Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Computers and Draws by Repetition

121 views
Skip to first unread message

Jeff Anderson

unread,
Jul 5, 1998, 3:00:00 AM7/5/98
to

No GM would let a 1400 player draw against him, but I can almost draw
with Fritz 5 at will now that I have found how vulnerable it is to draws
by repetition.

Here is one such example out of many:

1.e4 e5 2.Ne2?! Nc6 3.Nbc3 Bc5 4.Na4 Bf8 (This move itself is very lame,
but that is besides the point) 5.Nac3 Bc5 6.Na4 Bf8 7.Nac3 1/2-1/2.

Komputer Korner

unread,
Jul 6, 1998, 3:00:00 AM7/6/98
to

You are missing the point about a playing program. There is no reward
in defeating or drawing a program. They are meant as tools to help you
play better in the real world. Play normal chess against the program.
Play it the same way you would play an opponent at your club. This is
the way you will get better, or is your purpose in life to buy a chess
program so as to coax it to get better by learning against you? I had
thought that the purpose was that the human would learn from the
program and not the other way around. Maybe I am sadly mistaken here.
AAAAAAAAAAAAggggggghhhh!!!!!!!!!!!?????????????????

--
--
Komputer Korner
The inkompetent komputer

To send email take the 1 out of my address. My email address is
kor...@netcom.ca but take the 1 out before sending the email.
Jeff Anderson wrote in message <35A058...@foxitnternet.net>...

mclane

unread,
Jul 6, 1998, 3:00:00 AM7/6/98
to
>Jeff Anderson wrote in message <35A058...@foxitnternet.net>...
>>No GM would let a 1400 player draw against him, but I can almost draw
>>with Fritz 5 at will now that I have found how vulnerable it is to draws
>>by repetition.
>>
>>Here is one such example out of many:
>>
>>1.e4 e5 2.Ne2?! Nc6 3.Nbc3 Bc5 4.Na4 Bf8 (This move itself is very lame,
>>but that is besides the point) 5.Nac3 Bc5 6.Na4 Bf8 7.Nac3 1/2-1/2.

Fritz5 draw-problems are system-immanent and must have to do with the
search.
In many games fritz5 has strange draw-scores and lines.

best wishes

mclane


Howard E

unread,
Jul 7, 1998, 3:00:00 AM7/7/98
to

Jeff Anderson wrote in message <35A058...@foxitnternet.net>...
>No GM would let a 1400 player draw against him, but I can almost draw
>with Fritz 5 at will now that I have found how vulnerable it is to draws
>by repetition.
>
>Here is one such example out of many:
>
>1.e4 e5 2.Ne2?! Nc6 3.Nbc3 Bc5 4.Na4 Bf8 (This move itself is very lame,
>but that is besides the point) 5.Nac3 Bc5 6.Na4 Bf8 7.Nac3 1/2-1/2.

This inability of computers to recognize how strong an opponent they are
playing is rather amusing. Imagine a human player using this line in a
tournament against a computer. Since programs are so very strong today
they may as well avoid draws altogether if the eval for the second best move
is not too much lower than the "best" move.

Robert Hyatt

unread,
Jul 7, 1998, 3:00:00 AM7/7/98
to

Howard E <hex...@dlcwest.com> wrote:

: Jeff Anderson wrote in message <35A058...@foxitnternet.net>...


That's why I added the "rating" command to crafty. Playing on a server, this
is "critical information" to have.. so that you can speculate a little and
try to avoid draws against weaker opponents while not doing so against good
players...

There are special cases, as in simple K and P endings, you'd better be
careful and not try too hard to avoid draws as you might lose...


