Which Engine Plays Like Human?

Interesting view. I don't have much experience with Hiarcs and Shredder, but to me Rybka Human seems very natural. Maybe it's because I got used to analyzing with it.

Some might feel like a human, but of course no computer really thinks anything like a human. They just look at millions of moves and positions per second with near 100% tactical accuracy (at least for several moves deep) to compensate for not really knowing anything about chess.

That's not exactly true. They do know A LOT about chess. They do have a lot of positional knowledge. They know what to do.

I've never heard they had positional knowledge. Of course they have evaluation systems that generate a value for who is better, but wouldn't the rules it follows for evaluating (like putting in say .5 for the bishop pair, etc) have many exceptions it wouldn't be able to understand in a position with no short term tactics?

That's how I always thought chess computers worked, but I don't know them that well.

Hi, I'm not a big computer chess expert, but I've got sigma chess on my macbook and it plays pretty darn well...

The computer's "brute force" method is certainly very effective when you can calculate that fast. I've heard 99% or more of what they look at is completely irrelevant and stuff a good human player wouldn't consider, but despite how inefficient that is it just doesn't matter.

Apparently Rybka has some positional understanding due to the efforts of Larry Kauffman. That is supposed to be the reason that it is better than other programs that can calculate just as well. That is a fairly new thing, but still I suspect it relies very heavily on calculation.

How do you program positional understanding though?

 you can't

I don't know enough about programming to tell how, but I think it's probably possible. For example, a weak field is a field that is not controlled by the pieces or pawns. I think that it's possible to tell an engine to count how many pieces or pawns control a field. In those lines.

As an example, how do engines understand that the King is a crucially important piece ? They don't understand it in these terms, but they 'know' that the material value of the King is higher than all the other pieces taken together. A checkmate would naturally result in a win because more than 50 percent of the material that is on the board has been taken. Whether this is understanding or not is somewhat debatable.

You feed it with the knowledge. You punish a Nc3 move in a d4 opening before the c pawn has been moved with 0.02, you give 0.10 for open files occupied with rooks etc, that sort of thing. Of course, probably much more complicated than that, but you get the idea.

I don't think that's true. Brute force is a thing of the 80s. The best engines are the ones with the best pruning (cutting off non-promising lines).

Why is your handle philidor_position. There isn't one game in your archive where you played the philidor defense.  You played the Russian game in response to e4 virtually every single time.  Just curious.

Yes, to my understanding they make very small adjustments like this and then play something like 50,000 1 second games and analyse if there was an improvement or not.  If there was a net improvement supposedly Vas implemented the change weather it made the engine play worse in other positions or not.

philidor position refers to something different than the philidor defense. There are two philidor positions, both are theoretical endgame positions.

the one I knew when choosing the this handle was the one where the defending side could draw with a rook & king against a pawn & rook & king. Later I learned there was another one with a queen vs rook ending.

Philidor's position is an endgame position.

Interesting. The book was from 1996, so for computer chess it's probably completely outdated. But then computers might have the same problem us humans do, not considering a move that could turn out to be a brilliant, winning move(of course eliminating stuff is a must for humans, but for computers not as much). That happens sometimes, where the computer finds a very deep solution often starting off with a rediculous move that not even most grandmasters would consider.

There's another endgame philidor? What does that one look like?

There's quite a few famous endgame philidor positions/techniques.