Endgame statistics

does anyone know where to find endgame statistics on win/draw/loss for specific endgame configurations?  For example rook vs pawn.  There are a finite number of unique positions with that material, some are a win for the pawn, more are a draw, and most are a win for the rook.  Is there some way to use tablebases to find this?  Or is this online somewhere?  

You can probably use a sufficiently large database to pull some numbers from with a material search. Not sure if something already exists.

Here's one I used and liked on one study of mine:

http://www.shredderchess.com/online-chess/online-databases/endgame-database.html

http://scid.sourceforge.net/tbstats.html

I have the tablebases already on my computer.  Is there someway to query them directly based on material??

That's getting closer to what i'm looking for...

this shows that for KR-KP  there 1102 unique positions...

but is there someway to find what the win/draw/loss is for each of those 1102 positions?  

actually I read that closer, that is not unique number of possible positions.  That shows "the number of database games that reached a position with exactly that material"

and yes I have already tried doing this, but the tablesbases are not in a format I can open with chessbase (.emd and .md5)

This almost answers the question -- http://ilk.uvt.nl/icga/games/chess/endgames.php

This breaks out each type of endgame by percentage W/D/L based on turn to move and other things (distance to mate, etc), but it still doesn't show the number of unique positions (by FEN or GBR ideally).

Looking just at rook vs pawn (KRKP):

This might get you some of what you want, though I can't verify the trustworthiness of the site/files:

http://chess.jaet.org/endings/docs/

in particular the 3-5 Man Results and Maximals file on that page. Of course, that is partially limited since it doesn't include 6-piece tablebases.

I realized after I posted earlier, what I was suggesting really doesn't give you results with perfect play.

Yeah that's very similar to what I found...almost there...still searching...

Here is a small post/discussion on it:

http://chess.stackexchange.com/questions/6246/how-many-positions-are-there-in-n-pieces-endgame-tablebases-for-different-values

I don't think the tablebases will include anything about it due to cutting the space down due to symmetry.

funny because I was just reading the link in the article you just linked:

http://kirill-kryukov.com/chess/discussion-board/viewtopic.php?f=6&t=6639

I registered for an account on his website to figure out how to do this.

http://chess-db.com/public/research/endgame_statistics.html

That shows the likelihood of each kind of endgame - not the number of unique positions for each ending or the w/d/l. I think Martin has pointed me in the right direction and I hope to find the answer shortly

Ok getting closer to the truth!!  I found the answer to my first question, but not the second one.

The number of unique legal positions (NULP) for each kind of endgame can be found here:  

http://kirill-kryukov.com/chess/nulp/results.html

For king and rook vs king and pawn:  krkp - 9,041,298

I am still searching for W/D/L results for each of those positions, based on perfect play. The other results posted above for W/D/L used a database of actual games which means the play was not perfect.

http://www.mathematik.uni-stuttgart.de/~thiel/endgames/