Nc5

I'm interesting in seeing the position from time to time, could someone post a board with the game?

Go to the first page.

yeah it even gives houdini's eval.

Oh, ok :)

I don't really get Rfd8 by Houdini, the rook has no more potential on this square than it did on f8. In fact I think it has less.

Bb5!!!!

HOUDINI PLAYS NB4! VOTING NOW OPEN

I guess it left the option open for Nb5 in case white plays Bb5 (as in the game) ?  I know what you mean though, sometimes computer rook moves look pretty odd.

odd rook moves are to be feared, they may include strong prohylactic ideas. Or the lack or any other better moves :P

I'd rather think the engine doesn't know what it's doing, but now it seems clear that it was just defending the bishop.

In closed positions I've seen houdini make more or less random rook moves though.  Some middle game algorithm telling them to go to d8, then the next move c8, then a8, then back to d8.

wow thats odd. did Houdini win? maybe he was daring his opponent to overextend...

I was analysing a position, and letting houdini play the moves.  It can't "see" the plan, so it just "improves" it's pieces over and over.  Sometimes the moves are good prophylactic moves.  And if there is a pawn break or some kind of idea, it usually eventually plays it.  Of course after the position opens up, it's tactically extremely well prepared... but sometimes it didn't make a difference if the rook was on a8 or b8 heh.

lol I wonder if there is a way to take advantage of that

Hmm, only if you can more or less trick it into a position where it thinks it's better for some reason, but you have the practical chances.

Then proceed to play nearly flawlessly to prove it

The rook moves themselves I don't think you can take advantage of, just try to keep improving your pieces as well.

And although it's not as common an idea in a middlegame, there's definitely the idea of "doing as little as you can, as slow as you can" in endgames to throw your opponent off when you have a risk free edge to play with.

I wonder if the world's strongest players could beat houdini on a strong hardware at the strongest level... when was the last man vs machine mactch?

I've been told in ICCF, some players just use computers.  And apparently those players never rise above mid level because centaurs are always stronger.

So I'd have to assume that a strong player hand picking lines for houdini would beat houdini all alone.  Not the tactical positions of course, but those positions where the top 3 (or even top 8) moves are only a few centipawns difference.

Here's a tip,  sometimes when programmers make a change in an engine, and test the change by playing a few hundred or thousand games, if the overall winning goes up, then they keep the change... even if that means there are some positions it wont pick the best move anymore... this is to say that if two moves are evaluated differently by a tenth of a pawn, it may not mean anything if the position is closed or otherwise calls for long range thinking, like an endgame.

And you can notice this all the time when you play a few moves out in your computer's favorite line, and a few moves down it "changes it's mind"  And now that that position is in its memory, go back to the position you were looking at and notice how there's a different best move suggested :)

good points waffle. 

Another thing I was wondering, can computers make plans?

AI community would be thrilled at a program that plans :)

It just calculates a few zillion positions, and weighs the end positions with its formula.  The position that gets the highest number according to the formula gets chosen as best.  It's just an input output tool, the computer isn't really thinking in the way we think of thinking (heh, funny sentence).

But it will do logical human type moves such as preparing a pawn break or organizing it's pieces and conducting an attack, or even trading off an important piece and setting up a blockade.

But only when such maneuvers are within its search depth.  Most middlegame positions contain enough tactics that they're incredibly strong.  Most endgames are enough about overall strategy that computers do better to rely on EGTBs.  Also openings, at least the first few moves, often contain no tactics, so computers rely on opening books.