Positions engines get wrong ( please contribute )

Sorry, I didn't notice it first.It's not from a real game, it is a study by Chekhover.

I don't know about today's best engines, and I can check it out (because I don't have Houdini), but some years ago engines thought that endgames like this were won. LOL

Even Houdini with its poor endgame knowledge thinks that it is only +0.02. Stockfish and Rybka zero out without hesitation.

Ok ! Nice information !! So, the conclusion of this topic might be : Don't rely on computers every time !! LOL

The posted bishop vs. rook endgame is a real mistery to engines. If I play pfren's solution and build the fortress as suggested even then Stockfish happily moves the rook back and forth believing that the position is heavily winning ( +4.56 for black ):

Computers sometimes need some time to find the best move! Also, sometimes you see the symbol =, you play some of the suggested moves and sudenly you see +- 1,4 !!

What a Tactics Trainer rating you have! Unbelievable...

Note all those 3000 tactics trained rated. Obviously cheaters

Now I see the position which was posted, with the rook vs bishop endgames! There are some other positions like this, an example might be the following: In the final position, fritz 13 gives -+ 13.00 !!!  LOL

Howewer, many engines think the Spassky-Fischer endgame (rook vs 5 pawns) is won by black

Oh my god !! There are people with 3300 !! And I am trying to go to 2200 !! LOL

http://www.chess.com/forum/view/game-analysis/please-analyze7

I came across this post where the losing player is accusing the other of cheating. I'd like to know if any chess engine would play the moves 26. Nf4 and 27. Nxg6.  My first response at 26. is b4. 

As I said before: A stable positive evaluation by an engine does not mean that the position is winning. It means "this is most likely drawn by 50 move-rule". Houdini 3 even has an option in it's settings that makes the engine believe the 50-move-rule is a 10-move-rule for example. If you use this option to analyze the positions in question it will label all of them as 0.00 relatively quickly. 

Games like the Shirov-Game where the engine dismisses the winning move, on the other hand, are a nice example of the circumstance that engines still are very far away from perfect play. In this case their pruning makes them overlook the winning move. But let's be honest: How many human players would find such a move without massive in-depth analysis?

6k1/5n2/8/8/8/5n2/1RK5/1N6 w - - 0 1

White to move, mate in 262. I doubt any engine can see that deep.

All engines find the first move 1. Kd3 but none the second 2. Kd4. They play 2. Nc3 or 2. Nd2.

Imagine the look on everyone's faces when the GM proclaims, "mate in 261!" 

That position would be draw anyway(50 move rule).

It is unfortunate that players can claim a draw after 50 moves even when their opponent has a forced mate.

At least all engines agree that the French is the best defense and the Sicilian is an inaccuracy

The rule was introduced so that a chess game doesn't run to infinity.That's a sad part for players who have a winning position.In a recent National Championship in my country, white played accurate moves in the endgame and had a forced mate in two moves, but a draw was claimed because 50 moves had passed since the last piece was exchanged (a painful day for the white player).