A chess problem that computers can't solve.

OK, good to know. 

Try this puzzle
(computer can solve)

We do not even need a composition to prove that computers are very limited. After his game against Hikaru Nakamura in the London Chess Classic 2016  (www.chessgames.com/perl/chessgame?gid=1849553), Fabiano Caruana declared that in his preparation his computer did not see 21. Nf5! because it puts too much importance on material balance. So his brilliant queen offer (19. Qxf6) not only beat Nakamura but also the computer

That win thing is just stupid. If the opponent also plays on your side, then you may win. If they play like an average Chess player you'll still lose.

Correction: By losing I meant not winning, of course it's an easy draw (moving on white squares with a king).

The win thing is stupid but even if they play like a top GM you won't lose,it's an easy draw.

Correct me if I'm wrong, but a draw by agreement isn't what the article is going for. And for the outcome to be a "true" stalemate, white needs to have no possible moves left meaning he has to take the rooks with his pawns. So I'm assuming you get your king into a spot where when the queen jumps your pawn or something it leads to a stalemate?

Yeah you're right. it was just to show how stupid computers can be. It is of course an easy draw but people wanted to see if they could force a quick stalemate by trapping their own king. The pawns could still move of course and take the rooks but then they would be captured or trapped leaving the stalemate but there wasn't one anyway. ( at least I can't find one )

if the queen ends up on a4, its a stalemate.

if queen doesnt, and you get your king inside of blacks pawn cluster (with the queen leaving to try and get checkmate) that makes a stalemate scenario a lot easier.

Idk, ive only looked at it for 5ish min and i see lots of ways that white could end up in stalemate, just not sure if there is a "forced" stalemate

There are possible stalemates but the only forced one is the 50 moves

Solved 1st time using mobile phone?  Then if 3) .... Qe1+ then M in 18, 4) Kh2.

Or if 3) ....c4 then M in 20, 4) Rd1.

How do you get to " 3) .... Qe1+" the Q is trapped??

Sorry I was looking at #44.  I'll look at #1.

I skim-read the article and found no indication that it calls for a stalemate. Draw by 50 moves rule or repetition is perfectly fine.

Edit: I think the whole point of the position is that humans can (and Stockfish can't) see that if White just moves around the King on the light squares, Black has nothing but moving Bishops on the dark squares, and even if the 50 moves rule didn't exist, there would eventually be a repetition because there are only finitely many options to arrange these four pieces.

And of course White could win if Black was stupid enough to let go control of c7 and White moved the King between c8 and d7.

"Chess engines consider this to be a win for Black. For example, Stockfish 8 still evaluates it at -28 after five minutes of thinking" - this is a problem of how the author interprets engine evaluation. For me "engines consider this to be a win" is when evaluation says "Mate in N".

If you let the engine play this position as white he will achieve an easy draw just as a human will do. In no way this is "a position computers struggle with" as this article tries to claim.

As I see it the primary purpose of evaluation function is not to say who is going to win, not to predict an outcome of the game, but to compare relative values of the candidate moves and choose the best one. Then of course evaluation function is calibrated so that its absolute value means something for us: 0.0 means about equal, and +1.0 means about 1 pawn advantage. But that's approximation. When we interpret evaluation value as a hardcore prediction we can only blame ourselves.

No forced mate from initial position, except for eventual stalemate, draw, be it by repetition, 50 move or agreement.

White wins if black allows c6 through c7 to c8Q#, unlikely.

Black wins if it takes c7, then takes b2 with K or Q, again unlikely.

 WHite can achieve a win with a helpmate.

I have seen this kind of problem before. In the My diagram below, is a dead draw, no matter what happens. Still it shows that black is winning? .. The computer says -11.39,. But humans can see that it's a forced draw(50 move rule), once they understand the position. It has to do with the algorithms that were put in the chess engine. By that I mean, there's a code missing for closed games in chess engines.

I meant takes a2 not b2.

Hi Rasta_Jay #59

I agree with you, but could not set up as 'too many black pieces'.

I removed 1 black bishop giving -7.82  on material.

if black is stupid enough to move all three bishops off the line, white might be able to promote and win