A very hard nut to crack

pfren

Here is an extremely difficult endgame study, made by the Yugoslav composer Dragutin Djaja. White to play and draw. The solution is amazing- even computers take a very long time to discover white's idea.

 

iCloud

0mg, I'm stuck after 3...Ra4, xD Wow :]

asdfv

wtf, how is that stalemate

Justified08

:]

unityjeauxx

why doesn't white just play Nxd4 followed by Nc2 to stop the pawn? Seems like that would be a winning position although I haven't analyzed it thoroughly

pfren

2.Nxd4 a2 3.Nc2 loses at once to 3...Rb2.

musiclife

Wow, pretty sweet how the Knight guards the f5 square which indirectly keeps the g5 square safe.  Now I see why the K returns to h2 after kicking the Rh4+ w/ Kg3. 

mrguy888

Wow. Crazy stuff.

shashi85

how is it draw... please somebody tell me....

pfren

Black can play better in that last variation: 8...Kc7, when the a7 pawn is lost. However, white can draw with either 9.Rg6 or the surprising 9.Rb1 (all other rook moves lose). That said, it's extremely difficult to handle the white position without aid (online tablebases, that is). Mind you that when the study was composed, there were no computers, and no online tablebases.

You can check the position at the usual place:

http://www.k4it.de/index.php?topic=egtb&lang=en

 

The solution given by the composer is a clear draw, as well as an "easy" one- provided of course that someone finds 4.Rg8!! and 5.Nh6!!, which pop up at the Houdini moves list after just a couple of hours thought... Laughing

PrawnEatsPrawn

The solution given by the composer is a clear draw, as well as an "easy" one- provided of course that someone finds 4.Rg8!! and 5.Nh6!!, which pop up at the Houdini moves list after just a couple of hours thought... Laughing

 

I thought that maybe you were mistaken but Houdini seems to have no idea what is going on. Even when led up to the perpetual position, the latest Hoiudini on extreme hardware doesn't grasp the concept (I didn't bother to run for hours).

 

Thanks for showing this interesting position.

 

p.s. Although Houdini finds choumicha's draw in less than two seconds.

pfren

Houdini cannot compute the other draw, unless used under an UI which is using tablebases. It just gives an evaluation based on the material balance on the board, and the casual.

As for the engine's ability to find the composer's solution... look at these:

http://www.chessbase.com/newsdetail.asp?newsid=5634

http://www.chessbase.com/newsdetail.asp?newsid=5695

Of course the time mentioned is for 5.Nh6 alone, not for solving the whole study.

PrawnEatsPrawn

Your post clearly shows the benefits of letting more than one engine take a look at tricky positions.

 

Stockfish found the (a) solution fairly quickly: depth 32 in 150 seconds.

 

 

Justified08

how is ti  draw?


Surprised
pfren
Justified08 wrote:

 

how is ti  draw?

 

 



White threatens checking all the way from g8 to g1, and Black has nowhere to hide without allowing the a-pawn to queen.

marcofuics

a question: could white win the game?

RoffleMyWafflez

Way over my head.

I got the first two moves just by inituitively trying to stir up counterplay, but wasn't able to see the perpetual.  

CHCL

Nice.

p-wnfall7

kewl..

warrior689

wow!!!!!