Dear readers,
I purchased the 5-man tablebases from CB for my engines many years ago. I knew then that the CDs didn't include every 5-man ending because some are so rare and because the engines solved the others.
But in my correspondence chess games, I discovered some obviously erroneous Rybka 3 output in B+P vs. P endings. In positions with the wrong colored Bishop, Rybka claims wins. Stockfish has enough endgame knowledge to know when the position is drawn and when it's not. And it's correct output is confirmed by 6-man websites.
Yes, I've read that Rybka 3 has a known "wrong B" endgame problem but I assumed that this was minimized in its Monte Carlo mode because games are played to mate.
However, I saw examples of Rybka 3 Monte Carlo analysis where resignations occurred in drawn positions! These Monte Carlo games are thus not played out to mate. Naturally, this means my Rybka 3's output and its Monte Carlo output are bogus.
So, I need to download the B+P vs. P tablebases from somewhere to fix this. Do you have any recommendations?
All the best,
Steve
B+P vs. P Tablebase Needed
Moderators: hgm, Rebel, chrisw
-
- Posts: 2488
- Joined: Wed Mar 08, 2006 9:40 pm
- Location: Eden Prairie, Minnesota
- Full name: Stephen Ham
Re: B+P vs. P Tablebase Needed
If you've already been here then I ask for forgiveness: http://tablebase.sesse.net/
-
- Posts: 2488
- Joined: Wed Mar 08, 2006 9:40 pm
- Location: Eden Prairie, Minnesota
- Full name: Stephen Ham
Re: B+P vs. P Tablebase Needed
Hi Franklin,kingliveson wrote:If you've already been here then I ask for forgiveness: http://tablebase.sesse.net/
Thanks so very much for your kind offer to help.
At the risk of looking like a moron, how do I locate specifically B+P vs. P at that site? I didn't understand their file naming system there.
All the best,
Steve
-
- Posts: 6401
- Joined: Thu Mar 09, 2006 8:30 pm
- Location: Chicago, Illinois, USA
Re: B+P vs. P Tablebase Needed
http://tablebase.sesse.net/3-4-5/kbpkp.nbw.emdStephen Ham wrote:Hi Franklin,kingliveson wrote:If you've already been here then I ask for forgiveness: http://tablebase.sesse.net/
Thanks so very much for your kind offer to help.
At the risk of looking like a moron, how do I locate specifically B+P vs. P at that site? I didn't understand their file naming system there.
All the best,
Steve
http://tablebase.sesse.net/3-4-5/kbpkp.nbb.emd
I think you want exactly those two files (right click to save as)
Alternatively you can find them here:
http://www.olympuschess.com/egtb/nalimov/
Miguel
Re: B+P vs. P Tablebase Needed
I hope this helps:Stephen Ham wrote:Hi Franklin,kingliveson wrote:If you've already been here then I ask for forgiveness: http://tablebase.sesse.net/
Thanks so very much for your kind offer to help.
At the risk of looking like a moron, how do I locate specifically B+P vs. P at that site? I didn't understand their file naming system there.
All the best,
Steve
kbbkn
[D]8/8/3k2n1/8/8/1B6/2K2B2/8 b - - 0 1
kpkbb
[D]8/2b5/1k6/1b6/6P1/5K2/8/8 w - - 0 1
-
- Posts: 2488
- Joined: Wed Mar 08, 2006 9:40 pm
- Location: Eden Prairie, Minnesota
- Full name: Stephen Ham
Re: B+P vs. P Tablebase Needed
Thanks, Franklin. Yes, I knew that. It was the rest of the file names that threw me. But, I see now that I should have ignored that part.kingliveson wrote:I hope this helps:Stephen Ham wrote:Hi Franklin,kingliveson wrote:If you've already been here then I ask for forgiveness: http://tablebase.sesse.net/
Thanks so very much for your kind offer to help.
At the risk of looking like a moron, how do I locate specifically B+P vs. P at that site? I didn't understand their file naming system there.
All the best,
Steve
kbbkn
[D]8/8/3k2n1/8/8/1B6/2K2B2/8 b - - 0 1
kpkbb
[D]8/2b5/1k6/1b6/6P1/5K2/8/8 w - - 0 1
Thanks again!
Steve
-
- Posts: 2488
- Joined: Wed Mar 08, 2006 9:40 pm
- Location: Eden Prairie, Minnesota
- Full name: Stephen Ham
Re: B+P vs. P Tablebase Needed
Dear Miguel,michiguel wrote:http://tablebase.sesse.net/3-4-5/kbpkp.nbw.emdStephen Ham wrote:Hi Franklin,kingliveson wrote:If you've already been here then I ask for forgiveness: http://tablebase.sesse.net/
Thanks so very much for your kind offer to help.
