| |
Main
Date: 16 Nov 2006 07:01:57
From: SAT W-7
Subject: How come the same chess engines do not draw ???
|
When two of the same Chess engines play each other , Fritz 9 vs Fritz
9 for example ,
how come they do not draw every game ..
If they are all programmed the same way it seems like every game would
be a draw ..
Or is it that there are so many moves one mite pick a different move to
try and win than draw?
So then my question becomes if you are at the store and are going to
buy Fritz 9 and there are 3 or 4 to buy which one do you buy because
id want to get the one that wins the most because it tells me it is
stronger ...
|
|
| |
Date: 22 Nov 2006 07:30:30
From: [email protected]
Subject: Re: How come the same chess engines do not draw ???
|
Mr. Question wrote:
> <[email protected]> wrote in message
> news:[email protected]...
> >
> > The reason why the same program playing itself 'do not draw every game'
> > is because of the horizon effect in chess. The horizon effect in
>
> Well, I don't think that's the main culprit.
>
> The horizon effect effects both sides.
>
> White may find a good line but black discovers that it has a problem.
>
> But, black may find a good line and white discovers that it has a problem.
>
> So it applies equally to both sides. Assuming both programs are identical
> and their computer systems are identical and so on and so on.
>
> Each side is always seeing one move deeper than the other side.
>
> Assuming chess is a drawn game, then there are 50:50 odds that the horizion
> effect would effect either side in a significant way.
>
> If chess is a won game for one side, then obviously one side would tend to
> be effected a little less often than the other side.
>
>
> Really, the horizon effect is to blame only because a program can't see 50
> or 100 moves ahead and see the end of the game. They can't really see far
> enough to determine the effect their inaccurate positional evaluation will
> have on the game later.
>
>
>
> > computers is when a computer looks ahead and evaluates a move but fails
> > to look that one extra move ahead that would substantially *change*
> > that evaluation, making it a worse move than another choice. From a
> > human perspective, it would be the same as you playing (or not playing)
> > a combination becuase you looked 6-ply ahead and saw that it failed,
> > but in reality, had you looked one more ply you would have *seen* a
> > move that made the combination work.
> >
> > http://chess-training.blogspot.com
> >
>
>
>
> ----== Posted via Newsfeeds.Com - Unlimited-Unrestricted-Secure Usenet News==----
> http://www.newsfeeds.com The #1 Newsgroup Service in the World! 120,000+ Newsgroups
> ----= East and West-Coast Server Farms - Total Privacy via Encryption =----
You prove my point...
White can look x-ply, evaluate the best move, and play it. Black can
look the same x-ply, evaluate the best move, which busts white, and
play it, because black is now 1-ply further in the position that white
was when white evaluated. Now, in this phantom position, black gets an
insurmountable advantage. 0-1.
Horizon Effect.
That is how the horizon effect in chess works. you don't need 50 or 100
moves to see it, and that is impractical anyway. Most 'tournament time
control games' played by engines rarely go beyond 18-ply in the
middlegame. I speak from Fritz8 experience only...
http://chess-training.blogspot.com
|
| | |
Date: 22 Nov 2006 12:40:38
From: Mr. Question
Subject: Re: How come the same chess engines do not draw ???
|
<[email protected] > wrote in message
news:[email protected]...
>
> Mr. Question wrote:
>> <[email protected]> wrote in message
>> news:[email protected]...
>> >
>> > The reason why the same program playing itself 'do not draw every game'
>> > is because of the horizon effect in chess. The horizon effect in
>>
>> Well, I don't think that's the main culprit.
>>
>> The horizon effect effects both sides.
>>
>> White may find a good line but black discovers that it has a problem.
>>
>> But, black may find a good line and white discovers that it has a
>> problem.
>>
>> So it applies equally to both sides. Assuming both programs are
>> identical
>> and their computer systems are identical and so on and so on.
>>
>> Each side is always seeing one move deeper than the other side.
