Chess computers without opening book

Main

Date: 05 Dec 2006 14:21:09
From: Timid Demagogue
Subject: Chess computers without opening book

I had a discussion today about chess computers and their reliance on
opening books. I asserted that a human GM playing Deep Fritz 10
without its opening book would win most of the time.

Would anyone care to agree or disagree? I'd appreciate any elaboration
as to why, either way.

-td

Date:
From: Martin Brown
Subject: Re: Chess computers without opening book

Date: 12 Dec 2006 17:59:11
From: johnny T
Subject: Re: Chess computers without opening book

Sigh...

I was responding to the assertion in this thread about the thought
problem and the possibility of complete opening books. Please go back
and check that.

You keep arguing with me, not about this at all, but on whether there is
any value in opening books.

I agree opening books have value. I actually think they are
interesting in many ways, including the fact that there is essentially
2-3 guys that are so good, that they are the go-to guys, and really very
few others even have a chance in competing. I agree there are many good
reasons to make them, up to and including preparing for an opponents
strengths and weaknesses. I never said I didn't.

However, your quote.
__________
For any fixed depth the opening book could be made complete. But the
storage and computational requirements rapidly become impossible due to
the exponential growth of the game tree.
__________

This is a problematic statement. So much, I wish you would keep from
saying these things, because, whatever your point is, this confuses it.

Date: 12 Dec 2006 00:34:32
From: MaximRecoil
Subject: Re: Chess computers without opening book

Kenneth Sloan wrote:
> MaximRecoil wrote:
> > Kenneth Sloan wrote:
> >> Humans use this sort of knowledge routinely. There is no good reason to
> >> bar computers from using it - the *worst* reason is that computers do it
> >> better.
> >
> > What you mean to say is this:
> >
> > Humans use this sort of knowledge [of the apples variety] routinely.
> > There is no good reason to bar computers from using [this sort of
> > knowledge of the oranges variety].
>
> No, I don't.

Why did you reply if you weren't going to address the content of my
post?

"No I don't."?

You: "Two plus two equals five."
Me: "What you mean to say is, 'Two plus two equals four.'"
You: "No, I don't."

LOL

Date: 11 Dec 2006 14:41:27
From: Guy Macon
Subject: Re: Chess computers without opening book

tin Brown wrote:

>Table bases are complete for all the cases that they cover. But the
>storage volume goes up alarmingly with the number of pieces left on the
>board.

The figures I got from Bo Hyatt a while back are:

Complete 3 man Nalimov Tablebase...80 kB (measured)
Complete 3+4 man Nalimov Tablebase...30 MB (measured)
Complete 3+4+5 man Nalimov Tablebase...7.5 GB (measured)
Complete 3+4+5+6 man Nalimov Tablebase...1-2 TB (estimated)
Complete 3+4+5+6+7 man Nalimov Tablebase...200-600 TB (estimated)
Complete 3+4+5+6+7+8 man Nalimov Tablebase...40-180 PB (estimated)
Complete 3+4+5+6+7+8+9 man Nalimov Tablebase...????
Complete 3+4+5+6+7+8+9+10 man Nalimov Tablebase...????????

Opinions on these estimates would be most welcome, of course.

Date: 11 Dec 2006 19:11:47
From: Tony M
Subject: Re: Chess computers without opening book

On Mon, 11 Dec 2006 14:41:27 +0000, Guy Macon
<http://www.guymacon.com/ > wrote:

>
>
>
>tin Brown wrote:
>
>>Table bases are complete for all the cases that they cover. But the
>>storage volume goes up alarmingly with the number of pieces left on the
>>board.
>
>The figures I got from Bo Hyatt a while back are:
>
>Complete 3 man Nalimov Tablebase...80 kB (measured)
>Complete 3+4 man Nalimov Tablebase...30 MB (measured)
>Complete 3+4+5 man Nalimov Tablebase...7.5 GB (measured)
>Complete 3+4+5+6 man Nalimov Tablebase...1-2 TB (estimated)
>Complete 3+4+5+6+7 man Nalimov Tablebase...200-600 TB (estimated)
>Complete 3+4+5+6+7+8 man Nalimov Tablebase...40-180 PB (estimated)
>Complete 3+4+5+6+7+8+9 man Nalimov Tablebase...????
>Complete 3+4+5+6+7+8+9+10 man Nalimov Tablebase...????????
>
>Opinions on these estimates would be most welcome, of course.
>

The entire 6 man set, except for 5-1 sets, has now been generated.
They take up 1,146 GB.

http://kd.lab.nig.ac.jp/chess/tablebases-online/

There is a discussion forum specifically about tablebases that you may
find useful.

http://kd.lab.nig.ac.jp/chess/discussion-board/viewforum.php?f=6

Tony

Date:
From: Martin Brown
Subject: Re: Chess computers without opening book

Date: 11 Dec 2006 12:04:54
From: johnny T
Subject: Re: Chess computers without opening book

Ok, you have come into the middle and really not addressed the argument
at hand, using the examples at hand... Which makes for a very messy
thread.

But in that end, we shall sally forth! Onwards USENET NEWS!

tin Brown wrote:

> The completeness issue is very real. But it is possible to have an
> adequate opening book for most practical purposes given a sufficient
> amount of storage and some well chosen heuristics.

Adequate is not a very good scientific term. Unlike complete which is
not really lacking any ambiguity, adequate is fully in the eyes of the
beholder.

And at any rate, today, after the largest expansion of chess knowledge
in history (the last 25 years compared to the entire previous epoch)
chess openings/books have undergone constant revision. From the edges
to the core. What was once crap, became vogue, and suddenly crap again.
And every other measurement conceivable.

By reasonable measure, it is not adequate. By reasonable guess, there
is no definition of when it will become adequate. What is known, is
that there is a HUGE (impossibly large) physical divide between complete
and today, and somewhere in the middle, may or may not be adequate.