--
Robert Hyatt Computer and Information Sciences
hy...@cis.uab.edu University of Alabama at Birmingham
(205) 934-2213 115A Campbell Hall, UAB Station
(205) 934-5473 FAX Birmingham, AL 35294-1170

Har...@t-online.de

unread,
Jul 7, 1998, 3:00:00 AM7/7/98
to

quoting a mail from kor...@netcom.ca concerning Re: Computers and Draws by Repetition


> You are missing the point about a playing program. There is no reward
> in defeating or drawing a program. They are meant as tools to help you
> play better in the real world. Play normal chess against the program.
> Play it the same way you would play an opponent at your club. This is
> the way you will get better, or is your purpose in life to buy a chess
> program so as to coax it to get better by learning against you? I had
> thought that the purpose was that the human would learn from the
> program and not the other way around. Maybe I am sadly mistaken here.
> AAAAAAAAAAAAggggggghhhh!!!!!!!!!!!?????????????????

> Komputer Korner

Calm down, stupid computer. :-)
Wouldn't you take the chance to draw a 1000 points better player in any
tournament? Certainly there is no gain here except ELO-points but the aims
are different for every user...

> >No GM would let a 1400 player draw against him, but I can almost draw
> >with Fritz 5 at will now that I have found how vulnerable it is to
> draws
> >by repetition.
> >
> >Here is one such example out of many:
> >
> >1.e4 e5 2.Ne2?! Nc6 3.Nbc3 Bc5 4.Na4 Bf8 (This move itself is very
> lame,
> >but that is besides the point) 5.Nac3 Bc5 6.Na4 Bf8 7.Nac3 1/2-1/2.

Harald Faber


Don Getkey

unread,
Jul 8, 1998, 3:00:00 AM7/8/98
to
In article <6ns1ne$8u6$2...@juniper.cis.uab.edu>, Robert Hyatt
<hy...@crafty.cis.uab.edu> writes:

>That's why I added the "rating" command to crafty. Playing on a server, this
>is "critical information" to have.. so that you can speculate a little and
>try to avoid draws against weaker opponents while not doing so against good
>players...

Clever!


yours in chess,
Don

Coon Rapids MN USA

Heinrich Konrad B.-Snoerje

unread,
Jul 8, 1998, 3:00:00 AM7/8/98
to
Jeff Anderson wrote:
>
> No GM would let a 1400 player draw against him, but I can almost draw
> with Fritz 5 at will now that I have found how vulnerable it is to draws
> by repetition.
>
> Here is one such example out of many:
>
> 1.e4 e5 2.Ne2?! Nc6 3.Nbc3 Bc5 4.Na4 Bf8 (This move itself is very lame,
> but that is besides the point) 5.Nac3 Bc5 6.Na4 Bf8 7.Nac3 1/2-1/2.
What a stupid game ! What colour was fritz ?
I thought Fritz is one of the top programms ? You probably have run
it on the lowest playing thrength (no hash, no opening, one ply, easy
mode,.....), haven't you ?


--
PLEASE:
Don't send anwsers by email AND newsgroup, it cuts threads and it makes
me very angry. If you want to answer by email AND by newsgroup make a
little note like this: 'Posted and emailed' !!! Thank you.

Jeff Anderson

unread,
Jul 8, 1998, 3:00:00 AM7/8/98
to
> > Here is one such example out of many:
> >
> > 1.e4 e5 2.Ne2?! Nc6 3.Nbc3 Bc5 4.Na4 Bf8 (This move itself is very lame,
> > but that is besides the point) 5.Nac3 Bc5 6.Na4 Bf8 7.Nac3 1/2-1/2.

> What a stupid game ! What colour was fritz ?
> I thought Fritz is one of the top programms ? You probably have run
> it on the lowest playing thrength (no hash, no opening, one ply, easy
> mode,.....), haven't you ?

