WBEC Ridderkerk new results.

[Matthias Gemuh]

Geschrieben von: / Posted by: Matthias Gemuh at 15 February 2004 21:46:13:
Als Antwort auf: / In reply to: WBEC Ridderkerk new results. geschrieben von: / posted by: Leo Dijksman at 15 February 2004 17:05:51:



For the record, here is what has happened so far.
Someone asked me the question "Is DanChess a crafty clone?" I responded that I
had no idea, so they sent me an executable. I looked through the binary and I
found the following similarities:
1. Many identical arrays. IE things like the compact-attacks stuff to shift
diagonals for bishops, the various king-safety arrays that I use to scale the
various "defects" I find, and so forth. That was an immediate red flag. Ditto
for specific bit patterns such as the thing I use to detect the stonewall
attack, bitmaps for won king and pawn endings, and so forth. No doubt someone
could come up with the same ideas, or even read the crafty comments, but to do
things the _exact_ same way (ie even numbering the bits in a bad order for X86,
because Crafty was originally developed for the Cray which has an instruction
that effectively counts bits from MSB=0 to LSB=63, no bits set=64.
2. Dann then sent me the source. I looked at several pieces, and found that
there were too many similar pieces of code. IE Swap() was just one example.
The major differences between swap.cpp (DanChess) and swap.c (crafty) was that
(a) swap.c was rewritten to C++, and (b) the tree structure was removed and made
global since apparently he had no interest in copying the SMP stuff. If you
look at the two functions, they are identical. Dann and I didn't agree on this
as he believes that if you simply change variable names, that makes code
different. As an academician, I don't buy that. I looked at other pieces of
code, from pawn hashing, to pawn structure (ie he even does the
kingside/queenside defects stuff and stuffs it in the pawn hash) and so forth,
and the code is just like mine. Many parts of the eval show the same kind of
"translation" effects (c to c++) and some variable name changes, but that is
all.
3. In looking at it, the bitboard stuff is the same. The attack stuff (swap.c)
is the same. Much of the eval is identical. I looked at my evaluate.h and his
board_evaluate.cpp and there is _way_ too much identical code. I quit at that
point as there was little need to see if _more_ was identical, I was already
beyond what was reasonable.
Is it a clone?
You have to define clone for yourself. I think too much was taken directly from
the crafty source to call this an original program.
That is my opinion.
Yours may be different...
Feel free to post this on the winboard forum...




BigLion + Taktix

[Dann Corbit]

Geschrieben von: / Posted by: Dann Corbit at 15 February 2004 21:48:16:
Als Antwort auf: / In reply to: Robert Hyatt about Danchess and Crafty (in CCC) geschrieben von: / posted by: Matthias Gemuh at 15 February 2004 21:46:13:

For the record, here is what has happened so far.
Someone asked me the question "Is DanChess a crafty clone?" I responded that I
had no idea, so they sent me an executable. I looked through the binary and I
found the following similarities:
1. Many identical arrays. IE things like the compact-attacks stuff to shift
diagonals for bishops, the various king-safety arrays that I use to scale the
various "defects" I find, and so forth. That was an immediate red flag. Ditto
for specific bit patterns such as the thing I use to detect the stonewall
attack, bitmaps for won king and pawn endings, and so forth. No doubt someone
could come up with the same ideas, or even read the crafty comments, but to do
things the _exact_ same way (ie even numbering the bits in a bad order for X86,
because Crafty was originally developed for the Cray which has an instruction
that effectively counts bits from MSB=0 to LSB=63, no bits set=64.
2. Dann then sent me the source. I looked at several pieces, and found that
there were too many similar pieces of code. IE Swap() was just one example.
The major differences between swap.cpp (DanChess) and swap.c (crafty) was that
(a) swap.c was rewritten to C++, and (b) the tree structure was removed and made
global since apparently he had no interest in copying the SMP stuff. If you
look at the two functions, they are identical. Dann and I didn't agree on this
as he believes that if you simply change variable names, that makes code
different. As an academician, I don't buy that. I looked at other pieces of
code, from pawn hashing, to pawn structure (ie he even does the
kingside/queenside defects stuff and stuffs it in the pawn hash) and so forth,
and the code is just like mine. Many parts of the eval show the same kind of
"translation" effects (c to c++) and some variable name changes, but that is
all.
3. In looking at it, the bitboard stuff is the same. The attack stuff (swap.c)
is the same. Much of the eval is identical. I looked at my evaluate.h and his
board_evaluate.cpp and there is _way_ too much identical code. I quit at that
point as there was little need to see if _more_ was identical, I was already
beyond what was reasonable.
Is it a clone?
You have to define clone for yourself. I think too much was taken directly from
the crafty source to call this an original program.
That is my opinion.
Yours may be different...
Feel free to post this on the winboard forum...

