engine endgame stuff

[Aljoge]

I am new here. The reason is that I am interested in learning more about chess software, because it is used for own game analysis. Well: I use now Arena, working on ubuntu using wine, and using arena using windows XP. I just have a 32 bit computer. For several years I used crafty on a ubuntu-linux system. Now the step to Arena is made and an own database was created including 1 million of games.

This month one of my teammates posted an interesting article in our clubpaper. The statement was that at his computer the commercial versions of Fritz and Rybka are not finding the correct solutions for certain positions, mainly in the endgame. positions are reported in my blog: http://chesspeers.blogspot.com/

After this result: spike 1.2 will be my engine!!!

Are you familiar with that? How do your engines respond to this?

(hopefully it's according the forum rules)

[mercury]

It's quite well known engines have problems in some endgames. On my site echecs88 I tested several engines with 180 endgames positions. You can find my rating lists on this page. The best is Rybka 3, you can't say Spike is better just with testing 3 positions.
You have also the page of eigenmann testing 100 endgames positions.

[jdart]

Some of these are in the range of tablebases so programs that support those will solve them instantly. Without tablebases, a number of 5- and 6- man endgames are difficult and can require deep search to solve.

--Jon

[Mike S.]

The positions #1 and #3 are very familiar, but they are entirely outdated as computer chess test positions. They are from the era of chess computers. I am sure that almost all of these engines mentioned, solved 1 and 3 correctly even without endgame tablebases support.

If you are interested in comps + endgame analysis, you need to know Nalimov tablebases. Google for it to find the basics. They are free to download and 3+4-piece tbs. are ~30 MB only, and will help to solve these positions and many others with few pieces, by engines which can access them in the search (like Ruffian, Rybka or Spike mentioned on your blog, and many others).

See also site like

http://www.shredderchess.com/online-che ... abase.html

where you can access Nalimovs online and check positions up to 6 men.

So basically, this test consisting from 3 positions only (which is of course MUCH too small to judge) is testing position #3, which is a common problem, but not one of very much practical relevance...

Spike 1.2 is a good engine. - But from my tests and impressions, I can recommend for the endgame e.g. Bright 0.4a or Stockfish 1.5.1, which are stronger although they do not access the Nalimovs (Bright can access the Scorpio bitbases but if you are new to this configuration issues... it will also do ok without). Both of these engine support multicore processing, unlike Spike where the public version is limited to single core.

Both are freeware UCI engines. - I do not know though, if they will run in Wine.

http://members.ziggo.nl/allard.siemelink/bright/

http://homepages.tesco.net/henry.ablett/jims.html
(scroll down for the Stockfish executeable)

Another recommendable alternative is the free Rybka 2.2n2, the unrestricted Rybka demo available at Rybkachess.com. This version was before some elements of the endgame knowledge were removed from later versions. This engine can use the Nalimov tablebases with adjustable access intensity. (I'm not sure if the version 2.2 you mention already, is the same.)

P.S. For a general strength ranking of chess engines, see e.g.

http://computerchess.org.uk/ccrl/4040/
(the green and red engines are free)

http://www.husvankempen.de/nunn/

[royb]

Spike 1.2 is a good engine. - But from my tests and impressions, I can recommend for the endgame e.g. Bright 0.4a or Stockfish 1.5.1, which are stronger although they do not access the Nalimovs (Bright can access the Scorpio bitbases but if you are new to this configuration issues... it will also do ok without). Both of these engine support multicore processing, unlike Spike where the public version is limited to single core.

Both are freeware UCI engines. - I do not know though, if they will run in Wine.

Bright and Stockfish both run fine in wine. Stockfish also compiles easily (no modifications needed to the Makefile) on Linux to produce a native Linux 32-bit or 64-bit binary if your chess GUI can handle that. I use SCID for analysis myself which makes use of both Windows engines via wine or native Linux binaries for other engines that compile natively on Linux.

[Aljoge]

Tnx for the help. Next tuesday I have free time and figure this out. Mainly the bright and stockfish engines look interesting.