FENs to test engine endgame knowledge?

[xmas79]

Hi,
do you have a few FEN strings of "light" endgames to test engine endgame knowledge?
I implemented basic KPK endgame rules, KNBK mating tables, flat draws recognition etc... and wanted to test if everything is really OK. The very basic positions I tested works well, but I want to know if engine chooses to enter a drawn endgame by exchanging pieces when in disadvantage etc...

best regards,
Natale.

[velmarin]

Here I have something,
otherwise you serve.

Also you can make up your positions, it is sometimes more productive.
http://www.mediafire.com/download/olkmu1abm8vpzj9/TB.7z

[xmas79]

I tried some positions myself, and discovered some problems already (e.g. giving big bonus to pawn advance will prevent engine to promote in a simple KPK endgame).

So now I have some FENs... I don't have a PGN parser! lol!!!

Thank you very much,
Natale.

[velmarin]

For PGN you can use ARENA in windows



TB.epd
http://www.mediafire.com/?ofte1wdzl3fvon9

[Steve Maughan]

Thomas Petzke has a set on his site. See here:

http://www.fam-petzke.de/cp_download_en.shtml

Best,

Steve

[xmas79]

OK. Is there a time limit for the engine to find the best move? Do I have some way to automate this testing or I have to write my own code? e.g. Arena will "control" engine by sending new positions taken from epd file and waits until engine moves, and finally gives me statistics?

Thanks,
Natale.

[jdart]

Arena will "control" engine by sending new positions taken from epd file and waits until engine moves, and finally gives me statistics?

Fritz does this - not sure about Arena.

My engine has implemented an internal "test" command that processes EPD files. It is not hard to parse them.

--Jon

[tpetzke]

Hi Natale, in my sets actually there are no best moves. It is a evaluation test. The positions are classified in 3 sets (Wins, Losses and Draws).

Just give the position to your eval and look what it says. If it assigns a winning score for a won position everything is good, if it assigns a winning score for drawn/lost position you might want to have a look.

As no search is involved the positions are processed really fast.

You can generate an output like this (draws.epd)

1 6N1/8/1K6/3k4/8/8/8/4N3 w - - 0 1 score = cp 0
2 8/8/4K1N1/8/8/2N3k1/8/8 b - - 0 1 score = cp 0
3 6N1/8/8/8/1N6/8/4k3/7K b - - 0 1 score = cp 0
4 8/3N4/8/8/6k1/2N5/8/7K w - - 0 1 score = cp 0
5 5N2/8/8/8/2N4K/8/8/3k4 w - - 0 1 score = cp 0
6 4k3/5N2/8/8/8/6N1/3K4/8 w - - 0 1 score = cp 0
7 8/8/6N1/8/2K5/k7/8/N7 b - - 0 1 score = cp -745
...

The last one is one I would have a look at ...

Thomas...

[sje]

Encyclopedia of Chess Endgames #3

1,797 positions

https://dl.dropboxusercontent.com/u/316 ... D/ECE3.epd

[Steve Maughan]

Hi Thomas,

I hadn't realized these were for static evaluation - nice!

Steve