas) you would probably not want to allow assembly
>language programs as that tends to be grossly unfair to
>the other platforms being developed on. I really think
>you would have to specify a single machine to do this
>correctly and require the programs to be ANSI C. The
>platform should probably be Alpha's if they are available
>as they are the hottest machines for chess right now. And
>it wouldn't be fair for someone who is developing on INTEL
>for instance to face a competitor who has his own ALPHA at
>home and writes assembly code for it.
>Also the machines should be supplied and be exactly the same
>so there is no issue of who can afford to bring what.
>
>But you can see there are a lot of logistical problems here.
That's why I didn't want to enforce a strict uniform platform rule. I am
suggesting that we decide that a huge hardware advantage is not allowed in an
event called "WMCPC".
I don't like my opponent to have a 25% speed advantage, but I know I have still
good chances if I worked hard enough on my program.
But with a 2 times speed advantage, my opponent is already going to crush me,
even with an average program. I have already done the experiment of giving a top
program (say Genius or Rebel) half the time I gave to my program. God! I thought
I had a monster, because I won more than 65% of the games. More, the opponent
looked ridiculous. But this was pure bullshit.
Stricly uniform platform is too hard to organize. But it is not hard to get
something near uniform platform if we accept, say, a 25% range tolerance.
Let's take an upper clock speed limit, say 300MHz for Intel family and a 400MHz
limit for Alphas, so it's not too hard to find a computer that fits.
About assembly langage: I don't care. It's part of the the programmer's skills,
it has nothing to do with the computer's skills. So it's OK with me. My program
is C-only, and I don't think I would gain much by rewritting parts in assembly.
Maybe we could organize a vote to know:
1) What others think of a WMCPC nearly-uniform platform event
2) What are the limits we should set on clock speeds for the various processors
(x86, Alpha, PowerPC...)
Maybe we could also set a limit on hash tables.
Such an event would not require much money, because most of the competitors
could come with their own computer. If MIT provides the place, we are done...
My guess is that many programmers are dreaming of such a tournament.
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.