Chess vs. Poker vs. Computer

[David Sklansky]

"Perhaps you don't intend to imply that a player rated a lowly 1,000 will be tougher to beat by Fritz than a player rated 2,500 .. but that's what you've implied."

I didn't say that. I said that if the computer was playing its normal strategy against Mr. 1000 and the human expert was betting that it could beat Mr.1000 in fewer moves, it would certainly be favored to win the bet. Because the computer would not leave itself open to attacks that it itself recognizes but the human would ignore many of those theoretical threats to win faster. He would also lose the occasional game that the computer wouldn't.

To try to program a chess computer to beat mediocrioties faster than a human expert 200 points below it would we be a difficult task. To program it to beat ALL players faster including ones who made weak moves early to try to fool it is almost certainly insumountable.

[FortunaMaximus]

Even if those weak plays weren't intentional and the player was capable at just fast a speed of adapting and learning from the computer?

That could show why Chessmaster programs are used by a wide variety of players, even masters.

[David Steele]

To try to program a chess computer to beat mediocrioties faster than a human expert 200 points below it would we be a difficult task. To program it to beat ALL players faster including ones who made weak moves early to try to fool it is almost certainly insurmountable.

One thing I wonder about these kind of statements like "insurmountable", are they an indictment of current technology, because I seriously doubt there are any theoretical computer science reasons that limit this kind of task.

A more practical point, I don't see how a human could fool a computer into thinking it was weak without getting a losing position in the process ( and get quickly crushed ).

Also I kind of doubt the human would have the early win edge you claim, do you even play chess against computers?

Finally it seems that even a simple change to the evaluation function to favor shorter depth lines where there is very little difference in real evaluation, compared to the best line, might be enough to give the computer the fast-mate edge,if is not already superior ( which I suspect ).

D.

[StregaChess]

[ QUOTE ]
I didn't say that. I said that if the computer was playing its normal strategy against Mr. 1000 and the human expert was betting that it could beat Mr.1000 in fewer moves, it would certainly be favored to win the bet.

[/ QUOTE ]

I'm not really sure where you get this idea, computers crush weak players and it depends on how bad the 1000 is, if the 1000 is going to allow e4,bf4,Qf3 and Qxf7 mate, well no computer is going to play that way as its unsound so in that case expert winners faster....
However in most cases with reasonable play the computer would win faster.....

You have no proof to back up any claims that it would be otherwise....

[David Sklansky]

It is my understanding that a computer would never make a move that would give its opponent a forced checkmate regardless of the ingenuity it would require to see it.

My comments stem from that and similar understandings.

[AlienBoy]

I agree - I play chess quite alot, and I have often been able to mate a 1100 and below player in 7 moves (sometimes less) - but only because I veered from a "correct" opening. I don't think a computer would do that, because I don't think a computer would intuitively "know" that such a player would not be able to see the impending mate from bishop/knight.

And I have noticed that most computer chess programs choke when you go outside a "normal" opening line.


I believe the same issues are as if not more true in poker. Call it intuition perhaps - and I for one have never been able to reduce intuition to an If/Then statement.

AB

[David Steele]

And I have noticed that most computer chess programs choke when you go outside a "normal" opening line.

FALSE

You should upgrade from the PDP11 running Greenblat.

[StregaChess]

[ QUOTE ]
It is my understanding that a computer would never make a move that would give its opponent a forced checkmate regardless of the ingenuity it would require to see it.

My comments stem from that and similar understandings.

[/ QUOTE ]

That's true, however it would almost never come into play. A computer or a strong human would never have to worry about a weak player having a "possible" checkmate, it's just not very likely....

[David Sklansky]

But the same considerations are used to avoid less dire consequences.

Also a computer will not choose a line that will backfire if the opponnent sees an ingenious counter strategy that the computer sees. As long as lengthier strategies don't have this drawback.

In spite of my relative ignorance of both chess and computers, I'm still betting that the only way 2600 computers "crush" 1400s worse than 2400 humans do, is that their win percentage is higher. Add bonuses for fast wins and no change to the program, and the human does way better.

[StregaChess]

[ QUOTE ]
But the same considerations are used to avoid less dire consequences.

[/ QUOTE ]

Maybe we just don't have an optimal model in this question. So I still disagree.... let me try to get there in another direction and this is all from experience having started out as that 1100 decades ago and stopping in chess with a rating over 2300 back in the 1990's.

Here's how I've seen it flow...

An 1000 player against an expert/master is like getting hit in the face with a 50 lb sledgehammer, the 1000 facing a strong computer is like getting hit in the face with a 300 lb sledgehammer. The weak human never gets out of the box for other "conditions" to come into play.

If you change the question to a world champion level player, I would think he would get the job done sooner, even if the world champion was not as strong as a computer.