>>
>> Assuming chess is a drawn game, then there are 50:50 odds that the
>> horizion
>> effect would effect either side in a significant way.
>>
>> If chess is a won game for one side, then obviously one side would tend
>> to
>> be effected a little less often than the other side.
>>
>>
>> Really, the horizon effect is to blame only because a program can't see
>> 50
>> or 100 moves ahead and see the end of the game. They can't really see
>> far
>> enough to determine the effect their inaccurate positional evaluation
>> will
>> have on the game later.
>>
>>
>>
>> > computers is when a computer looks ahead and evaluates a move but fails
>> > to look that one extra move ahead that would substantially *change*
>> > that evaluation, making it a worse move than another choice. From a
>> > human perspective, it would be the same as you playing (or not playing)
>> > a combination becuase you looked 6-ply ahead and saw that it failed,
>> > but in reality, had you looked one more ply you would have *seen* a
>> > move that made the combination work.
>> >
>> > http://chess-training.blogspot.com
>> >
>>
>>
>>
>> ----== Posted via Newsfeeds.Com - Unlimited-Unrestricted-Secure Usenet
>> News==----
>> http://www.newsfeeds.com The #1 Newsgroup Service in the World! 120,000+
>> Newsgroups
>> ----= East and West-Coast Server Farms - Total Privacy via Encryption
>> =----
>
> You prove my point...
>
> White can look x-ply, evaluate the best move, and play it. Black can
> look the same x-ply, evaluate the best move, which busts white, and
> play it, because black is now 1-ply further in the position that white
> was when white evaluated. Now, in this phantom position, black gets an
> insurmountable advantage. 0-1.
No, not really.
The search extensions make that a lot more complicated.
The program may be searching 9 full plies, but will follow key lines deeper,
due to search extensions.
White may do 9 full plies and then 7 plies of extensions to reach a quiscent
position.
Black may do 9 full plies and then only need 6 plies of extensions to reach
that same quiscent position.
In which case both programs are reaching the same point. (Although it'll
take more time for white, which will effect its time controlled search
later.)
OR, the key move may be at the 10th full ply. While wont see it, but Black
will.
And that also assumes a perfect evaluator. That the evaluator is st
enough to notice the key positional evaluation. That it wont miss it or
mis-evaluate it.
Then you have all the other factors that have already been mentioned in my
first message. Opening books, bias against drawing, evaluator treats white
a little differently than black (some programs do, because white makes the
first move of the game.) etc. etc.
>
> Horizon Effect.
>
> That is how the horizon effect in chess works. you don't need 50 or 100
> moves to see it, and that is impractical anyway. Most 'tournament time
> control games' played by engines rarely go beyond 18-ply in the
> middlegame. I speak from Fritz8 experience only...
>
> http://chess-training.blogspot.com
>
Not relevant. The 'horizon effect' is an artifical term (for a real event)
that occurs because the program *can't* see to the end of the game.
I just used 50 or 100 moves as an off the top of my head number. That's not
important.
The horizon effect is when an imperfect evaluator tries to evaluate a
position with a perfect score. Basically, it guesses. If it sees something
it doesn't like, it tries to push it over the horizon.
----== Posted via Newsfeeds.Com - Unlimited-Unrestricted-Secure Usenet News==----
http://www.newsfeeds.com The #1 Newsgroup Service in the World! 120,000+ Newsgroups
----= East and West-Coast Server Farms - Total Privacy via Encryption =----
|
| |
Date:
From: Martin Brown
Subject: Re: How come the same chess engines do not draw ???
|
|
| | |
Date: 22 Nov 2006 12:38:41
From: Mr. Question
Subject: Re: How come the same chess engines do not draw ???
|
"tin Brown" <
|
| |
Date: 20 Nov 2006 06:10:25
From: [email protected]
Subject: Re: How come the same chess engines do not draw ???
|
SAT W-7 wrote:
> When two of the same Chess engines play each other , Fritz 9 vs Fritz
> 9 for example ,
> how come they do not draw every game ..