>> Physics will prevent any meaningful and complete opening book. There is
>> absolutely no point to argue that it will be, except for the sake of
>> argument.
>
> The point of an opening book is to guide the chess engine through deep
> waters where classic lookahead at tournament time controls is
> insufficient to fully see into the long term consequences of a
> particular move. There is circumstantial evidence that at the moment
> the top opening book authors are refining books to include trap lines
> that tempt other engines into playing an obvious strong move visible to
> the evaluation function running under normal tournament time controls
> but which is refuted by some deeper book variation.

First, you are totally missing the freaking point here. Go back and
reread to catch up.

Secondly, you are missing the freaking point.

> This means that in an engine vs engine match the engine with the best
> book takes the other out of book with an apparently weaker move
> inviting an obvious refutation that it knows from its own prepared
> deeper analysis does not work.

So what? This has nothing to do with the previous question, which had
to do with "solved chess" and comparing that to solving chess on the
fly. The tactical uses of opening books were not under question.

And again "best" here is in the eye of the winner. You have found a
single case, and determined that one was "best". Which I suppose is one
means of measuring, but I am sure that there are others.
>
> NB these types of trap variation and refutation are well within the
> capabilities of the engines to precompute as a part of pre-match
> preparation.

And somehow, you use this example with "adequate". And I would use
this example to destroy the term "adequate".

>> The table bases tend to be complete, in that they follow all paths to
>> the end of the game for given situations.
>
> Table bases are complete for all the cases that they cover. But the
> storage volume goes up alarmingly with the number of pieces left on the
> board.

Well, um... Duh. However, they are complete, in their purview, by
definition. Which is the priy difference between them and opening
books.

> The risk of dragons will be ever present, but the way the current
> engines play at full strength obtaining a true 50cp advantage out of
> the opening book would probably be good enough.

Wow... "a true 50cp advantage"

You know, philosophically, that has nearly no meaning whatsoever. It is
a measure of the existing state of the game by a given observer's
evaluation. There is nearly no agreement on what objectively 50 cp
means. Not on depth, quality or much of anything. It is only an
internal ruler which an engine can help determine which move is better
than another. As a matter of fact, that whole 50cp thing is precisely
the difference between one engine and another.

And even with the amazing blunder, you have still missed the point.
Namely, you cannot make a constraint on the game in real-life, based on
the thought problem of a complete book, because physics prevents that
from existing. For similar reasons, you can discount time travel when
making constraints between humans and computers.

Date: 10 Dec 2006 22:08:21
From: MaximRecoil
Subject: Re: Chess computers without opening book

Kenneth Sloan wrote:
> Humans use this sort of knowledge routinely. There is no good reason to
> bar computers from using it - the *worst* reason is that computers do it
> better.

What you mean to say is this:

Humans use this sort of knowledge [of the apples variety] routinely.
There is no good reason to bar computers from using [this sort of
knowledge of the oranges variety].

How do you come to such a conclusion anyway? The "apples variety" of
this sort of knowledge (human memory) is both different and inferior in
many ways to the "oranges variety" of this sort of knowledge (tangible
reference material).

Now with humans vs. computers, we have a case of only one side being
able to utilize the "apples", and both sides being able to utilize the
"oranges"; and your proposition in the name of fairness (or whatever)
is "let one side use apples and let the other side use oranges"? Why
wouldn't, "let both sides use oranges" be the more logical choice?

Or, you can deny them both oranges because of the longstanding
prohibition against them, and say "tough luck" to the computer
builders/programmers for not figuring out a way to implement apples
into their machines, because humans already have apples intrinsically
and apples have always been a part of legal chess.

Date: 11 Dec 2006 22:34:58
From: Kenneth Sloan
Subject: Re: Chess computers without opening book

MaximRecoil wrote:
> Kenneth Sloan wrote:
>> Humans use this sort of knowledge routinely. There is no good reason to
>> bar computers from using it - the *worst* reason is that computers do it
>> better.
>
> What you mean to say is this:
>
> Humans use this sort of knowledge [of the apples variety] routinely.
> There is no good reason to bar computers from using [this sort of
> knowledge of the oranges variety].

No, I don't.

--
Kenneth Sloan [email protected]
Computer and Information Sciences +1-205-932-2213
University of Alabama at Birmingham FAX +1-205-934-5473
Birmingham, AL 35294-1170 http://www.cis.uab.edu/sloan/

Date: 07 Dec 2006 16:18:53
From: Ange1o DePa1ma
Subject: Re: Chess computers without opening book

"Timid Demagogue" <[email protected] > wrote in message
news:[email protected]...
>I had a discussion today about chess computers and their reliance on
> opening books. I asserted that a human GM playing Deep Fritz 10
> without its opening book would win most of the time.
>
> Would anyone care to agree or disagree? I'd appreciate any elaboration
> as to why, either way.

Anyone who subscribes to Chess Today or any other service that provides 100s
or 1000s of GM games per month will notice that a good many second-tier (and
below) GM games are out of anything remotely recognizable as "book" after
several moves. Others are reviving very old, discredited lines. In most
cases it appears their goal is to get into a reasonably even middle game,
which most of the time they do. I don't see how a computer that did the same
would be so weakened that it couldn't trounce anyone rated 2600 and below.
In fact, being out of book might be an advantage. Remember Miles' a3 against
Karpov.

The computer would get to a playable middle game. It wouldn't blunder pieces
or pawns out of the opening. It would probably not win in 38 moves, it might
take 68 moves, but it would win.

