Has Magnus Identified a flaw in computers?

While listening to several reports of Anand-Carlsen 6, I heard that computers had the game "dead even", until Magnus made one move.  Has magnus identified a flaw in computer logic?  Or did I hear incorrectly?

Computers can be all over the place statistically speaking. I have seen statistics fluctuate insanely with just one move.

A joke was made the other day by Aronian, computers think you are winning until it decides you are loosing. 

Doesn't sound realistic. Since the chess engines check almost all the lines, I don't think they wouldn't be able to see a single move ahead.

Even if this would be the case it still wouldn't mean there is a flaw in computers. There might be a flaw in computers, but the first place to start looking would be in the chess-engine. But, in the past there have been flaws in computers and given a modern cpu has 4.000.000.000 transistors. Yeay, there's probably a flaw here and there. Chances those flaws are in the instructionset though, are pretty slim.

Engines can be blind to certain moves. I have seen a correspondence chess game in which a 2300+ fide master resigned with black. I have checked the position and Houdini valued it at -0.4 for black. How come that you resign in a -0.4 position? Then I made the move for white which was obvious to a human observer (other than Houdini suggested). After a little bit of thinking Houdini changed its mind and concluded that black was heavily losing.

They do not check all the lines, or anywhere near it.

They only have time for a relatively a very small number of lines compared to the zillions that are possible.

People today are waking up to that, and finding the kind of configurations that the engines cannot see over the horizon of.

it is possible that A very intelligent and gifted fellow like Carlsen has cracked to some extent the way computers think in their bitse and Bytesy way.

Thank you to those who replied!

To clarify, when I mentioned computer in my topic, I meant the engine itself.

Indeed if an engine evaluation changes significantly after one move that it did not predict, there is a flaw in logic in the programming somewhere.  It would be interesting to determine how to exploit that.  The future could possibly have computerA vs computerB, where computerA recognizes computerB's weakness, and directs the game that way!

The computer was right, it was a draw if not for some very poor decision making by Anand

i think that the computer was right they are better than magnus

Just put the following position into your Houdini:

The above position is valued by Houdini in single PV mode +0.4 for white yet white is winning by force.

Interesting position for Comp-Analysis.

Stockfish started with +1 but the evaluation changed to +3 when the analysis reached a depth of 42/70.

Ummm...it was not Carlsen's move that changed the evaluation from dead even to losing...it was Anand's, Ra4.  The evaluation was dead on, it was not a mistake that the engine did not "see" Carlsen's winning until that point.  

Carlsen only "won" this game in the sense that he did not make a mistake and waited for Anand to crack, exactly like the previous game, where Anand was also dead even until a move in the mid 40s, Rc1+.

In both games, Anand has played very well right up until he gives it away.  Carlsen is not beating Anand in that sense, by accumulating small positional advantages over time...he is just presenting conundrums to Anand and holding even until Anand plays inaccurately.

@btickler Interesting.  Yes, I could have misheard, or my timing could have been off as to when the computer evaluation changed.  I thought that it was after Carslen had moved that it was announced that the computer eval had went from drawn to better for Carlsen.  It still poses an interesting question as to computer weaknesses, or "blind spots".

I have seen times where the evaluation jumped after a move was made. Often with more depth the engine would have seen it eventually, but in some cases there may have been overly aggressive pruning.

This could be considered a flaw, but to "fix" the flaw would reduce the engine's strength.

If this were a case of the engine not seeing the winning line because of its "event horizon", then what you would more likely see scores like 0.07 Carlsen, 0.11 Anand's move, then 0.41 Carlsen, etc.

That is not what is happening in these games.  It's 0.07 Carlsen and then suddenly 0.41 after Anand's move.

Houdini is very bad in endgames relative to other engines, so that could be the cause of the incident in question. I've seen so many positions where the Houdini3 evaluation in an endgame position is all over the map and not settling down, while Stockfish4 immediately zeros in on the correct evaluation within 5 seconds. (Edit: the point isn't that Stockfish4 can evaluate all positions in seconds, because it can't, but that it seems to always be much superior to Houdini3 in endgames whenever I've compared them.)

It is not the programming which makes computers strong, it is their ability to analyse so many moves so quickly - to number crunch.

But they come nowhere near being able to analyse all available moves.

I am a novice in such matters but I suspect that if you took away a programme's opening book and endgame database Carlsen (and many other strong GMs) would be favourite to beat it. In the matches which take place they have their opening book, that starts them down the right line and their tactical nouse makes them unbeatably strong in the middlegame (unless they can be manoevred into closed positions requiring strategic grasp).

Carlsen's intuitive grasp and avoidance of heavily analysed opening lines may make him a bit more able to avoid falling foul of their tactical strength. But I can't see him having any great interest in testing that.

It would be interesting to know exactly where this happenend, to know if it's due to event horizon issues or pruning issues. If it's the first, I suspect the computer simply wasn't given enough time to evaluate the position.

I think the OP is saying that computers first evaluated the position as a draw, and then right after Magnus made a move, they said that Magnus was winning. So Anand making bad moves isn't the problem here.