>
> If they are all programmed the same way it seems like every game would
> be a draw ..
>
> Or is it that there are so many moves one mite pick a different move to
> try and win than draw?
>
> So then my question becomes if you are at the store and are going to
> buy Fritz 9 and there are 3 or 4 to buy which one do you buy because
> id want to get the one that wins the most because it tells me it is
> stronger ...
First, there is no difference in the actual Disc you buy between the 3
or 4 you can choose from. Same program.
The reason why the same program playing itself 'do not draw every game'
is because of the horizon effect in chess. The horizon effect in
computers is when a computer looks ahead and evaluates a move but fails
to look that one extra move ahead that would substantially *change*
that evaluation, making it a worse move than another choice. From a
human perspective, it would be the same as you playing (or not playing)
a combination becuase you looked 6-ply ahead and saw that it failed,
but in reality, had you looked one more ply you would have *seen* a
move that made the combination work.
http://chess-training.blogspot.com
|
| | |
Date: 21 Nov 2006 13:59:16
From: Mr. Question
Subject: Re: How come the same chess engines do not draw ???
|
<[email protected] > wrote in message
news:[email protected]...
>
> The reason why the same program playing itself 'do not draw every game'
> is because of the horizon effect in chess. The horizon effect in
Well, I don't think that's the main culprit.
The horizon effect effects both sides.
White may find a good line but black discovers that it has a problem.
But, black may find a good line and white discovers that it has a problem.
So it applies equally to both sides. Assuming both programs are identical
and their computer systems are identical and so on and so on.
Each side is always seeing one move deeper than the other side.
Assuming chess is a drawn game, then there are 50:50 odds that the horizion
effect would effect either side in a significant way.
If chess is a won game for one side, then obviously one side would tend to
be effected a little less often than the other side.
Really, the horizon effect is to blame only because a program can't see 50
or 100 moves ahead and see the end of the game. They can't really see far
enough to determine the effect their inaccurate positional evaluation will
have on the game later.
> computers is when a computer looks ahead and evaluates a move but fails
> to look that one extra move ahead that would substantially *change*
> that evaluation, making it a worse move than another choice. From a
> human perspective, it would be the same as you playing (or not playing)
> a combination becuase you looked 6-ply ahead and saw that it failed,
> but in reality, had you looked one more ply you would have *seen* a
> move that made the combination work.
>
> http://chess-training.blogspot.com
>
----== Posted via Newsfeeds.Com - Unlimited-Unrestricted-Secure Usenet News==----
http://www.newsfeeds.com The #1 Newsgroup Service in the World! 120,000+ Newsgroups
----= East and West-Coast Server Farms - Total Privacy via Encryption =----
|
| | | |
Date: 22 Nov 2006 08:13:55
From: David Richerby
Subject: Re: How come the same chess engines do not draw ???
|
Mr. Question <[email protected] > wrote:
> Assuming chess is a drawn game, then there are 50:50 odds that the
> horizion effect would effect either side in a significant way.
There's no reason to assume that this is true. For example, from the
position after 1.e4, it could be that Black has to play a long
sequence of `only moves' to hold the draw (i.e., if he makes a single
mistake during the first forty moves, say, the draw evaporates and
White can force the win), while White has several winning moves in
each of the positions on the tree. In that case, Black is much more
likely to be hit by the horizon effect than White.
On the other hand, it could be that White has to play the long
sequence of only moves. There's no way of telling which scenario, if
either, is true and I see no reason to assume that the two scenarios
are equally likely.
Dave.
