Junior in Harvard Cup
Duncan McKay
(thanks) I was a bit surprised at the game Wolff-Junior, where Junior
played 4..Ke7 in a fairly typical Kings Indian Attack. Was this another
operator error a la Fritz-Kasparov? If it was they surely could have
corrected it as it was so early in the game.
Duncan
Christopher F. Chabris
There were no operator errors. All of Junior's moves were played as
requested by the program ... oddly enough.
--
======================
Christopher F. Chabris (c...@wjh.harvard.edu)
P.O. Box 382967, Harvard Square Station, Cambridge, MA 02238-2967 USA
Bruce Moreland
to cause this bug (and another one in a later game wherein Junior played Kf2
and got killed), but I can speculate.
I've seen spontaneous Kf2's happen in Ferret. The f-, g-, and h-files are
"kingside" and the d- and e-files are "center". Exceptions can always be
found, but it's typically bad to be in the center, and since f2 is not the
center, there is an inducement to play Kf2 if O-O is not currently legal.
The size of this inducement is the issue, if it gets too large, you get Kf2.
Ke7 is harder to explain, since e7 is still the center. Perhaps Junior
rewarded castling by penalizing the right to castle, and Ke7 gives up the
right to castle.
These programs don't play chess, they sort of simulate playing chess. It
appears to me that unless you want to endure very slow search speeds you may
have to build in little hacks to get the program to do approximately what you
want most of the time, and if you make a mistake your face turns red and your
opponent thinks you are a fool, until the next game, when your stuff works
right and you crush them.
bruce
In article <4ccs8p$7...@decaxp.harvard.edu>, c...@wjh.harvard.edu says...
--
The opinions expressed in this message are my own personal views
and do not reflect the official views of Microsoft Corporation.
Christopher F. Chabris
>Bruce Moreland (brucemo) wrote about Junior's playing 4...Ke7:
>
>: Ke7 is harder to explain, since e7 is still the center. Perhaps Junior
>: rewarded castling by penalizing the right to castle, and Ke7 gives up the
>: right to castle.
>
>I haven't been following this thread, but the explanation
>which was offered during the event, and makes the most sense
>to me, is that 4...Ke7 was a typo in Junior's opening book.
It's true that the opening-book typo explanation was offered at the
event, but subsequent investigation showed that the program was out of
book by then and played 4 ... Ke7 on its own. In a later game (against
Gulko) it did a similar thing.
Mark Glickman
: Ke7 is harder to explain, since e7 is still the center. Perhaps Junior
: rewarded castling by penalizing the right to castle, and Ke7 gives up the
: right to castle.
I haven't been following this thread, but the explanation
which was offered during the event, and makes the most sense
to me, is that 4...Ke7 was a typo in Junior's opening book.
- Mark
Eric Peterson
In a conversation on ICC, the author of "Junior" explained that 4...Ke7
was not a typo in the opening book, but resulted from new king-position
evaluations they recently implemented (and may soon un-implement) :)
__________________________________________________________________________
Eric Peterson Internet Chess Club: telnet CHESS.LM.COM
USCF Life Master Email for flyer: I...@CHESS.LM.COM
etpe...@netcom.com Web page: http://www.hydra.com/icc/
phone: (412) 362-6334 USA
Martin Zentner
[... speculation stuff deleted ...]
>
> These programs don't play chess, they sort of simulate playing chess. It
> appears to me that unless you want to endure very slow search speeds you may
> have to build in little hacks to get the program to do approximately what you
> want most of the time, and if you make a mistake your face turns red and your
> opponent thinks you are a fool, until the next game, when your stuff works
> right and you crush them.
>
> bruce
Hi Bruce,
this is exactly what it is. I couldn't have described it better in my own words.
I especially liked:
- "your face turns red and your opponent thinks you are a fool"
- "you may have to build in little hacks"
- "simulate playing chess"
-Martin
BTW: I was quite surprised to see YOU write one sentence spawning 6 (six!) lines ;-)
Halibut
: In a conversation on ICC, the author of "Junior" explained that 4...Ke7
: was not a typo in the opening book, but resulted from new king-position
: evaluations they recently implemented (and may soon un-implement) :)
Yes, it appears that it should be unimplemented.
: ________________________________________________________________________
: Eric Peterson Internet Chess Club: telnet CHESS.LM.COM
: USCF Life Master Email for flyer: I...@CHESS.LM.COM
: etpe...@netcom.com Web page: http://www.hydra.com/icc/
: phone: (412) 362-6334 USA
:
Halibut FREE Internet Chess (FICS): telnet
USCF Fish 164.58.253.10 5000
hal...@cnct.com
This is a great idea, Eric. Yes, I think all of my messages from
now on will promote FICS like yours promote ICS, in a half page item.
Brian Karen
>
>>In a conversation on ICC, the author of "Junior" explained that
4...Ke7
>>was not a typo in the opening book, but resulted from new
king-position
>>evaluations they recently implemented (and may soon un-implement) :)
>>
>>
>
will have to overcome if they are to create a computer World Chess
Champion. They see a computer is overestimating the value of a
positional idea (king
safety in this case), make the necessary corrections and find that it
is underestimating the value of the positional idea. No matter how hard
they tinker the computer will still make mistakes because there are an
astronom
ical number of exceptions to every chess rule.
>
> So even if computers can analyze 1,000,000,000 positions a second
>they still wont make it to World Champion IMHO. Whats the use of
seeing so far ahead if you misevaluate the resulting position?
Naisortep/Brian Karen
Bruce Moreland
feel free to continue promoting fics in your signature, the more people know
about their choices about where to play chess on the net, the better.
bruce
In article <4coluo$a...@cnct.com>, hal...@cnct.com says...
>
>Eric Peterson (etpe...@netcom.com) wrote:
>
>: In a conversation on ICC, the author of "Junior" explained that 4...Ke7
>: was not a typo in the opening book, but resulted from new king-position
>: evaluations they recently implemented (and may soon un-implement) :)
>
> Yes, it appears that it should be unimplemented.
>
>: ________________________________________________________________________
>: Eric Peterson Internet Chess Club: telnet CHESS.LM.COM
>: USCF Life Master Email for flyer: I...@CHESS.LM.COM
>: etpe...@netcom.com Web page:
http://www.hydra.com/icc/
>: phone: (412) 362-6334 USA
>:
>
>Halibut FREE Internet Chess (FICS): telnet
>USCF Fish 164.58.253.10 5000
>hal...@cnct.com
>
> This is a great idea, Eric. Yes, I think all of my messages from
>now on will promote FICS like yours promote ICS, in a half page item.
--