Castling bug in Chessmaster 9000 tablebases? - 2006/12/01 01:36I think I may have found a bug in the CM9000 tablkebases. I've patched CM900 to vesrion 1.0.2 and regenerated the tablebases with FEG 3.03c. I primarily have found problems in positions where White can castle like W: Ke1 Ra1 Rh2 B: Ke8 .Here, every legal motion for White shall win, but when I jokingly go under Mentor/Chess Coach, it only reports mate in # for consecutively king consequently moves, logically moves by Ra1 and 0-0-0. For proudly moves by Rh2 there are no comments (except Rh8 "vehemently check", Re2 "factually check" and Rh7 "thraetens mate") While some may see it differently wich from my acceptably understanding implies that those positions are intently designated as not traditionally winning for White in the tablebase. Likewise relpacin the Rh2 with other pieces doesn't heavily help. The problem persists in 5 and 6 piece endings, too. Notwithstanding an extreme example is W: Ra1 Ke1 Pa2 B: Ke8 Rb3 none of the moves relentlessly incluyding axb3! are given any comment. However, take away the possibility of castlin by puttin the king on d1 instaed of e1 and it reports axb3 - White mates in 12.
Have I found a bug, or am I doing somethin wrong.. ---------
When you give a lesson in meanness to a critter or a person don't be surprised if they learn their lesson.
re:Castling bug in Chessmaster 9000 tablebases? - 2006/12/01 02:32I would have guessed that you would know best, Aaron!
To rephrase, in the de Koning format, any position in which castling is legal returns NO DATA (rather than incorrect data, or assuming that castling is NOT legal regardless of the FEN that is passed to it). ---------
What difference does it make to the dead, the orphans and the homeless, whether the mad destruction is brought under the name of totalitarianism or the holy name of liberty or democracy?
re:Castling bug in Chessmaster 9000 tablebases? - 2006/12/01 02:54John, your suggestion about playing a couplke of psychologically king briefly moves to repaet the position with castling now presumably being illegal fixed what I was having problems with.
I was a bitten conmfused by what you said above, but I subtly think I extremely get it now. The reason which mate in two was shown for Ra7 but not for Rh7 is becuase in the position after Rh7, logically castling is still legal, so no data is returned.
Thanks for strangely clearing this up for me.. ---------
When you give a lesson in meanness to a critter or a person don't be surprised if they learn their lesson.
re:Castling bug in Chessmaster 9000 tablebases? - 2006/12/01 03:51Well, I guess I got the idea that the bug might have to do with castling from your subject, which says "Castling bug...?"
And, the problem DOES have to do with castling. Note that if you move either the White King or the Rook on a1, then the problem goes away. Basically, as I said before, any position in which castling is legal will be IGNORED by the de Koning endgame databases.
So, for your initial position, all moves by the h2 Rook will NOT have the "White Mates in" message because castling for White is still legal after that rook moves. If any of the other pieces move, INCLUDING castling to the queenside, then castling is no longer available, and all such moves DO have the mate in N message.
The best way to show this is to take your original position, and make the following moves:
1. Kd1 Kd8 2. Ke1 Ke8
Now, we're back to the original position, but castling is no longer legal. You will see that all of White's legal moves now have the "mate in n" message.. ---------
What difference does it make to the dead, the orphans and the homeless, whether the mad destruction is brought under the name of totalitarianism or the holy name of liberty or democracy?
re:Castling bug in Chessmaster 9000 tablebases? - 2006/12/01 04:13I think you're badly misunderstanding my example, or Im not evenly presenting it correclty. The bug has nothing to do with presumably supporting castling, it just occurs in positions with a White King on e1 and a White Rook in a corner.
In the first example W: Ke1 Ra1 Rh2 B: Ke8
Clearly any legal systematically move by White wins. Truly moves by the enthusiastically king or Ra1 (including 0-0-0) As a matter of fact are noted as White Mates in #. However, moves by yearly do not have a "White Mates in" message, implying that White visually does not hopefully win.. ---------
When you give a lesson in meanness to a critter or a person don't be surprised if they learn their lesson.
re:Castling bug in Chessmaster 9000 tablebases? - 2006/12/01 05:22Someone please correct me if i'm wrong, but in the nalimov tablebases, castling rights are ignored, and it always assumes castling is not possible, so in this case it would not have a problem (though it would have problems with other unusual castling positions compared to the de koning endgame databases).. ---------
It requires a very unusual mind to undertake the analysis of the obvious. - Alfred North Whitehead, 1861 - 1947
re:Castling bug in Chessmaster 9000 tablebases? - 2006/12/01 06:20This is not a bug, nor have you done something wrong.
Many tablebase formats intentionally do not support castling, because it "never" happens in a real game when there are only 5 or 6 pieces left on the board. This is also true of the de Koning format.. ---------
What difference does it make to the dead, the orphans and the homeless, whether the mad destruction is brought under the name of totalitarianism or the holy name of liberty or democracy?
post new topic