At the risk of looking like a moron, how do I locate specifically B+P vs. P at that site? I didn't understand their file naming system there.
All the best,
Steve
http://tablebase.sesse.net/3-4-5/kbpkp.nbb.emd
I think you want exactly those two files (right click to save as)
Alternatively you can find them here:
http://www.olympuschess.com/egtb/nalimov/
Miguel
Thank you so very much! I'm in your debt.
All the best,
Steve
-
- Posts: 20943
- Joined: Mon Feb 27, 2006 7:30 pm
- Location: Birmingham, AL
Re: B+P vs. P Tablebase Needed
All of the 3-4-5 piece tables are on my ftp box... ftp.cis.uab.edu/pub/hyatt/TB/3-4-5Stephen Ham wrote:Dear readers,
I purchased the 5-man tablebases from CB for my engines many years ago. I knew then that the CDs didn't include every 5-man ending because some are so rare and because the engines solved the others.
But in my correspondence chess games, I discovered some obviously erroneous Rybka 3 output in B+P vs. P endings. In positions with the wrong colored Bishop, Rybka claims wins. Stockfish has enough endgame knowledge to know when the position is drawn and when it's not. And it's correct output is confirmed by 6-man websites.
Yes, I've read that Rybka 3 has a known "wrong B" endgame problem but I assumed that this was minimized in its Monte Carlo mode because games are played to mate.
However, I saw examples of Rybka 3 Monte Carlo analysis where resignations occurred in drawn positions! These Monte Carlo games are thus not played out to mate. Naturally, this means my Rybka 3's output and its Monte Carlo output are bogus.
So, I need to download the B+P vs. P tablebases from somewhere to fix this. Do you have any recommendations?
All the best,
Steve
Bob
Re: B+P vs. P Tablebase Needed
Dr. Hyatt, I actually got my copy from your server...will consider hosting the 6 men egtb?bob wrote:All of the 3-4-5 piece tables are on my ftp box... ftp.cis.uab.edu/pub/hyatt/TB/3-4-5Stephen Ham wrote:Dear readers,
I purchased the 5-man tablebases from CB for my engines many years ago. I knew then that the CDs didn't include every 5-man ending because some are so rare and because the engines solved the others.
But in my correspondence chess games, I discovered some obviously erroneous Rybka 3 output in B+P vs. P endings. In positions with the wrong colored Bishop, Rybka claims wins. Stockfish has enough endgame knowledge to know when the position is drawn and when it's not. And it's correct output is confirmed by 6-man websites.
Yes, I've read that Rybka 3 has a known "wrong B" endgame problem but I assumed that this was minimized in its Monte Carlo mode because games are played to mate.
However, I saw examples of Rybka 3 Monte Carlo analysis where resignations occurred in drawn positions! These Monte Carlo games are thus not played out to mate. Naturally, this means my Rybka 3's output and its Monte Carlo output are bogus.
So, I need to download the B+P vs. P tablebases from somewhere to fix this. Do you have any recommendations?
All the best,
Steve
Bob
-
- Posts: 20943
- Joined: Mon Feb 27, 2006 7:30 pm
- Location: Birmingham, AL
Re: B+P vs. P Tablebase Needed
A few years ago our department chair was unwilling to commit enough funding to put together an FTP box large enough to host them. Takes a few terabytes (I don't know the exact size now, I did when the old FTP box failed...)kingliveson wrote:Dr. Hyatt, I actually got my copy from your server...will consider hosting the 6 men egtb?bob wrote:All of the 3-4-5 piece tables are on my ftp box... ftp.cis.uab.edu/pub/hyatt/TB/3-4-5Stephen Ham wrote:Dear readers,
I purchased the 5-man tablebases from CB for my engines many years ago. I knew then that the CDs didn't include every 5-man ending because some are so rare and because the engines solved the others.
But in my correspondence chess games, I discovered some obviously erroneous Rybka 3 output in B+P vs. P endings. In positions with the wrong colored Bishop, Rybka claims wins. Stockfish has enough endgame knowledge to know when the position is drawn and when it's not. And it's correct output is confirmed by 6-man websites.
Yes, I've read that Rybka 3 has a known "wrong B" endgame problem but I assumed that this was minimized in its Monte Carlo mode because games are played to mate.
However, I saw examples of Rybka 3 Monte Carlo analysis where resignations occurred in drawn positions! These Monte Carlo games are thus not played out to mate. Naturally, this means my Rybka 3's output and its Monte Carlo output are bogus.
So, I need to download the B+P vs. P tablebases from somewhere to fix this. Do you have any recommendations?
All the best,
Steve
Bob