Master Om wrote on 04/01/10 at 04:15:43:
So please any one Analyze this and tell me who can find the answer.
8/1p6/1Pp2N1q/p1Ppk2p/P3p3/3PPpPp/3K1P1P/1R6 w - - 0
It doesn't seem so difficult:
1.Ng4+ hxg4 2.d4+ Kf5 3.Rh1 Qh8 4.Ke1 Qa8 5.Kf1 Qa6+ 6.Kg1 and the king just shuffles between f1 and g1.
Engines won't find
1.Ng4+, but they will find the defense afterward. Most of them will grossly misevaluate the position, but that's not really a problem if they find the right moves. The score never goes up despite how deeply it looks, meaning that its search cannot find any way to progress. That also means that everything is working correctly, but the engines just do not have any way to properly evaluate the concept of "progress." It gives the otherwise proper evaluation weights to the queen, a few mobility penalties for the rook in the corner, and that's about it.
Very few engines have implemented explicit blocked-pawn code, namely Crafty and maybe Shredder, if I'm not mistaken. They prefer to handle blocked pawns practically by cooking their opening books to avoid them, and giving early evaluation penalties to closing the position. If such situations are never allowed to arise, then they will not have to bother trying to solve their engines' short-sightedness. This anti-human concept may not be so helpful in analysis, but it's the easiest way to deal with it.
Also, it's been a known problem for a long, long time and doesn't really prove anything. The situation can be summed up as... the engine programmers know about it but just don't care. Exploit it if you can, but there's no reason to sacrifice overall strength in every other position with cumbersomely slow evaluation terms just to handle such extreme cases.
Thus as you guessed Shredder is the only engine which can find the answer w/o any help.