--
David Richerby Strange Watch (TM): it's like
www.chiark.greenend.org.uk/~davidr/ a precision chronometer but it's
totally weird!
|
| |
Date:
From: Martin Brown
Subject: Re: How come the same chess engines do not draw ???
|
|
| | |
Date: 20 Nov 2006 09:09:40
From: David Richerby
Subject: Re: How come the same chess engines do not draw ???
|
tin Brown <
|
| | |
Date: 16 Nov 2006 12:57:45
From: SAT W-7
Subject: Re: How come the same chess engines do not draw ???
|
Thank You for that info.....
|
| |
Date: 16 Nov 2006 11:03:20
From: Mr. Question
Subject: Re: How come the same chess engines do not draw ???
|
"SAT W-7" <[email protected] > wrote in message
news:[email protected]...
> When two of the same Chess engines play each other , Fritz 9 vs Fritz
> 9 for example ,
> how come they do not draw every game ..
>
> If they are all programmed the same way it seems like every game would
> be a draw ..
There are a lot of reasons why two identical programs may not draw.
Lots of random factors and even a few non-random factors.
1) two identical computers aren't actually identical. One may run slightly
faster, which means that system may search just a few nodes more and find
the right move. Not a massive influence, but every bit helps when a program
is in a tough spot. There have been lots of games lost because a computer
didn't have an extra second or two to discover a move was bad.
2) Pure chance. It's possible a program may make a move that looks good,
but the oponent program's search will see that it has a problem and take
advantage of it.
3) Programs often have randomization built in. If two moves have identical
scores, then the program might pick one at random.
4) Programs often have learning. Usually involving the opening book. So
one program may learn a little better than the other version.
5) Opening book differences. The books might not be entirely balanced.
Meaning it might have more better moves for white than for black.
6) The openings in the book may not be...., well, the book may have a
preference for White.
7) Since white goes first, that program will pick the best book move it can
find (with randomization.) That may put the black program at a slight
disadvantage.
8) The program may have a bias against drawing. Better to play for the win
than to draw.
9) in computer vs. computer play with identical programs, it's possible that
white might win more than black. Especially if the program is biased
against drawing.
And there are more. Those are just off the top of my head. And when the
programs aren't identical (but with just ever so slight tweaks among
versions), things become even more complicated.
To put it simply, computer vs. computer play is not really representative of
how well the program plays against humans. Kind of like comparing two
brands of computers when you actually need a photocopier. Not really all
that relevant.
The reality is that any major program you buy today will be significantly
strong on today's hardware.
Heck, even the chess programs for cell phones and PDA's are stronger than
most of the programs of 10 years ago. There are cell phones with more
processing power than the Cray-1 super computer that CrayBlitz used to run
20 years ago. (Not to mention better algorithms than were available 20
years ago.)
The reality is that, with today's hardware, there are lots and lots of very
strong programs. Even crafty, fruit, toga, ProDeo, etc. are quite strong on
modern hardware and are all free. Just plug them into WinBoard or Arena and
you have a decent gui to play with.
>
> Or is it that there are so many moves one mite pick a different move to
> try and win than draw?
>
> So then my question becomes if you are at the store and are going to
> buy Fritz 9 and there are 3 or 4 to buy which one do you buy because
> id want to get the one that wins the most because it tells me it is
> stronger ...
>
----== Posted via Newsfeeds.Com - Unlimited-Unrestricted-Secure Usenet News==----
http://www.newsfeeds.com The #1 Newsgroup Service in the World! 120,000+ Newsgroups
----= East and West-Coast Server Farms - Total Privacy via Encryption =----
|
| |
Date: 16 Nov 2006 16:45:52
From: David Richerby
Subject: Re: How come the same chess engines do not draw ???
|
SAT W-7 <[email protected] > wrote:
> When two of the same Chess engines play each other , Fritz 9 vs
> Fritz 9 for example , how come they do not draw every game ..
Perhaps playing White gives an advantage. Perhaps the losing side
randomly chose a slightly inferior line in the opening book. Perhaps
some random timing issue caused White Fritz to overlook something
Black Fritz saw.
Dave.
--
David Richerby Portable Sadistic.com (TM): it's like
www.chiark.greenend.org.uk/~davidr/ an E-commerce portal but it wants to
hurt you and you can take it anywhere!
|
|
