7-piece tablebase win in 517 moves

[George Tsavdaris]

I am also not sure about the nightrider rules but I think that it was explained in the programming subforum.

The Nightrider is one of the two most popular fairy chess pieces by chess problemists. It was invented by the famous British chess problem composer T.R. Dawson. The technical name for the piece is Knightrider, but nightrider was already a meaningful English word, making it a more suitable name for the piece.

Movement
The Nightrider can make a move like a Knight, but then can continue to move in the same direction. Thus, it can make one or more successive Knight-leaps, all in the same direction: the spaces visited by all but the last jump must be empty.

Vocabulary: Rider
The Nightrider is a Knight-rider. A rider is a piece that can repeat the same type of leap in a single direction--until it captures a piece or its next leap would be to an occupied space. The Nightrider repeats the Knight's leap in a single direction. It may continue leaping like a Knight in one direction until it captures a piece or is blocked from further leaping. Although the Nightrider isn't used in standard Chess, that game has three riders of its own. The Rook is a Wazir-rider, the Bishop is a Ferz-rider, and the Queen is a compound of these two. As a Wazir-rider, a Rook repeats the single-space orthogonal leap of a Wazir until it is blocked from moving any further. Likewise, a Bishop repeats the single-space diagonal leap of a Ferz until it is blocked.


Look at the amazing Chessvariants.org-PIECECLOPEDIA for all Chess type pieces for more details.

[ernest]

many of the 7-piece wins have not been published yet.

Nice contribution, but some of it is "old" news.

Marc Bourzutschky
Posted: Fri May 26, 2006 2:42 am Post subject: New Endgame Record: 517 moves

[smirobth]

I am putting together a list of the longest wins found in the chess tablebases.

I once did an EGTB for KbmKh (= Bishop + Commoner against Nightrider). The maximum DTC was 308 moves, which is quite exceptional for a 5-men end-game (note that DTM will be significantly longer, as KbmK is not so easily won either). Even the average DTC was 265.5 moves! Nightrides are quite difficult to catch!

Unfortunately I lost the maximin position when the HD of my laptop crashed. I could easily recalculate it, though, if you are interested; that would only take a few minutes.

I really like the way you address endgame play by computers. I think that with the increasing power of hardware ... you could use your system to write appropriate EGTB's on the fly. You would first need a table of win, lose or draw data ...

? Perhaps I don't understand what you are saying. Wouldn't such a "table of win, lose or draw data" in fact already be the bulk of any EGTB? There would be little left (just the distance to conversion info) to write "on the fly", and that information would already have been known in order to create a "table of win, lose or draw data" in the first place.

[M ANSARI]

I am not really sure ... but I would think that a win, lose or draw of a certain position would be much less if all that was needed is just the data of whether it wins loses or draws. Because you only include that this is possible ... you don't have to include every iteration or every possible move to reach the win draw or loss. Dr. Muller I remember, had a special module that would find very quickly the actual best moves once a certain criteria of data has been met. It would be similar to Freezer application where you give it a set of data to start with and then it will find the best play. Let us say you have King against bishop and knight in the center of the board. If no piece is en prise and if there is no stalemate you can immediately say that side A wins. The actual sequence of moves would take dramatically more data to store than simple win loss or draw. So now that the module knows it is a win ... it will use the special set of data to find the win. This type of very limited module calculation would fail miserably as a stand alone engine ... but would be great for finding the sequence of a win in such a situation. That is what I meant.

[Ovyron]

And don't forget its possible Piece representation

[GothicChessInventor]

many of the 7-piece wins have not been published yet.

Nice contribution, but some of it is "old" news.

Marc Bourzutschky
Posted: Fri May 26, 2006 2:42 am Post subject: New Endgame Record: 517 moves

I understand that, but all the websites post only:

1. The FEN diagram and DTM/DTC, or
2. The position and the move list, or
3. The position and the DTM/DTC

I haven't seen a replay utility for all of the positions gathered in one place. So that is what I am building.

[GothicChessInventor]

I am not really sure ... but I would think that a win, lose or draw of a certain position would be much less if all that was needed is just the data of whether it wins loses or draws.

Storing win,loss,draw only requires 2 bits per position.

Binary 00 = draw
Binary 01 = win
Binary 10 = loss

With 2 bits per position, you can store 4 positions in a byte.

So 01010010 = 82 = R if you would look at that byte in a text editor.

But, using s technique called run-length encoding, you can crank your efficiency way up. For example, when I solved the 132 billion endgame positions in checkers for all piece configurations with 8 or fewer pieces with one side having no more than 4 pieces, I stored win-loss-draw info only and compressed the entire database to more then 30 positions PER BYTE.

This was not ".zip" type encoding, it was able to be decompressed at run-time without noticeably slowing down the search.

In chess, this type of data would be very important in endgames such as Queen + Pawn vs. Queen and Rook + Pawn vs. Rook. Think of all the draws in those endgame classes that basically defy an "algorithm." Using the run-length encoding technique, you could have access to this information very quickly, and your program would be able to recognize the wins/draws from a great distance. If it was down a pawn, it could most likely find one path to secure a draw. If it was a pawn up, you would not be able to swindle a draw from it.

[GothicChessInventor]

George your avatar cracks me up! George Bush losing a chess game to a pineapple!!

[S.Taylor]

I am putting together a list of the longest wins found in the chess tablebases. While most of you may have seen some long 6-piece tablebase wins, many of the 7-piece wins have not been published yet.

http://www.gothicchess.com/javascript_8 ... dings.html

The page above shows the position, the line of optimal play with best defense, and a link to a javascript utility to let you play through the ending.

You mean it could not possibly have been achieved in any less, against best defence!?

[S.Taylor]

Amazing !!! I think the 50 move rule should certainly be modified now. This shows how rich chess really is and how clueless we all are about its inner secrets.

I wouldn't be so sensationalistic as to call it "inner secrets"!