Instead of crossposting, I think it would be very good to read the entire thread. It starts here (reading requres a free CCC account):
http://www.talkchess.com/forums/1/message.html?349222



my ftp site {remove http:// unless you like error messages}

[Matthias Gemuh]

Geschrieben von: / Posted by: Matthias Gemuh at 15 February 2004 22:05:50:
Als Antwort auf: / In reply to: Re: Robert Hyatt about Danchess and Crafty (in CCC) geschrieben von: / posted by: Dann Corbit at 15 February 2004 21:48:16:

Instead of crossposting, I think it would be very good to read the entire thread. It starts here (reading requres a free CCC account):
http://www.talkchess.com/forums/1/message.html?349222

First I wanted to post a link.
Then I remembered that CCC needs an account (though free)
http://www.talkchess.com/forums/1/message.html?349253

[Tom Likens]

Geschrieben von: / Posted by: Tom Likens at 16 February 2004 03:30:16:
Als Antwort auf: / In reply to: Re: Dann as judge is too liberal. geschrieben von: / posted by: Dann Corbit at 15 February 2004 21:40:46:

Anyway, I think Dann's judgement is extremely liberal.
Dr. Hyatt certainly knows why he speaks of cut an paste.
I wouldn't go that far, but it sure looks like search and replace.
Would be OK if starting point for that is Nero.
I can clone Crafty and end up with only 10% total code size by
using similar data structures and mimmicking Crafty's 10 key routines.

/Matthias.

I understood from Dann's post that 70% of Danchess code is not similiar to crafty.

Many parts are completely different. All of this is neither here nor there. He is doing a complete rewrite to 0x88. He has some ideas that are fresh and interesting and found nowhere else.
Amazingly, he already has a working version. It's just as strong as the old one.
Daniel Shawul is a very smart person. Expect big things from him. He does not have a lot of experience. But that will gather like a snowball down a hill.

Sounds like he fits the classic profile of a chess programmer, full of fresh ideas and lots of energy
The question is if these on-the-paper ideas actually work.
Only time will tell.
The are already many smart people in chess programming, the question is
not if he is 'smart' but if he is 'smarter'

Probably determination is even more important than that. Some of the smartest chess programmers simply got tired of it.

Yes, what is Mr. Moreland up to these days!
--tom

[Dann Corbit]

Geschrieben von: / Posted by: Dann Corbit at 16 February 2004 19:53:58:
Als Antwort auf: / In reply to: Re: Dann as judge is too liberal. geschrieben von: / posted by: Tom Likens at 16 February 2004 03:30:16:

Anyway, I think Dann's judgement is extremely liberal.
Dr. Hyatt certainly knows why he speaks of cut an paste.
I wouldn't go that far, but it sure looks like search and replace.
Would be OK if starting point for that is Nero.
I can clone Crafty and end up with only 10% total code size by
using similar data structures and mimmicking Crafty's 10 key routines.

/Matthias.

I understood from Dann's post that 70% of Danchess code is not similiar to crafty.

Many parts are completely different. All of this is neither here nor there. He is doing a complete rewrite to 0x88. He has some ideas that are fresh and interesting and found nowhere else.
Amazingly, he already has a working version. It's just as strong as the old one.
Daniel Shawul is a very smart person. Expect big things from him. He does not have a lot of experience. But that will gather like a snowball down a hill.

Sounds like he fits the classic profile of a chess programmer, full of fresh ideas and lots of energy
The question is if these on-the-paper ideas actually work.
Only time will tell.
The are already many smart people in chess programming, the question is
not if he is 'smart' but if he is 'smarter'

Probably determination is even more important than that. Some of the smartest chess programmers simply got tired of it.

Yes, what is Mr. Moreland up to these days!
--tom

And Mr. Kittinger and Mr. Beal and Mr. Berliner and Mr. Hsu and Mr. Campbell and Mr. Heinz and Mr. ...




my ftp site {remove http:// unless you like error messages}

[Tom Likens]

Geschrieben von: / Posted by: Tom Likens at 17 February 2004 04:00:58:
Als Antwort auf: / In reply to: Re: Dann as judge is too liberal. geschrieben von: / posted by: Dann Corbit at 16 February 2004 19:53:58:

[--snip--]

Probably determination is even more important than that. Some of the smartest chess programmers simply got tired of it.

Yes, what is Mr. Moreland up to these days!

And Mr. Kittinger and Mr. Beal and Mr. Berliner and Mr. Hsu and Mr. Campbell and Mr. Heinz and Mr. ...

Sigh, well it *is* a hard problem.
--tom

[Tord Romstad]

Geschrieben von: / Posted by: Tord Romstad at 17 February 2004 13:14:41:
Als Antwort auf: / In reply to: Re: Dann as judge is too liberal. geschrieben von: / posted by: Dann Corbit at 16 February 2004 19:53:58:

Probably determination is even more important than that. Some of the smartest chess programmers simply got tired of it.

Yes, what is Mr. Moreland up to these days!

And Mr. Kittinger and Mr. Beal and Mr. Berliner and Mr. Hsu and Mr. Campbell and Mr. Heinz and Mr. ...

Please don't forget Mr. Lang, arguably the most prominent of them all!
Tord

[Tom Likens]

Geschrieben von: / Posted by: Tom Likens at 17 February 2004 16:33:33:
Als Antwort auf: / In reply to: Re: Dann as judge is too liberal. geschrieben von: / posted by: Tord Romstad at 17 February 2004 13:14:41:

Probably determination is even more important than that. Some of the smartest chess programmers simply got tired of it.

Yes, what is Mr. Moreland up to these days!

And Mr. Kittinger and Mr. Beal and Mr. Berliner and Mr. Hsu and Mr. Campbell and Mr. Heinz and Mr. ...

Please don't forget Mr. Lang, arguably the most prominent of them all!
Tord

This is a good point. Richard Lang did amazing things with his Chess Genius series of
programs. I remember when he beat Kasparov on a Pentium 90!!
--tom