Re: Computer vs. human matches and Chess256

Den 2007-08-19 23:15:47 skrev Guy Macon <http://www.guymacon.com/>:




M Winther wrote:

Computer vs. human matches is an unfair competition while the computer
has recourse to a huge opening library. Thus, the program doesn't need
to know much about play in the first phase of the game. The opening
problems are already solved, more or less, while the opening books are
that comprehensive. The chess engines, typically, begin thinking only
at the beginning of the middlegame.

It would be a more fair competition if Chess256 was employed in
computer vs. human matches. Neither human nor computer would know any
theory. But it would still be regular chess, and not Fischer Random,
which is much different.

In Chess256 the rules are the same as in Fide chess, and the piece
array is the same. The only difference is the pawn chain. The opening
setup for the pawns is chosen randomly while satisfying the condition
that the pawns must be either placed on the second or third rank.
There are 256 possible configurations. All of them are sound and
balanced, and fully playable. Black's setup mirrors white's.

While Black's setup mirrors white's, this should result in a proper
evaluation of strength. One is free to choose opening strategy,
whether to create an unbalanced or a balanced position, etc. In this
way the program is forced to grapple with problems specific to the
opening, which pertain to development, initiative, and opening
strategy.

http://hem.passagen.se/melki9/randompawn.htm

Good post! Innovative thinking.

I have three comments:

[1] Why did you post this to rec.games.chess.misc and not to
rec.games.chess.computer? that's where computers and chess
should be discussed. Followups set.




Because human-computer matches have had a great impact in the
human chessplayers world, as well. It has affected how we view
chess, and how the public views chess. Moreover, this forum is
much bigger.



[2] I would like to examine the argument that "all 256 possible
configurations are sound, balanced, and fully playable." This
could be tested with a computer; run 256 computer-vs-computer
games with each of the configurations and see whether the black
to white win/loss/draw ratios stay reasonably close. At 10
minutes per game this would take a little over a year for one
computer and about a month for ten computers, with partial
results a lot earlier. It is not a given that all positions
where black mirrors white are fair.


Computer's are horridly bad at evaluating strategical aspects. It
is a very inefficient method. Although the computers will certainly
create good theory, it will exclude many variations that are good
or even better, simply on the grounds that it has no understanding
of the strategical advantages of those lines. For instance, accepting
a position with a bad bishop and get a very solid position instead.

I'd argue that the problem with Chess256 is not unbalanced positions.
The problem is, rather, that the positions are too balanced, i.e., black
is often well prepared already in the opening setup. He has already
initiated a defensive setup on lines of the French (e6), Pirc (d6), etc. So
Chess256 would tend to favour a black defensive setup while some
pawns are already placed on the third rank, whereas white is not as
keen on having his pawns on the third rank, if he is playing for a win,
that is. Thus, games would tend to be equal. However, in practice,
the initial situation is well suited for fighting chess while chessplayers
aren't knowledgeable in all these different defensive setups. They
have to start working immediately, and they always have the choice to
create unbalanced positions.


[3] The proposed solution doesn't solve the described problem.
The computer can simply start with an opening library that is
256 times bigger. It would tend to encourage machine-generated
opening books rather than building the books from master-level
games, which is IMO a Good Thing, but it would not remove the
advantage the computer gets from having a huge opening book.
It would instead increase that advantage.



That would be a huge project, and it would tend to create opening
books understandable only to a silicon brain. Humans, with their
superior strategical understanding would be able to play opening
setups that the computers would not understand how to handle in the
most efficient way. So humans and computers would tend to go
different ways. That's a good development, instead of having
computers beat grandmasters on their own turf, simply because
the computers know the grandmasters' opening repertoire.

Mats

.

Relevant Pages

  • Re: Computer vs. human matches and Chess256
    ... has recourse to a huge opening library. ... Black's setup mirrors white's. ... computer and about a month for ten computers, ... games would tend to be equal. ...
    (rec.games.chess.misc)
  • Re: COMPUTERS MAKE OLD CHESS OBSOLETE??
    ... Computers don't play the opening *all that* badly. ... positions even in Fischer Random chess, ...
    (rec.games.chess.computer)
  • Re: COMPUTERS MAKE OLD CHESS OBSOLETE??
    ... >>>> No doubt Fischer Random chess would level the playing field more, ... >>> Computers don't play the opening *all that* badly. ... Computers play the openings extremely well. ...
    (rec.games.chess.computer)
  • Re: COMPUTERS MAKE OLD CHESS OBSOLETE??
    ... the opening with some positional advantage or another. ... beating the GMs. ... computers do use their openings databasewhen play- ... I am no expert in FIDE chess opening theory. ...
    (rec.games.chess.computer)
  • Re: Recommendation for repetoire book(s)
    ... You say "opening, middle, endgame." ... To me, with weaker players, they've already accepted defeat at chess, so ... >> Back when I played, computers were about 2300 at best, but we knew what ... > phase of the game that is novel, ...
    (rec.games.chess.analysis)