Studys where engines wont help

Im interested in chess studies where engines cant find the solution. Do you have some? Post with solution please.

Here is one i created (I saw the drawing idea in another study). In the line 1.bxc6? 5...fxg2! wins for black.

http://de.lichess.org/yyrvOcCY

hello a very beautiful study awesomer, can you please explain this line
where white plays g3 instead of the king queen fork
im in interested in knowing
how black enters the position if white wants to keep the
knight as it protects a8, is it through teh a4 diagonal as well, perhaps an early exchange on b5
thank you for your time
http://en.lichess.org/WgvcecZXZgEp

this is another line i just played but maybe is something like this
where black enters the position
http://en.lichess.org/UrTvHpe0eqZ2
too tired to figure it out and besides I dont have the help of the engine if it involves a queen sacrifice.
maybe with a b pawn push black could enter the position, sacrificing the queen, exchanging for the rook.

something like this seems to be the line
http://en.lichess.org/X6PFEZqsAoLQ
seems machine is clueless about the value of the bishop which
is zero in this game, maybe negative for the stalemate problem

@awesomer

In most of these type of studies engines don't really understand that the position is a draw and by simply calculating material points they think one side is better.

A perfect example would be this postion

http://el.lichess.org/analysis/rkb5/1p6/pBp5/PPPp4/1K1Pp1P1/4Pp1p/5P1P/8_w_-_-#0

#2 in the line 1.g3 (i gave this line but the import mechanism strips variations) black plays 1...cxb5! as you already found out.

If white then takes with rook on b5, black plays Ka6, Qd8 (to protect a5 and d5) then brings his bishop to c6 via f7-e8 (so that Nb6 can not take it), then plays Qe8 then takes Bxa4, then activates his queen and king and advances a-pawn. Probably there are more solutions.

If white takes with a-pawn on b5 then Qe8, attacking b5 and preparing Qa8-a5, then Ka7 to c7, Then a5-a4, then Qa8-a5. The rest should be not difficult.

#5 this is a nice study. I knew this already but not with the pawn on g4.

If you dont mind, i added the solution here:
http://de.lichess.org/FGGdN9QY

@5, @7

Indeed, the engines are still somewhat clueless about that kind of closed or drawn positions.

See below how stockfish 8 is so greedy about material and just gives away the game incredibly easily after that drawn position. I know this is not a full powered engine on lichess but it's still kind of funny.

http://en.lichess.org/gGvuQbCmstXg

And there are even some open positions that engines fail, especially the complicated end game kind of situations but they are improving on that aspect better than the closed positions.

For instance there are a few positions in the following video that most engines fail most of the time if they don't get enough time to calculate. But my stockfish got the right answers after a few failing tries and proved itself eventually :)

www.youtube.com/watch?v=N20YQZ7xmjE

It's an amazing study btw.

Exactly that's the point. When the position is closed and there is no way for either side to make progress, computers dont realize that. They simply stick with their first evaluation after counting the material. On the other hand i think that in a position R+p vs Q where the side with the rook has cut off the enemy king and queen, so the game is drawn a strong computer should be able to understand the position. I don't have stockfish though so someone has to check this :)

#9 yes there are some positions where the engines can still make progress. EG. many endgames with bishops of different colors. Im quite sure a bunch of them can be calculated to be definitely draw using a relatively simple algorithm. Work in this field of chess programming has not even started yet.

#8 thanks for the video link, will watch it next month when my conn is faster again :-)