Date: 08 Dec 2006 12:46:07
From: David Richerby
Subject: Re: Chess computers without opening book

Ange1o DePa1ma <[email protected] > wrote:
> Anyone who subscribes to Chess Today or any other service that
> provides 100s or 1000s of GM games per month will notice that a good
> many second-tier (and below) GM games are out of anything remotely
> recognizable as "book" after several moves. Others are reviving very
> old, discredited lines. In most cases it appears their goal is to
> get into a reasonably even middle game, which most of the time they
> do. I don't see how a computer that did the same would be so
> weakened that it couldn't trounce anyone rated 2600 and below. [...]

An excellent point. Thank you.

Dave.

--
David Richerby Old-Fashioned Cat (TM): it's like a
www.chiark.greenend.org.uk/~davidr/ cuddly pet but it's perfect for your
grandparents!

Date: 07 Dec 2006 05:43:19
From: Sanny
Subject: GetClub has 4 moves Opening.

GetClub plays opening for first 4 moves only. It gives it lot of
advantage oven Human Opponents as it makes quick and correct moves.

GetClub Game at: http://www.GetClub.com/Chess.html

Without Opening also the GetClub Game will play nice games. But I
prefer it to have openings.

Bye
Sanny

Play Chess at: http://www.GetClub.com/Chess.html

Date: 07 Dec 2006 12:56:17
From: Guy Macon
Subject: Re: Chess computers without opening book

[email protected] wrote:

>I think *eventually* we will have to have the comp play without an
>opening book to give the GM's a shopt at beating it. After all, the
>opening books that the comps use allow the comps to avoid having to
>calculate the most complicated and difficult phase of the game - the
>opening. This would truly be the Man vs. Machine match we want to see
>with no 'pre-programmed' moves for the computer, i.e., the opening
>book.

actually, the endgame tablebase is what allows the computer
to avoid having to calculate the most complicated and difficult
(for a computer) phase of the game.

If you are to cripple the computer, where will it end? A limit
on the computer's CPU clock speed? Giving the human 100X more
time on the chess clock?

Guy Macon
<http://www.guymacon.com >

Date: 07 Dec 2006 14:00:33
From: Thomas T. Veldhouse
Subject: Re: Chess computers without opening book

Guy Macon <http://www.guymacon.com > wrote:
>
> If you are to cripple the computer, where will it end? A limit
> on the computer's CPU clock speed? Giving the human 100X more
> time on the chess clock?
>

What makes the end game the most difficult portion of the game to calculate?
It seems to me it should be one of the easiest. In fact, keeping that fact in
mind, it seems tablebases are a way to simply speed things up because they are
easy to pre-compute.

--
Thomas T. Veldhouse
Key Fingerprint: D281 77A5 63EE 82C5 5E68 00E4 7868 0ADC 4EFB 39F0

Date: 07 Dec 2006 15:01:17
From: David Richerby
Subject: Re: Chess computers without opening book

Thomas T. Veldhouse <[email protected] > wrote:
> Guy Macon <http://www.guymacon.com> wrote:
>> If you are to cripple the computer, where will it end? A limit
>> on the computer's CPU clock speed? Giving the human 100X more
>> time on the chess clock?
>
> What makes the end game the most difficult portion of the game to
> calculate?

The fact that the endgame is about forming long-term plans.

> It seems to me it should be one of the easiest. In fact, keeping
> that fact in mind, it seems tablebases are a way to simply speed
> things up because they are easy to pre-compute.

You contradict yourself. If tablebases were easy to pre-compute, they
couldn't provide a significant speed-up. As it is, tablebases are
conceptually easy to pre-compute (in the same way that long
multiplication is conceptually easy) but fulfilling that concept takes
a certain amount of effort (what is 3423746829367346 x 23926384767343,
by the way?).

Dave.

--
David Richerby Miniature Postman (TM): it's like a
www.chiark.greenend.org.uk/~davidr/ man who delivers the mail but you can
hold in it your hand!

Date: 07 Dec 2006 16:45:54
From: Thomas T. Veldhouse
Subject: Re: Chess computers without opening book

David Richerby <[email protected] > wrote:
> Thomas T. Veldhouse <[email protected]> wrote:
>> Guy Macon <http://www.guymacon.com> wrote:
>>> If you are to cripple the computer, where will it end? A limit
>>> on the computer's CPU clock speed? Giving the human 100X more
>>> time on the chess clock?
>>
>> What makes the end game the most difficult portion of the game to
>> calculate?
>
> The fact that the endgame is about forming long-term plans.

For a human it is about this.

>
>> It seems to me it should be one of the easiest. In fact, keeping
>> that fact in mind, it seems tablebases are a way to simply speed
>> things up because they are easy to pre-compute.
>
> You contradict yourself. If tablebases were easy to pre-compute, they
> couldn't provide a significant speed-up. As it is, tablebases are
> conceptually easy to pre-compute (in the same way that long
> multiplication is conceptually easy) but fulfilling that concept takes
> a certain amount of effort (what is 3423746829367346 x 23926384767343,
> by the way?).
>

It is easy to recompute. Anybody can write a program [with enough skill] and
they will always get the same results. There is no contradiction. By the
way, it doesn't appear you meant "easy to pre-compute" above, as it clealy is
... it just takes a long time to cover finite variations, which is why they
are "pre-computed"; call it busy work.

With the rest of the chess game, each engine [and each human] will get
different results for any given position based upon how they play.

--
Thomas T. Veldhouse
Key Fingerprint: D281 77A5 63EE 82C5 5E68 00E4 7868 0ADC 4EFB 39F0

Date: 07 Dec 2006 12:01:59
From: johnny T
Subject: Re: Chess computers without opening book

Thomas T. Veldhouse wrote:

> It is easy to recompute. Anybody can write a program [with enough skill] and
> they will always get the same results.

