GMTonyKosten wrote on 06/06/06 at 09:06:01:

It does as it may screw up the evaluation, eg overvalueing passed pawns may lead a computer to go for a line in which he faulty evaluates the position after move 20.

HgMan wrote on 06/03/06 at 18:33:23:

Not an expert on computerchess myself, but well versed in programming. It would mainly depend on the actual evaluation (and to the amount of plys of course). Eg it gives an IQP an x amount of points, but substracts if blocked or adds when it could be pushed. That said chess programs havent really become much more advanced since the first ones. The evaluation is more sophisticated, but mainly the hardware is what makes them better.

I know computers are data-crunchers that calculate really fast and that the knock against most programs is that their positional understanding remains limited.  But how does that play out in practice?   

Take, for example, a game with an IQP.  While the computer can calculate effectively, it doesn't really know what it should do--at least, it doesn't follow the principles of the position.  Whereas a human knows that in general that the side without the IQP should try to trade off pieces and avoid a kingside attack, the computer plays stronger tactically motivated moves.  But at what point are the two one and the same?  That is to say, if a computer calculates the best move in a given position, how often does it "discover" the positionally soundest move?