Opening Bug in Nejmet and Pharaon

Archive of the old Parsimony forum. Some messages couldn't be restored. Limitations: Search for authors does not work, Parsimony specific formats do not work, threaded view does not work properly. Posting is disabled.

Re: Opening Bug in Nejmet and Pharaon

Postby Uri Blass » 31 May 2003, 12:48

Geschrieben von: / Posted by: Uri Blass at 31 May 2003 13:48:20:
Als Antwort auf: / In reply to: Re: Opening Bug in Nejmet and Pharaon geschrieben von: / posted by: Richard Pijl at 31 May 2003 11:35:53:
I don't know if this problem has been discussed before. I did a search but it didn't turn up anything.
The problem is easily simulated if you let the engines play White. It can also be simulated when the programs play black. Engines are set to use the opening books.
1. Engines play 1. e4 or 1. d4
2. You move 1. ...Nc6
3. Then move 2. ...Nb8
4. Repeat the knight moves
5. Engines will draw with 3-move repetition
Are there any other engines which show similiar bug?
This can happen with engine that store their book by positions instead of trees.
Early versions of the Baron had the same bug. Since someone used this on ICC to get a draw from the Baron, the Baron doesn't play repetitions while in book.
Richard.
Hi Richard,
i have also only a positional book, but Holmes would never play Ng1, and this ssems to be the bug and not the opening book. In the games from Telmo, various engines seems to have absolutely the same bug. Very curious that various engines are playing Ng1.
So i have checked the engines with open source, which are older than Nejmet and the other ones:
Arasan,GNU,Phalanx,Faile and Crafty doesn't have these bug but Exchess has absolutely the same bug.
Andreas
The position after 1.e4 occurs in the book. After 1..Nc6 2.Nf3 Nb8 then 3.Ng1 would probably be the only move that leads to a book position. Unless repetitions are excluded...
Richard.
Do you do one ply search to find position that is in book.
No human thinks like this so I do not understand why programmers do it.
It is not the natural way to think about book.
The natural way is to think about it as a list of positions with moves.
If you do not like moves then you can have 2 books.
book A is for position that you use book and book B is for positions that
you know as good positions so if the position is in Book A you can search for a move that leads to book B.
Uri
Uri Blass
 

Re: Opening Bug in Nejmet and Pharaon

Postby Richard Pijl » 31 May 2003, 14:05

Geschrieben von: / Posted by: Richard Pijl at 31 May 2003 15:05:24:
Als Antwort auf: / In reply to: Re: Opening Bug in Nejmet and Pharaon geschrieben von: / posted by: Uri Blass at 31 May 2003 13:48:20:
I don't know if this problem has been discussed before. I did a search but it didn't turn up anything.
The problem is easily simulated if you let the engines play White. It can also be simulated when the programs play black. Engines are set to use the opening books.
1. Engines play 1. e4 or 1. d4
2. You move 1. ...Nc6
3. Then move 2. ...Nb8
4. Repeat the knight moves
5. Engines will draw with 3-move repetition
Are there any other engines which show similiar bug?
This can happen with engine that store their book by positions instead of trees.
Early versions of the Baron had the same bug. Since someone used this on ICC to get a draw from the Baron, the Baron doesn't play repetitions while in book.
Richard.
Hi Richard,
i have also only a positional book, but Holmes would never play Ng1, and this ssems to be the bug and not the opening book. In the games from Telmo, various engines seems to have absolutely the same bug. Very curious that various engines are playing Ng1.
So i have checked the engines with open source, which are older than Nejmet and the other ones:
Arasan,GNU,Phalanx,Faile and Crafty doesn't have these bug but Exchess has absolutely the same bug.
Andreas
The position after 1.e4 occurs in the book. After 1..Nc6 2.Nf3 Nb8 then 3.Ng1 would probably be the only move that leads to a book position. Unless repetitions are excluded...
Richard.
Do you do one ply search to find position that is in book.
No human thinks like this so I do not understand why programmers do it.
It is not the natural way to think about book.
The natural way is to think about it as a list of positions with moves.
If you do not like moves then you can have 2 books.
book A is for position that you use book and book B is for positions that
you know as good positions so if the position is in Book A you can search for a move that leads to book B.
Correct.
Because it makes the bookstructure very easy. Just zobristkeys ...
Perhaps. But then you may miss some transpositions into the book.
I agree it may be better though. But I lack the motivation to change it ...
This way, once you leave the book, you'll never return to it. I don't like that.
Richard.
Richard Pijl
 