This is true, and not the point at all. End game tablebases have major
advantages in time... It takes too long to recreate them on the fly
for most uses, and since as you say, there is no contradiction, there is
no reason NOT to have them precomputed, and major reasons to (IE, under
current hardware, they cannot be computed quick enough for even use
under classic conditions).

Date: 08 Dec 2006 14:28:09
From: Guy Macon
Subject: Re: Chess computers without opening book

johnny T wrote:
>
>Thomas T. Veldhouse wrote:
>
>> It is easy to recompute. Anybody can write a program [with enough skill] and
>> they will always get the same results.
>
>This is true, and not the point at all. End game tablebases have major
>advantages in time... It takes too long to recreate them on the fly
>for most uses, and since as you say, there is no contradiction, there is
>no reason NOT to have them precomputed, and major reasons to (IE, under
>current hardware, they cannot be computed quick enough for even use
>under classic conditions).

In like manner, once the decision has been made to have a program play
without an opening book, there is no reason to have the moves that it
would make for the first N moves precomputed, and major reasons to
do so -- to gain clock time and to do a deeper search than would be
possible under the game's time controls.

Note that what I just described is *not* the same thing as the usual
opening book that most computers use. It's a *computed* opening
book.

Guy Macon
<http://www.guymacon.com/ >

Date: 08 Dec 2006 16:43:09
From: johnny T
Subject: Re: Chess computers without opening book

Guy Macon wrote:

> Note that what I just described is *not* the same thing as the usual
> opening book that most computers use. It's a *computed* opening
> book.

But with opening books, computed or not, but lets say computed, will run
into completeness problems. Tablebases, by definition are complete.
There are no other moves to be seen.

Opening books, on the other hand, are either naturally, or computer
pruned. The only contain the best paths, believed at the times. But
that does not, however, always lead to the truth.

BTW, Computers already do to some extent what you are claiming, in
Ponder mode. They look ahead to get an advanced view on what the
opponent is going to do, and attempt to not think twice about as much
stuff as possible.

Date: 09 Dec 2006 10:22:02
From: Guy Macon
Subject: Re: Chess computers without opening book

johnny T wrote:
>
>Guy Macon <http://www.guymacon.com/> wrote:
>
>But with opening books, computed or not, but lets say computed,
>will run into completeness problems. Tablebases, by definition
>are complete. There are no other moves to be seen.
>
>Opening books, on the other hand, are either naturally, or computer
>pruned. The only contain the best paths, believed at the times. But
>that does not, however, always lead to the truth.

Think of a tablebase as a collection of data trees reaching backwards
from checkmate positions to N-man positions. Now think of opening
books as a data tree reaching forward from the opening position.
(If the two ever meet, chess will be solved, but that isn't going
to happen anytime soon...).

There is no reason why a computed opening book couldn't be
complete -- in the sense of covering the entire tree -- for
the first N moves. After that some pruning is needed to allow
a deeper book on even lines and not on lines where one side
throws away a queen and two rooks.

In like manner, a collection of tablebases can have, say,
a complete set of 3+4+5 man tablebases but only some 6-man
tablebases, keeping KRP vs krp and pruning out KQRRP vs, k.

These are different sorts of "competeness", of course, so
I wouldn't want to push the analogy too far.

Guy Macon
<http://www.guymacon.com/ >

Date: 09 Dec 2006 07:13:25
From: johnny T
Subject: Re: Chess computers without opening book

Guy Macon wrote:

> There is no reason why a computed opening book couldn't be
> complete --

Physics...

There are very real practical differences between opening books and
tablebases. The completeness issue is real, and to ignore them... Is
just sophistry and faith.

Physics will prevent any meaningful and complete opening book. There is
absolutely no point to argue that it will be, except for the sake of
argument.

The table bases tend to be complete, in that they follow all paths to
the end of the game for given situations.

The twain shall never meet, the beginning will never be complete, it
does not move the issue one iota to ignore the truth. Dragons may lie
at the edge of the map, but we shall NEVER get there.

Date: 12 Dec 2006 14:16:17
From: Thomas T. Veldhouse
Subject: Re: Chess computers without opening book

johnny T <[email protected] > wrote:
> Guy Macon wrote:
>
>> There is no reason why a computed opening book couldn't be
>> complete --
>
> Physics...
>

Statistics .. or mathmatics perhaps. Not much to do with Physics in chess and
less you throw the pieces at each other.

--
Thomas T. Veldhouse
Key Fingerprint: D281 77A5 63EE 82C5 5E68 00E4 7868 0ADC 4EFB 39F0

Date: 12 Dec 2006 17:49:53
From: johnny T
Subject: Re: Chess computers without opening book

Thomas T. Veldhouse wrote:
> johnny T <[email protected]> wrote:
>> Guy Macon wrote:
>>
>>> There is no reason why a computed opening book couldn't be
>>> complete --
>> Physics...
>>
>
> Statistics .. or mathmatics perhaps. Not much to do with Physics in chess and
> less you throw the pieces at each other.
>
Physics, as in we are physically unable to store that kind of
information. There are physics in information structures. Time to
access, and physical storage are two examples of that. Sometimes the
physics are so large that the thought problem is meaningless.

Date: 07 Dec 2006 19:38:23
From: David Richerby
Subject: Re: Chess computers without opening book

Thomas T. Veldhouse <[email protected] > wrote:
> David Richerby <[email protected]> wrote:
>> Thomas T. Veldhouse <[email protected]> wrote:
>>> It seems to me it should be one of the easiest. In fact, keeping
>>> that fact in mind, it seems tablebases are a way to simply speed
>>> things up because they are easy to pre-compute.
>>
>> You contradict yourself. If tablebases were easy to pre-compute,
>> they couldn't provide a significant speed-up. As it is, tablebases
>> are conceptually easy to pre-compute (in the same way that long
>> multiplication is conceptually easy) but fulfilling that concept
>> takes a certain amount of effort (what is 3423746829367346 x
>> 23926384767343, by the way?).
>
> It is easy to recompute. Anybody can write a program [with enough
> skill] and they will always get the same results.

