Weird analysis by engines

I played a game against the AI in Chessmaster 11, and got interested in this "blunder" I made, supposedly my worst move of the game.

Here is the position, it's black to move. The engine analysis is 0,00.

I can't see anything wrong with the position, so I opened the PGN in Arena and let Stockfish analyze it, too. The same thing happened! Moves before and after are +10 to +15 for white, but this position is 0,00. On two different engines, that is.

Am I missing something in the position, or what is happening? Is there something wrong with the engines?

The only thing I can think of is tri-fold repetition, but it isn't, and the game continued normally.

Thanks for any ideas.

You set the position up upside down in your computer.

Think how much better black would do with the move and a pawn about to queen on e2.

Looks pretty 0.00 to me too.

 Thanks for the idea, but that is not the case.

Here is the full game, the position in question is after white's move 37.

Please don't laugh at my play, I'm a complete beginner. Maybe someone could throw this into their program and let some engines analyze the position after white's move 37. I know it's not the best move, but flat even is a bit harsh.

The fact that you are able to see that the computer evaluation is ridiculous is a good thing. You are up a rook and a bishop and of course it is not a flat draw.

My guess is that you are coming up against the limits of your computer's memory. Endgames are very difficult for computers since the ability to see many moves is suddenly at a premium.

Engines usually evaluate a repetition as 0.00. It saw the position after black's 35th and 37th move are the same.

Yeah, I see now that black did repeat the position by making the move, so I guess it was kind of a bad move from me to give him the chance to repeat even once, even though I did not make any more repetitions.

Now that I think about it, 0,00 seems to be a sensible way to make the engine want to repeat when losing, and avoid repetitions when winning. So whether or not I fall for the repetition "trap", the engine will want to try it when losing, and thus evaluates it as a potentially drawing move. Way better than anything else black has available at this point.

That would also explain why more than one engine did the same thing. It seems to be a common trick of the trade to evaluate all repeats as 0,00.

Thank you TBentley

I had a similar problem with Stockfish recently so I asked on the dev forum. It is indeed a trick of the trade. 

https://groups.google.com/forum/?fromgroups=#!topic/fishcooking/MixMrNso02I