Re: Opening Bug in Nejmet and Pharaon

Postby Uri Blass » 31 May 2003, 14:56

Geschrieben von: / Posted by: Uri Blass at 31 May 2003 15:56:15:
Als Antwort auf: / In reply to: Re: Opening Bug in Nejmet and Pharaon geschrieben von: / posted by: Richard Pijl at 31 May 2003 15:05:24:
I don't know if this problem has been discussed before. I did a search but it didn't turn up anything.
The problem is easily simulated if you let the engines play White. It can also be simulated when the programs play black. Engines are set to use the opening books.
1. Engines play 1. e4 or 1. d4
2. You move 1. ...Nc6
3. Then move 2. ...Nb8
4. Repeat the knight moves
5. Engines will draw with 3-move repetition
Are there any other engines which show similiar bug?
This can happen with engine that store their book by positions instead of trees.
Early versions of the Baron had the same bug. Since someone used this on ICC to get a draw from the Baron, the Baron doesn't play repetitions while in book.
Richard.
Hi Richard,
i have also only a positional book, but Holmes would never play Ng1, and this ssems to be the bug and not the opening book. In the games from Telmo, various engines seems to have absolutely the same bug. Very curious that various engines are playing Ng1.
So i have checked the engines with open source, which are older than Nejmet and the other ones:
Arasan,GNU,Phalanx,Faile and Crafty doesn't have these bug but Exchess has absolutely the same bug.
Andreas
The position after 1.e4 occurs in the book. After 1..Nc6 2.Nf3 Nb8 then 3.Ng1 would probably be the only move that leads to a book position. Unless repetitions are excluded...
Richard.
Do you do one ply search to find position that is in book.
No human thinks like this so I do not understand why programmers do it.
It is not the natural way to think about book.
The natural way is to think about it as a list of positions with moves.
If you do not like moves then you can have 2 books.
book A is for position that you use book and book B is for positions that
you know as good positions so if the position is in Book A you can search for a move that leads to book B.
Correct.
Because it makes the bookstructure very easy. Just zobristkeys ...
Perhaps. But then you may miss some transpositions into the book.
I agree it may be better though. But I lack the motivation to change it ...
This way, once you leave the book, you'll never return to it. I don't like that.
I never thought about in that way.
I thought that you need to have the FEN for every position.
I do not like zobrist key because it means that if you have a big book (say 2^24 positions) and 64 bits for hash key then you may have hash collision in the book(small chances but practical chances)
Another problem is that in every move when you are not in book you may
have probability of 1/2^40 to believe that you are in book and that probability may increase when the book increase.
Summery:It seems to me bad design decision and I prefer even to play without book.


No
When I leave book A I can return book A.
I do not do it that way because my book is of moves with no positions in it
because I also had not the motivation to change it without some help.
Uri
Uri Blass
 

Re: Opening Bug in Nejmet and Pharaon

Postby Gian-Carlo Pascutto » 31 May 2003, 15:38

Geschrieben von: / Posted by: Gian-Carlo Pascutto at 31 May 2003 16:38:03:
Als Antwort auf: / In reply to: Re: Opening Bug in Nejmet and Pharaon geschrieben von: / posted by: Uri Blass at 31 May 2003 13:48:20:
No human thinks like this so I do not understand why programmers do it.
When I play chess, I'm not thinking in terms of alpha and beta either.
--
GCP
Gian-Carlo Pascutto
 