It is also `easy' to factorize large numbers. Anyone can write a
program [with enough skill] and they will always get the same results.
Therefore, encryption is `easy' to crack.

Further, it is `easy' to solve chess (i.e., produce the perfect opening
book). Anyone can write a program [with enough skill] that simply
analyzes every possible line to the point of win, loss or draw.

Indeed, one could say this about every finite problem.

Perhaps your definition of `easy' is askew?

Dave.

--
David Richerby Metal Smokes (TM): it's like a pack
www.chiark.greenend.org.uk/~davidr/ of cigarettes that's made of steel!

Date: 07 Dec 2006 19:49:36
From: Thomas T. Veldhouse
Subject: Re: Chess computers without opening book

David Richerby <[email protected] > wrote:
> It is also `easy' to factorize large numbers. Anyone can write a
> program [with enough skill] and they will always get the same results.
> Therefore, encryption is `easy' to crack.
>

It is EASY because it has been done ... many times, and always with the same
result. It only make sense to optimize the process given the opportunity of
having enough RAM and/or storage. That IS what has been done.

A human chess player is free to memorize end games, given enough time and
brain capacity. It is NOT the fault of the computer player if its human
opponent is not up to the task.

--
Thomas T. Veldhouse
Key Fingerprint: D281 77A5 63EE 82C5 5E68 00E4 7868 0ADC 4EFB 39F0

Date: 07 Dec 2006 14:23:39
From: Guy Macon
Subject: Re: Chess computers without opening book

Thomas T. Veldhouse wrote:
>
>Guy Macon <http://www.guymacon.com> wrote:
>
>> If you are to cripple the computer, where will it end? A limit
>> on the computer's CPU clock speed? Giving the human 100X more
>> time on the chess clock?
>>
>
>What makes the end game the most difficult portion of the game to calculate?
>It seems to me it should be one of the easiest.

The proof is in the pudding. Tablebases increase the playing strength
more than opening books do.

>It seems tablebases are a way to simply speed things up because
>they are easy to pre-compute.

Opening books are also easy to pre-compute. So far getting opening
books from study of human games makes for a stronger book, but that
won't be true forever; eventually the programs will be able to calulate
better opening books.

Date: 12 Dec 2006 17:50:38
From: Gian-Carlo Pascutto
Subject: Re: Chess computers without opening book

Guy Macon wrote:
> Thomas T. Veldhouse wrote:
>> Guy Macon <http://www.guymacon.com> wrote:
>>
>>> If you are to cripple the computer, where will it end? A limit
>>> on the computer's CPU clock speed? Giving the human 100X more
>>> time on the chess clock?
>>>
>> What makes the end game the most difficult portion of the game to calculate?
>> It seems to me it should be one of the easiest.
>
> The proof is in the pudding. Tablebases increase the playing strength
> more than opening books do.

If you have any evidence of that you should provide it.

As far as I know, nobody has been able to show that tablebases increase
the playing strength of an engine.

Just because it sounds logical doesn't mean it's true.

--
GCP

Date: 07 Dec 2006 14:51:36
From: Thomas T. Veldhouse
Subject: Re: Chess computers without opening book

Guy Macon <http://www.guymacon.com/ > wrote:
>>
>>What makes the end game the most difficult portion of the game to calculate?
>>It seems to me it should be one of the easiest.
>
> The proof is in the pudding. Tablebases increase the playing strength
> more than opening books do.

Only the endgame. Like I said, they are easy to precompute ... there is no
rocket science with tablebases. It is faster to use precomputed tablebases
than computing it on the fly ... why bother if you always get the same
results.

>
> Opening books are also easy to pre-compute. So far getting opening
> books from study of human games makes for a stronger book, but that
> won't be true forever; eventually the programs will be able to calulate
> better opening books.
>

Opening books are NOT easy to pre-compute. They are based upon games played
and positions analyzed, which is far different then working backward from
endgame. Openning books are not all inclusive of the first few moves, table
bases ARE all inclusive of the last few moves. I currently run with five man
tablebases and the computer should be able to play a five piece endgame
perfectly.

--
Thomas T. Veldhouse
Key Fingerprint: D281 77A5 63EE 82C5 5E68 00E4 7868 0ADC 4EFB 39F0

Date: 08 Dec 2006 14:21:54
From: Guy Macon
Subject: Re: Chess computers without opening book

Thomas T. Veldhouse wrote:
>
>Guy Macon <http://www.guymacon.com/> wrote:
>>>
>>>What makes the end game the most difficult portion of the game to calculate?
>>>It seems to me it should be one of the easiest.
>>
>> The proof is in the pudding. Tablebases increase the playing strength
>> more than opening books do.
>
>Only the endgame.

And opening books only help in th opening. Tablebases increase the
ratings for playing entire games more than opening books do.

>Opening books are NOT easy to pre-compute.

Yes they are.

>They are based upon games played and positions analyzed, which is
>far different then working backward from endgame.

You are confusing computed opening books with opening books compiled
from other sources. A chess computer can be set up to calculate what
it would have computed in OTB no-book play and to save the results as
a pre-computed opening book.

This isn't done very often because gathering data from games makes
for a better book, but that is likely to change.

>Openning books are not all inclusive of the first few moves, table
>bases ARE all inclusive of the last few moves.

*Computed* opening books are all-inclusive of the first few moves.
They are no different from having the computer calculate those first
few moves in real time.

