Stockfish tells Anand was in a winning position when he drew it

Engines play way different from humans. They generally focus on increasing the score of the position without an appropriate plan or reason.

nice

I asked H4 to play out from this position:

8/6p1/3k1p1p/3r4/1p2K3/3P3P/1R4P1/8 b - - 1 43

Did I get it wrong?  If I did, my abject apologies.  I can post the scores from Houdini winning with black if that helps.  I don't have a horse in this, frankly, I WANT humans to be smarter than the engines!

Yes, since Ke4 was not played. It would have been a horrible blunder but once such a move is registrered by the Arbiter placing the king on e4 to indicate the game was drawn it is usually impossible to make people believe that it wasn't played :-)

Magnus would have played it out. And Magnus might have won. That's why he's better than everyone else.

you are talking too much about stockfish.

if u think it is child then y can't u beat it.

Actually, the post you quote is a humorous comment about the endgame abilities of those engines.

Houdini in endgames is the only modern engine I have ever seen make mistakes.

It's not humorous at all... factly, it's the reality.

Houdini playes the endgames like crap (especially the rook endgames), Stockfish is only marginally better, and Komodo, while clearly better than these two, still is unable to evaluate properly many endgame positions that are crystal clear even to players below the FM/IM level.

This is no surprise to anyone who knows how an engine is working- it's probably a surprise to any casual player who thinks that engines play "perfectly".

Correct me if im wrong, but dont chess engines calculate brute force taactics?  And if none exist, the will often just make a random move.  Wouldnt this explain chess engines weak end game performance?

Although material gain and mating searches are what they do better/faster/deeper than human, they also evaluate each position they encounter for known positional advantages.  For instance, they will look at king safety, piece protection, rooks on open files, etc. and give subtle values to these on all positions they evaluate.  The values that they give are tuned to databases and learning algorithms over other data to be good predictors of the value those positional elements give to winning a game.  All engines are basically equivalent in their ability to search for material advantage and mate, but it is this positional valuation that makes different engines unique.

It is also that positional evaluation that makes engines fairly poor at the "complex" endings (5 or more pieces, say).  Even though there are known strategies to simplify a number of mildly complex endgames, and strategy books are filled with the tricks that will bring about your best result, engines just look at positional cues that are almost completely irrelevant in the endgame.  For instance, a rook is almost always on an open file near the end, and it really doesn't matter.  Usually what matters is whether the rook is tied to the defense of a piece/pawn or has freedom of movement, which is entirely different as it may occur on a rank or a file.  Similarly, knight are often outpost squares as no pawn may challenge it, but again that doesn't say whether it is useful or not.  Mate can be 40+ moves away but be pretty easy to see if you know how to ignore all the irrelevant information and don't need to just blindly search all possible moves.

No. Modern engines never do brute force calculations- this would cost a lot of time, even in very simple positions. They use advanced pruning techniques, but that "advanced" is a very vague term, and quite dependent to the engine programmer's skill.

I could write thousands of words about this, but for the moment, I prefer staying on topic.

Thank You pfren...if by "thousands of words" you were referring to ex0du5's post.  Yea...i read it.  It was alo of words, not desrcibing anything with the added benefit of not answering my question.

I apologise, I_Am_Second.  I had not realised you were mentally handicapped and unable to understand basic communication.  I am sorry that those words hurt your feeble brain.

Brute force tactics are the material gain and mating searches I mention right up front, pointing out that contrary to your weak grasp of reality, engines do much more.  I then described that "much more", which is also the correct explanation of why engines are not good at endgames, directly addressing the next part quoted.

I responded to every sentence of yours I quoted.  Clearly, though, you are not someone who appreciates insight or thought and just wanted a slap on the back and ignorant approval of your misguided view of the world.  In the future, I will remember your dullness and avoid throwing too many words at your tl;dr little comprehension.

His thousand of words probably refers to that kind of stuff. Navigating through the links in "other algorithms" gives you an idea of how much improvement there can be over brute force.

Apology accepted