Author: stuart taylor
Date: 18:30:57 10/25/00
Go up one level in this thread
On October 23, 2000 at 11:41:10, stuart taylor wrote:
> I would like to suggest what I think would be a most intersting and informative
>system for a chess computer tournament.
>PLEASE READ CAREFULLY!
> 1. All the new top programs.
>2.Their previous versions.
>3.Preferably, their previous last TWO versions, in fact, (but at least the
>previous).
>4.All play all, either 8 or 16 games x3 (each for 5 minute, 30 minute and
>tournament timed cycles).
>5. If 8 games x3 each-each program would have one Q-pawn opening, one K-pawn
>opening one books off opening! and one Fisherandom, both as white and black
>which is 8 games.
>6.Or twice that.(the "books-off" can be started off once with d4 and once with
>e4. And the books on, can be "directed" a little to come to games of two
>different natures for each of the two d4 or e4 games.
>7.Whether the 8 or 16 game all play all arrangements are made, it should be 3
>times that much. Once for 5 mins. Once for 30 mins. And once for 40 in 120 Mins.
>games=altogether 24 or 48 games for each program all play all.
>8.All on equal hardware.
>
>If anyone can arrange such a tournament, I think it would be the most
>informative as well as the most entertaining system such a tournament could ever
>be!
> I can hardley even see a rating list to be of more interest than such a
>tournament- so can someone PLEASE do it, or very similar (if you can find any
>improvements).
> With deep analysis, not only of the games themselves, but of the exact results
>and the natures of the results, I think a lot of conclusions or strong
>indications can be seen.
>Any opinions?
>Anyone able/willing to do it?
>S.Taylor
To summarize, and make a bit more clear,
There are 2 main new types of idea.
1.Trying to enter previous 2 versions, will show very clearly how much actual
overal improvements are being made. e.g. when Fritz 7.00 will be entered, even
though it may do well against GT, and Shredder 5 etc.(if), It might not do
better than Fritz 6a. In many cases, we have seen that Fritz 6a was not better
than 5.32, but we must certainly be able to see if Fritz 7.00 will atleast be
vastly better than Fritz 5.32. This is one of many things you can see with such
a tournament, or test.
2.I also suggested 4 types of opening to all be included. 1.KP,2.QP,3.No opening
book,4.Fischerandom.
From all this you can see different types of opening skills. number3.Might
show great new openings and novelties, and number4.can show how well all the
knowledge can be applied in even more testing circumstances. In all cases, don't
forget, all programs are being equally fairly tested, but are being more
critically tested.
In each case, all these 4 types of opening should be done from the White
perspective, AND from the black perspective, which means 8 games between each
two programs.
All 8 games can be done once as game in 5m. Once Game in 30, and once
tournament time. Altogether 24 games between each combination of 2 programs, Or
if twice as much (for a really great test) 48.
I'm also suggesting that you might get clear rating truths without so many
boring hundreds of games. It would take a brilliant deep mathematical mind,
however, to make great insightful conclusions. But this suggested tournament may
provide great opportunity for such insight.
Again, can this ever be done, or even in part?
Very interested, in seeing,
S.Taylor
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.