Guy Macon
<http://www.guymacon.com/ >

Date: 08 Dec 2006 14:58:55
From: Thomas T. Veldhouse
Subject: Re: Chess computers without opening book

Guy Macon <http://www.guymacon.com/ > wrote:
>
> You are confusing computed opening books with opening books compiled
> from other sources. A chess computer can be set up to calculate what
> it would have computed in OTB no-book play and to save the results as
> a pre-computed opening book.
>

I most certainly am not confusing that. A human being memorizes opening book
just like a computer does, reguardless of the source [which is usually based
upon historical analysis of games played by grand masters]. A computer has
the ability to completely and accurately remember an openning book; something
a human being fails at. Tough cookies for the human.

> *Computed* opening books are all-inclusive of the first few moves.
> They are no different from having the computer calculate those first
> few moves in real time.
>

Have any examples of "computed" openning books? To my knowledge, they are all
compiled from historical sources.

--
Thomas T. Veldhouse
Key Fingerprint: D281 77A5 63EE 82C5 5E68 00E4 7868 0ADC 4EFB 39F0

Date: 06 Dec 2006 19:45:39
From:
Subject: Re: Chess computers without opening book

Timid Demagogue wrote:
> I had a discussion today about chess computers and their reliance on
> opening books. I asserted that a human GM playing Deep Fritz 10
> without its opening book would win most of the time.
>
> Would anyone care to agree or disagree? I'd appreciate any elaboration
> as to why, either way.
>
>
> -td

I agree with the original poster. First of all, an engine without a
book is predictable as hell and it would be child's play for a GM to
lead an engine into a losing position.

Most people cannot even imagine what a really first-rate opening book
looks like and is capable of doing: they've never seen or experienced
one, so they base their opinion on what they have seen. Opening books
are just an essential part of the competitive package and always will
be. What I don't understand is letting the opponent see the book;
that's crazy!

Date: 06 Dec 2006 06:05:24
From: [email protected]
Subject: Re: Chess computers without opening book

Gian-Carlo Pascutto wrote:
> k Houlsby wrote:
>
> > It doesn't have to be (which is kinda my point). Its being used confers
> > more of an advantage to a human than to a machine... familiar
> > territory, and all that.... if a machine gets into unfamiliar
> > territory, it can calculate its way out... better than a human...
>
> Them not being used would lead to games that drift into unfamiliar and
> "weird" positions soon. Since such matches have a commercial goal and
> must please an audience, this is undesirable.
>
> The Mainz Chess Classic has had a few Chess960 (Fischerrandom) matches
> between humans and computers. Obviously no opening books there.
>
> --
> GCP

"weird" positions? How so? There is no such thing as a weird position
in OTB chess. Perhaps in problem composition, yes. But I would say that
any position could be deemed reasonable from OTB play, and that is not
a reason to avoid opening books - there are so many other better and
more valid reasons to do so.

I think *eventually* we will have to have the comp play without an
opening book to give the GM's a shopt at beating it. After all, the
opening books that the comps use allow the comps to avoid having to
calculate the most complicated and difficult phase of the game - the
opening. This would truly be the Man vs. Machine match we want to see
with no 'pre-programmed' moves for the computer, i.e., the opening
book.

Chess Training?: http://chess-training.blogspot.com

Date: 06 Dec 2006 03:52:53
From: Mark Houlsby
Subject: Re: Chess computers without opening book

k Houlsby wrote:
> Gian-Carlo Pascutto wrote:
> > k Houlsby wrote:
> >
> > > It doesn't have to be (which is kinda my point). Its being used confers
> > > more of an advantage to a human than to a machine... familiar
> > > territory, and all that.... if a machine gets into unfamiliar
> > > territory, it can calculate its way out... better than a human...
> >
> > Them not being used would lead to games that drift into unfamiliar and
> > "weird" positions soon. Since such matches have a commercial goal and
> > must please an audience, this is undesirable.
> >
> > The Mainz Chess Classic has had a few Chess960 (Fischerrandom) matches
> > between humans and computers. Obviously no opening books there.
> >
> > --
> > GCP
>
> Yes. Good points Gian-Carlo, thanks. It's true that this Kramnik match
> was about selling Fritz.
>
> The Adams-Hydra match, on the other hand, was not about selling Hydra,
> so the conditions can vary, as you know. I suppose that the OP was
> talking about "no-holds-barred" matches against machines' being
> unfeasible from now on, and she may well be correct.
>
> Certainly machine vs. machine matches *already* have a tendency to
> reach the "weird" positions which you mentioned.
>
> How long before Chess960 opening theory reaches the point at which it
> spawns its own theory, and therefore its own opening books, I wonder?
> :-)
>
> OT p.s. Does Sjeng/Deep Sjeng have any events soon?
>
> --
> MH

Oops! Excuse me: I meant "...*he* may well be correct"... Sorry

Date: 06 Dec 2006 03:50:09
From: Mark Houlsby
Subject: Re: Chess computers without opening book

Gian-Carlo Pascutto wrote:
> k Houlsby wrote:
>
> > It doesn't have to be (which is kinda my point). Its being used confers
> > more of an advantage to a human than to a machine... familiar
> > territory, and all that.... if a machine gets into unfamiliar
> > territory, it can calculate its way out... better than a human...
>
> Them not being used would lead to games that drift into unfamiliar and
> "weird" positions soon. Since such matches have a commercial goal and
> must please an audience, this is undesirable.
>
> The Mainz Chess Classic has had a few Chess960 (Fischerrandom) matches
> between humans and computers. Obviously no opening books there.
>
> --
> GCP

Yes. Good points Gian-Carlo, thanks. It's true that this Kramnik match
was about selling Fritz.

The Adams-Hydra match, on the other hand, was not about selling Hydra,
so the conditions can vary, as you know. I suppose that the OP was
talking about "no-holds-barred" matches against machines' being
unfeasible from now on, and she may well be correct.

Certainly machine vs. machine matches *already* have a tendency to
reach the "weird" positions which you mentioned.