Re: Opening Bug in Nejmet and Pharaon

Postby Richard Pijl » 31 May 2003, 15:39

Geschrieben von: / Posted by: Richard Pijl at 31 May 2003 16:39:35:
Als Antwort auf: / In reply to: Re: Opening Bug in Nejmet and Pharaon geschrieben von: / posted by: Uri Blass at 31 May 2003 15:56:15:
I don't know if this problem has been discussed before. I did a search but it didn't turn up anything.
The problem is easily simulated if you let the engines play White. It can also be simulated when the programs play black. Engines are set to use the opening books.
1. Engines play 1. e4 or 1. d4
2. You move 1. ...Nc6
3. Then move 2. ...Nb8
4. Repeat the knight moves
5. Engines will draw with 3-move repetition
Are there any other engines which show similiar bug?
This can happen with engine that store their book by positions instead of trees.
Early versions of the Baron had the same bug. Since someone used this on ICC to get a draw from the Baron, the Baron doesn't play repetitions while in book.
Richard.
Hi Richard,
i have also only a positional book, but Holmes would never play Ng1, and this ssems to be the bug and not the opening book. In the games from Telmo, various engines seems to have absolutely the same bug. Very curious that various engines are playing Ng1.
So i have checked the engines with open source, which are older than Nejmet and the other ones:
Arasan,GNU,Phalanx,Faile and Crafty doesn't have these bug but Exchess has absolutely the same bug.
Andreas
The position after 1.e4 occurs in the book. After 1..Nc6 2.Nf3 Nb8 then 3.Ng1 would probably be the only move that leads to a book position. Unless repetitions are excluded...
Richard.
Do you do one ply search to find position that is in book.
No human thinks like this so I do not understand why programmers do it.
It is not the natural way to think about book.
The natural way is to think about it as a list of positions with moves.
If you do not like moves then you can have 2 books.
book A is for position that you use book and book B is for positions that
you know as good positions so if the position is in Book A you can search for a move that leads to book B.
Correct.
Because it makes the bookstructure very easy. Just zobristkeys ...
Perhaps. But then you may miss some transpositions into the book.
I agree it may be better though. But I lack the motivation to change it ...
This way, once you leave the book, you'll never return to it. I don't like that.
I never thought about in that way.
I thought that you need to have the FEN for every position.
I do not like zobrist key because it means that if you have a big book (say 2^24 positions) and 64 bits for hash key then you may have hash collision in the book(small chances but practical chances)
Another problem is that in every move when you are not in book you may
have probability of 1/2^40 to believe that you are in book and that probability may increase when the book increase.
Summery:It seems to me bad design decision and I prefer even to play without book.


No
When I leave book A I can return book A.
I do not do it that way because my book is of moves with no positions in it
because I also had not the motivation to change it without some help.
Uri
Chances of 1/(2^40) are too small to consider. It would mean that in one out of more than 4 Billion games(considering 1 of 2^32 games to get to a familiar number) you'll get problems. Small enough of a chance for me.
Your choice.
Of course, I agree.
Richard.
Richard Pijl
 

Re: Opening Bug in Nejmet and Pharaon

Postby Gian-Carlo Pascutto » 31 May 2003, 15:43

Geschrieben von: / Posted by: Gian-Carlo Pascutto at 31 May 2003 16:43:47:
Als Antwort auf: / In reply to: Re: Opening Bug in Nejmet and Pharaon geschrieben von: / posted by: Uri Blass at 31 May 2003 15:56:15:
Another problem is that in every move when you are not in book you may
have probability of 1/2^40 to believe that you are in book and that >probability may increase when the book increase.
Summery:It seems to me bad design decision and I prefer even to play without >book.
Assuming we probe for 20 moves, and probe 20 subpositions every move, the probablity of a collision in a book with 2^24 entries is 1 in 2 748 779 069 games.
Yes, obviously unacceptable. Better do away with Zobrist hashing
alltogether.
--
GCP
Gian-Carlo Pascutto
 

Re: Amy's turn

Postby Thorsten Greiner » 31 May 2003, 16:35

Geschrieben von: / Posted by: Thorsten Greiner at 31 May 2003 17:35:06:
Als Antwort auf: / In reply to: Amy's turn geschrieben von: / posted by: Telmo at 31 May 2003 05:41:38:

Thanks for the hint ... will fix it in next release :-)
-Thorsten
Thorsten Greiner
 

Re: Opening Bug in Nejmet and Pharaon

Postby Uri Blass » 31 May 2003, 16:55

Geschrieben von: / Posted by: Uri Blass at 31 May 2003 17:55:15:
Als Antwort auf: / In reply to: Re: Opening Bug in Nejmet and Pharaon geschrieben von: / posted by: Gian-Carlo Pascutto at 31 May 2003 16:38:03:
No human thinks like this so I do not understand why programmers do it.
When I play chess, I'm not thinking in terms of alpha and beta either.
--
GCP
Yes but this is a different case.
Book is not something that you need to do in a different way when you have more time and I see no reason not
to remember it in a similiar way to humans.
The fact that it is possible to use harddisk also mean that there is no reason
to take the small risk of using only zobrist keys to check if the position is in book.
Uri
Uri Blass
 

