stetagy to defeat comps... specificly

I have seen Nacurama Hikura destroy some comps, how does he do it???? what is the strategy to beat comps?

i would say Karate, i heard glass cant harm them. Iron fists.

I think it rests in understanding what kinds of positions they're poor at analyzing, or where they're prone to misjudging an imbalance and steering a game towards one of those positions to try to gain an advantage you can milk for the remainder of the game.

The Horizon Effect is a good place to start learning about the limitations ... though advances in engine design are trying to mitigate this one-and-only weakness.

thats pretty interesting.

I'm not sure it's the "one-and-only" weakness.  More strategic positional considerations are difficult for most if not all engines, which rely solely on brute force tactics, which is where inaccurate position assessments tend to occur.

There are also considerations for how engines use time, but with faster and faster hardware and better and better pruning algorithms I have to think that our ability to manipulate this in any meaningful fashion would come more at the expense of our game than the computer's.

Yeah, when you see Naka's games, he basically closes all the position and waits for the computer to make something stupid by fear of a draw.

Agreed ... it's not accurate to group everything into a "oh its the horizon effect" bundle.  Inaccurate evaluations can also make the beast do poorly. Specially when the time controls are shorter. (The Naka games were mostly blitz as I recall?)

Though honestly, have you recently seen a Man vs. Machine game at slow time controls where the latest-and-greatest engine messed up in the manner you described? It's a rarity post-2006 (Kramnik - Deep Fritz) where Kramnik did employ anti-computer strategies to pull off a few draws.

what were those "anti-computer strategies"?

Yep, if you make the point of your play beyond their horizon (opening play, positional play in general, endgame play) and avoid the short them tactics, then they'll just have a worse position.

Of course that's easier said than done heh.

Get a closed position and win with an attack or endgame.  Both involves plans beyond the horizon of its calculation so the computer is literally blind to the possibility of losing until it's too late heh.

I'm no expert here, but I'm pretty sure it's suicidal to try playing highly tacticall and/or questionable openings vs PCs. According to advice from masters that I've read, try to keep the game as positional as possible. Click on these links for more info:

http://www.stumbleupon.com/su/1Dhrla/homepage.ntlworld.com/adam.bozon/computers.htm/

http://www.chesslines.com/tips-to-beating-chess-computers/

thanks for the links!

http://www.stumbleupon.com/su/1Dhrla/homepage.ntlworld.com/adam.bozon/computers.htm/ was especialy helpful.

The old strategies don't really work much against today's computers.

The best bets are to get the computer to play an opening that it has completely misevaluated, such as the King's Indian, then win with a grand strategy rather than tactics.

One thing that Kasparov and Kramnik both discovered is that computers are much stronger than humans when Queens are on the board. So trade off Queens as early as possible. The last thing you want is to play Q+Pawn ending against an engine. 

Beyond that, pawn storms are still extremely difficult to caclulate, which is why the King's Indian remains a good choice against an engine from either side. The engines are unlikely to find the best moves for Black regardless of which side they are on. 

It's also possible to find positions where an exchange sacrifice is winning. But the key in all of these is to do a great deal of research before you play the engine. Do some heavy work before you even start. Find a line that the computer is likely to play but that has a serious flaw. 

As Pfren has pointed out numerous times in other threads, computers are really quite weak in complex endgames. If you can survive to the endgame, that is where your best chances lie.  (With 6 pieces or fewer, they are quite literally, perfect.)