Fritz 5 had the black pieces, Fritz made all those moves with it's book
(it was out of book after 2.Ne2). It was playing on it's highest
level. The time control was G/30 on a Pentium 133.

bruce moreland

unread,
Jul 9, 1998, 3:00:00 AM7/9/98
to
On Tue, 07 Jul 1998 02:07:51 GMT, "Howard E" <hex...@dlcwest.com>
wrote:

> This inability of computers to recognize how strong an opponent they are
>playing is rather amusing. Imagine a human player using this line in a
>tournament against a computer. Since programs are so very strong today
>they may as well avoid draws altogether if the eval for the second best move
>is not too much lower than the "best" move.

This is a serious problem in server play. I've had mine take rep
draws against weak players because it thought it was worse, and I've
had it take reps in positions where it was a whole lot worse, but the
opponent only had a second or two in a zero-increment game, and I'll
even call the draw for them.

There are some players out there who play for both of these
situations, knowing exactly what they are doing.

Just another way humans can adapt to play computers.

bruce


bruce moreland

unread,
Jul 9, 1998, 3:00:00 AM7/9/98
to

I did the same thing to Power Chess, in a Marshall Gambit, with myself
as white. Black came out of book down a pawn, so it let me repeat
right off.

Hard to prevent against an opponent who knows about this problem.

bruce


Quenton Fyfe

unread,
Jul 9, 1998, 3:00:00 AM7/9/98
to
I wonder if Deep Blue suffers from this problem.....

I wonder if Garry Kasparov knew about this tactic.....

Regards

Quenton Fyfe

bruce moreland wrote in message <35a552b0...@news.seanet.com>...

Robert Hyatt

unread,
Jul 9, 1998, 3:00:00 AM7/9/98
to
Quenton Fyfe <que...@excelsiordirect.com> wrote:
: I wonder if Deep Blue suffers from this problem.....

: I wonder if Garry Kasparov knew about this tactic.....

: Regards

: Quenton Fyfe


Remember that it is solvable. If you look at the source to crafty,
file=draw.c, you will find the function DrawScore(). This would normally
return 0.00 unless you have the "contempt factor" set for some reason.

But in the case of Crafty, this is more complex. In the opening, it will
generally return something close to -1.00 for draws. In the middlegame,
it will return a -.5, and in endgames, 0.00 as expected. But all of these
are modified by the clock time remaining for the opponent, so that when the
opponent is low on time, it will accept draws only when it thinks it is
really quite a ways down in score...

So, there is really nothing that says a program *must* suffer from the
"early draw syndrome". It is just that Fritz hasn't been "brought up
on a server like ICC" so that Frans would see this happen *many* times.
Because it would be fixed, then.

Har...@t-online.de

unread,
Jul 9, 1998, 3:00:00 AM7/9/98
to

quoting a mail from sno...@gmx.net concerning Re: Computers and Draws by Repetition

> > No GM would let a 1400 player draw against him, but I can almost draw
> > with Fritz 5 at will now that I have found how vulnerable it is to draws
> > by repetition.
> >

> > Here is one such example out of many:
> >
> > 1.e4 e5 2.Ne2?! Nc6 3.Nbc3 Bc5 4.Na4 Bf8 (This move itself is very lame,
> > but that is besides the point) 5.Nac3 Bc5 6.Na4 Bf8 7.Nac3 1/2-1/2.
> What a stupid game ! What colour was fritz ?
> I thought Fritz is one of the top programms ? You probably have run
> it on the lowest playing thrength (no hash, no opening, one ply, easy
> mode,.....), haven't you ?

I checked this on K6-200 44MB hash 3min/move.


Harald Faber


mclane

unread,
Jul 10, 1998, 3:00:00 AM7/10/98
to
"Heinrich Konrad B.-Snoerje" <sno...@gmx.net> wrote:
>What a stupid game ! What colour was fritz ?
>I thought Fritz is one of the top programms ? You probably have run
>it on the lowest playing thrength (no hash, no opening, one ply, easy
>mode,.....), haven't you ?


The draw-trick works not only with fritz. Genius is also very
problematic and will have problems if you repeat the position for no
reason. Suddenly, you can force a draw.

best wishes

mclane


0 new messages