How long before Chess960 opening theory reaches the point at which it
spawns its own theory, and therefore its own opening books, I wonder?
:-)

OT p.s. Does Sjeng/Deep Sjeng have any events soon?

--
MH

Date: 12 Dec 2006 18:01:13
From: Gian-Carlo Pascutto
Subject: Re: Chess computers without opening book

k Houlsby wrote:
> Gian-Carlo Pascutto wrote:
>> k Houlsby wrote:
>>
>>> It doesn't have to be (which is kinda my point). Its being used confers
>>> more of an advantage to a human than to a machine... familiar
>>> territory, and all that.... if a machine gets into unfamiliar
>>> territory, it can calculate its way out... better than a human...
>> Them not being used would lead to games that drift into unfamiliar and
>> "weird" positions soon. Since such matches have a commercial goal and
>> must please an audience, this is undesirable.
>>
>> The Mainz Chess Classic has had a few Chess960 (Fischerrandom) matches
>> between humans and computers. Obviously no opening books there.
>>
>> --
>> GCP
>
> Yes. Good points Gian-Carlo, thanks. It's true that this Kramnik match
> was about selling Fritz.
>
> The Adams-Hydra match, on the other hand, was not about selling Hydra,
> so the conditions can vary, as you know.

Actually, I would say that the match was certainly arranged to promote
Hydra. Maybe they aren't selling it, but prestige projects (which Hydra
is) must certainly be promoted.

It should also be noted that Hydra's opening book was truncated at move
10, so your point is somewhat moot.

> Certainly machine vs. machine matches *already* have a tendency to
> reach the "weird" positions which you mentioned.

I think the audience has more chance to recognize what is normal and
what is not in a Sicilian, than if the machine would open 1.h4!

And decades of human chess practice has made us pretty sure 1.e4 is
better than 1.h4, so why not play it? :)

> OT p.s. Does Sjeng/Deep Sjeng have any events soon?

The only upcoming event that I know of is Paderborn and that is in the
middle of Christmas holidays. No, thanks!

We'll be there for the 2007 World Champs, though. Whenever they are :P

--
GCP

Date:
From: Martin Brown
Subject: Re: Chess computers without opening book

Date: 06 Dec 2006 10:06:43
From: David Richerby
Subject: Re: Chess computers without opening book

Timid Demagogue <[email protected] > wrote:
> I had a discussion today about chess computers and their reliance on
> opening books. I asserted that a human GM playing Deep Fritz 10
> without its opening book would win most of the time.

I don't see where you find the evidence for such a strong assertion.
Deep Fritz has just beaten Kramnik +2-0=4. Kramnik would have had to
have won at least four more games than he did in order to `win most of
the time', which is a massive turn-around. I find it hard to believe
that opening theory would make such a big difference, as the evalu-
ation functions these days are usually programmed to avoid closed
positions anyway. Also, the opening book to some extent benefits the
GM by him to use his book knowledge deeper into the game.

Note also that, when Hydra beat Michael Adams +5-0=1, it was using a
relatively shallow opening book -- only to move 10, as I recall.

Dave.

--
David Richerby Broken Sadistic Windows (TM): it's
www.chiark.greenend.org.uk/~davidr/ like a graphical user interface but it
wants to hurt you and it doesn't work!

Date: 07 Dec 2006 12:48:22
From: Guy Macon
Subject: Re: Chess computers without opening book

If the creators of a chess-playing program decide for whatever
reason not to use an opening book but rather to calculate every
move, it still makes sense to have a special kind of opening book
on the machine; one that contains only calculated moves that were
calculated using more time than there is available during an
actual game. The program (or multiple copies of the program) can
sit there 24/7 building up that book, then on game day can make
what are acy=tually calculated moves instantly, thus giving it
more time on the clock.

--
Guy Macon
<http://www.guymacon.com/ >

Date: 08 Dec 2006 08:50:02
From: Kenneth Sloan
Subject: Re: Chess computers without opening book

Guy Macon wrote:
> If the creators of a chess-playing program decide for whatever
> reason not to use an opening book but rather to calculate every
> move, it still makes sense to have a special kind of opening book
> on the machine; one that contains only calculated moves that were
> calculated using more time than there is available during an
> actual game. The program (or multiple copies of the program) can
> sit there 24/7 building up that book, then on game day can make
> what are acy=tually calculated moves instantly, thus giving it
> more time on the clock.
>
> --
> Guy Macon
> <http://www.guymacon.com/>
>

If computers are going to play without opening books, the FIRST thing to
be done by the authors of current programs is to write, debug, and tune
a completely new evaluation function to cover the early stages of the
game. One of the reasons that computers (and humans!) used "opening
books" is that there are moves known to be good (from experience, from
deep analysis, etc.) which appear to be BAD based on general evaluation
principles.

As a tiny example, I play the Classical French. There are *many*
opening sequences which take a long time to recognized as "playable".
If confronted with them over-the-board, it might take a long time to
convince a player that it is OK to go into those lines. In many
instances, the *only* feeble "book knowledge" that I use is: "it's OK to
play that (key) move".

Humans use this sort of knowledge routinely. There is no good reason to
bar computers from using it - the *worst* reason is that computers do it
better.