Re: Opening Bug in Nejmet and Pharaon

Postby Dieter Bürßner » 31 May 2003, 18:33

Geschrieben von: / Posted by: Dieter Bürßner at 31 May 2003 19:33:34:
Als Antwort auf: / In reply to: Re: Opening Bug in Nejmet and Pharaon geschrieben von: / posted by: Gian-Carlo Pascutto at 31 May 2003 16:38:03:
When I play chess, I'm not thinking in terms of alpha and beta either.
Are you sure? Just another thought (that may not contradict to what you said)
I have the impression, that many chess players are thinking with alpha-beta-search (with very variable depth). I remember that I have read one article in Rochade Europa, where one GM (perhaps it was R. Hübner) was describing his impressions from a game. I can only paraphrase. He said things, like I calculated this variation to end. White could refute this [some black answer to an earlier white move in the thinking process] by ... And more comments like this. To me it sounded, as he described the alpha-beta algorithm.
Often Analysis shown with variants is very much like the search tree searched by alpha-beta search (and the evaluation of the endnodes is given by, =, +-, etc.)
Regards,
Dieter
Dieter Bürßner
 

Re: Opening Bug in Nejmet and Pharaon

Postby Gian-Carlo Pascutto » 31 May 2003, 18:55

Geschrieben von: / Posted by: Gian-Carlo Pascutto at 31 May 2003 19:55:37:
Als Antwort auf: / In reply to: Re: Opening Bug in Nejmet and Pharaon geschrieben von: / posted by: Dieter Bürßner at 31 May 2003 19:33:34:
When I play chess, I'm not thinking in terms of alpha and beta either.
Are you sure? Just another thought (that may not contradict to what you said)
There a similarities, but the process ends up being very different
(much more selective, variable depth, much less nps :) ).
It makes no sense for a human to try to exactly emulate the computer,
just as it doesn't make sense for the computer to exactly emulate the
human.
--
GCP
Gian-Carlo Pascutto
 

Re: Bingo! Poor Nejmet

Postby Telmo » 31 May 2003, 20:37

Geschrieben von: / Posted by: Telmo at 31 May 2003 21:37:22:
Als Antwort auf: / In reply to: Re: Bingo! Poor Nejmet geschrieben von: / posted by: Dr.WAEL DEEB at 31 May 2003 10:09:14:
Hi,
Bingo??Do you really get happy when you make such a silly draw!?
I suspect it refers to reproducing a reported bug, not happiness in achieving draws as such.
Regards,
Mogens
Hi Mogens,
If it so,then I apologize!My aim is not to upset Telmo!
Regards,
Dr.WAEL DEEB
Wael, my guess was that you feel some envy: you have to fight hard to get a draw against Winboard engines, and I -due to better opening preparation- got it in a few moves...
Telmo
 

Re: Bingo! Poor Nejmet

Postby Dr.WAEL DEEB » 31 May 2003, 20:55

Geschrieben von: / Posted by: Dr.WAEL DEEB at 31 May 2003 21:55:35:
Als Antwort auf: / In reply to: Re: Bingo! Poor Nejmet geschrieben von: / posted by: Telmo at 31 May 2003 21:37:22:
Hi,
Bingo??Do you really get happy when you make such a silly draw!?
I suspect it refers to reproducing a reported bug, not happiness in achieving draws as such.
Regards,
Mogens
Hi Mogens,
If it so,then I apologize!My aim is not to upset Telmo!
Regards,
Dr.WAEL DEEB
Wael, my guess was that you feel some envy: you have to fight hard to get a draw against Winboard engines, and I -due to better opening preparation- got it in a few moves...
Hi,
Yes,I confess....it's true!How do you do it,brother!?
Regards,
Dr.WAEL DEEB
Dr.WAEL DEEB
 

Previous

Return to Archive (Old Parsimony Forum)

Who is online

Users browsing this forum: No registered users and 11 guests

cron