--
Kenneth Sloan [email protected]
Computer and Information Sciences +1-205-932-2213
University of Alabama at Birmingham FAX +1-205-934-5473
Birmingham, AL 35294-1170 http://www.cis.uab.edu/sloan/

Date: 09 Dec 2006 10:05:49
From: Guy Macon
Subject: Re: Chess computers without opening book

Kenneth Sloan wrote:
>
>Guy Macon <http://www.guymacon.com/> wrote:
>
>> If the creators of a chess-playing program decide for whatever
>> reason not to use an opening book but rather to calculate every
>> move, it still makes sense to have a special kind of opening book
>> on the machine; one that contains only calculated moves that were
>> calculated using more time than there is available during an
>> actual game. The program (or multiple copies of the program) can
>> sit there 24/7 building up that book, then on game day can make
>> what are acy=tually calculated moves instantly, thus giving it
>> more time on the clock.
>
>If computers are going to play without opening books, the FIRST thing to
>be done by the authors of current programs is to write, debug, and tune
>a completely new evaluation function to cover the early stages of the
>game. One of the reasons that computers (and humans!) used "opening
>books" is that there are moves known to be good (from experience, from
>deep analysis, etc.) which appear to be BAD based on general evaluation
>principles.

Excellent point! One wonders whether a computer using a computed
opening book created with an opening-optimized evaluation function
would be stronger than one using an ioening book based on human play.

For one thing, the computed book would be far wider and deeper as
the computer (or multiple computers) kept chugging away at making
the book bigger 24/7. The collection of computers could also spend
a day or a week evaluating each move for selected lines.

>As a tiny example, I play the Classical French. There are *many*
>opening sequences which take a long time to recognized as "playable".
>If confronted with them over-the-board, it might take a long time to
>convince a player that it is OK to go into those lines. In many
>instances, the *only* feeble "book knowledge" that I use is: "it's OK to
>play that (key) move".

Have you ever tried letting your favorite program ponder one of those
moves for several days? It would be interesing to see if it found the
key move, and how long it took to do so.

Guy Macon
<http://www.guymacon.com/ >

Date: 06 Dec 2006 03:07:24
From: David Kane
Subject: Re: Chess computers without opening book

"David Richerby" <[email protected] > wrote in message
news:ZR*[email protected]...
> Timid Demagogue <[email protected]> wrote:
>> I had a discussion today about chess computers and their reliance on
>> opening books. I asserted that a human GM playing Deep Fritz 10
>> without its opening book would win most of the time.
>
> I don't see where you find the evidence for such a strong assertion.
> Deep Fritz has just beaten Kramnik +2-0=4. Kramnik would have had to
> have won at least four more games than he did in order to `win most of
> the time', which is a massive turn-around. I find it hard to believe
> that opening theory would make such a big difference, as the evalu-
> ation functions these days are usually programmed to avoid closed
> positions anyway. Also, the opening book to some extent benefits the
> GM by him to use his book knowledge deeper into the game.

Don't forget that in this match, Kramnik was able to see the Fritz
screen while Fritz was in book.

Date: 05 Dec 2006 20:25:17
From: Kenneth Sloan
Subject: Re: Chess computers without opening book

Date: 05 Dec 2006 14:45:44
From: Mark Houlsby
Subject: Re: Chess computers without opening book

Timid Demagogue wrote:

> On Dec 5, 5:30 pm, "k Houlsby" <[email protected]> wrote:
>
> > First of all it depends upon which GM. Second, the sheer *calculating
> > power* of today's software/hardware is so great that (it has been
> > argued) removing the openings books doesn't necessarily make very much
> > difference... frequently "book" moves are overruled by engines on
> > account of their being tactically unsound...
> >
> > -mh
>
> Your first point is a given-there is always the human element to
> consider. Regarding your second point, if a book move with years of
> theory and application behind it is considered unsound by the computer,
> why isn't the theory being rewritten for everyone?

It is. Read an Informator sometime.

Either way, if
> removing the opening book won't make a difference, I don't understand
> why they are used in Man vs. Machine events in the first place.
>
> -td

It doesn't have to be (which is kinda my point). Its being used confers
more of an advantage to a human than to a machine... familiar
territory, and all that.... if a machine gets into unfamiliar
territory, it can calculate its way out... better than a human...

-mh

Date: 06 Dec 2006 06:58:00
From: Gian-Carlo Pascutto
Subject: Re: Chess computers without opening book

k Houlsby wrote:

> It doesn't have to be (which is kinda my point). Its being used confers
> more of an advantage to a human than to a machine... familiar
> territory, and all that.... if a machine gets into unfamiliar
> territory, it can calculate its way out... better than a human...

Them not being used would lead to games that drift into unfamiliar and
"weird" positions soon. Since such matches have a commercial goal and
must please an audience, this is undesirable.

The Mainz Chess Classic has had a few Chess960 (Fischerrandom) matches
between humans and computers. Obviously no opening books there.

--
GCP

Date: 05 Dec 2006 14:39:09
From: Timid Demagogue
Subject: Re: Chess computers without opening book

On Dec 5, 5:30 pm, "k Houlsby" <[email protected] > wrote:

> First of all it depends upon which GM. Second, the sheer *calculating
> power* of today's software/hardware is so great that (it has been
> argued) removing the openings books doesn't necessarily make very much
> difference... frequently "book" moves are overruled by engines on
> account of their being tactically unsound...
>
> -mh

Your first point is a given-there is always the human element to
consider. Regarding your second point, if a book move with years of
theory and application behind it is considered unsound by the computer,
why isn't the theory being rewritten for everyone? Either way, if
removing the opening book won't make a difference, I don't understand
why they are used in Man vs. Machine events in the first place.

-td

Date: 05 Dec 2006 14:30:01
From: Mark Houlsby
Subject: Re: Chess computers without opening book

Timid Demagogue wrote:

> I had a discussion today about chess computers and their reliance on
> opening books. I asserted that a human GM playing Deep Fritz 10
> without its opening book would win most of the time.
>
> Would anyone care to agree or disagree? I'd appreciate any elaboration
> as to why, either way.
>
>
> -td

First of all it depends upon which GM. Second, the sheer *calculating
power* of today's software/hardware is so great that (it has been
argued) removing the openings books doesn't necessarily make very much
difference... frequently "book" moves are overruled by engines on
account of their being tactically unsound...

-mh