Chess vs. Poker vs. Computer

[StregaChess]

[ QUOTE ]
Chess computers CANNOT beat a weak player just as efficiently as a human player. Human players can set -EV traps that pay off against bad players. Computers are too good/stupid to do that.

[/ QUOTE ]

You are dead wrong, good computers CRUSH weak human players.

Regarding the whole computer vs man at chess, the deep blue match is not all it's cracked up to be. They basically cheated by changing the program during the match, agreeing to share information and withheld and whole bunch of bs...
In correspodence chess a human would still crush a computer...

[gull]

I am not dead wrong.

I agree that good computers crush weak human players.

My argument is that good human players can win more quickly and efficiently.

Against a bozo, 1. e4 2. Bc4 3. Qf3 isn't that bad. You play some bad moves for a big payoff. Computers are too smart to be that stupid.

[Shandrax]

1. g4 e5 2. f3 loses even faster, but does it mean that black has an advantage in the starting position?

I don't think that trying to find the fastest to win or the most idiotic way to lose proves anything in chess. Please do not mention the Hikaru Nakamura now.

[David Sklansky]

"This is what seperates poker from chess.. a chess computer can beat a weak player just as effeciently as a grandmaster can."

Because there is no bonus for beating somebody badly. But say there were extra points for checkmating in less than say 45 moves. If tournaments were comprised of good and bad players alike, a computer would lose to humans it could beat head up.

[Knockwurst]

[ QUOTE ]
"This is what seperates poker from chess.. a chess computer can beat a weak player just as effeciently as a grandmaster can."

Because there is no bonus for beating somebody badly. But say there were extra points for checkmating in less than say 45 moves. If tournaments were comprised of good and bad players alike, a computer would lose to humans it could beat head up.

[/ QUOTE ]

David, you are undoubtedly wrong about this. First, a chess computer can be programmed to play any number of different ways or open in any number of different ways. Thus, if there is an opening that is objectively weaker against a stronger player, but allows the computer to win in fewer moves against a weaker player, the computer can simply be programmed to do just that.

Indeed, a common program retailing for about $100 such as Fritz is so strong that it could play an inferior opening and still beat many mid-level grandmasters in a tournament like you proposed or heads up.

These programs are known for finding mates in the shortest number of moves which humans have missed in analysis. So there is no reason to think that the program would do anything other than find the quickest mate once it has gone into the opening it has been programmed to play.

Simply put, with a strong enough computer the most efficient response to a given chess move could be calculated whether its the first or last moves of the game and the human has no advantage here as he/she cannot more efficiently exploit a weak move.

Back to the more interesting question -- Can a program eventually beat the best poker players in the world? I sure would like to see you weigh in on that one.

[FortunaMaximus]

[ QUOTE ]
Back to the more interesting question -- Can a program eventually beat the best poker players in the world? I sure would like to see you weigh in on that one.

[/ QUOTE ]

Definitely not. No neural net's going to come close to an optimal human brain for pattern recognition and psychology. That's what will always set the human poker player apart.

HU, yeah, this program could crush 99% of opponents. But at a full table, not so much.

[bigpooch]

CHESS
-----

Yes, there were some strange occurences in the game where
Kasparov missed a fantastic draw against Deep Blue on the
Black side of the Ruy Lopez (e.g., how could a "bot"
overlook that a king move, although natural to a human
player, was a blunder that would lead to a repetition of
position?) although I still have some doubts. If I had to
choose whether if IBM had put in a human player at some
points of the match (or overruled the move that Deep Blue
chose) or if they did not, unfortunately, I think they did
for this game for at least two moves!

If you want me to be more specific about the move numbers of
the game, I can show you the reasoning behind my belief. As
for the other games, I didn't see this because it either
escaped my attention, or it simply didn't happen.

I totally agree that if you take two best "bots" and the
eight best correspondence players of the world (now) to
play a double round-robin, I am very confident that a bot
would not win the tournament. Also, I wouldn't be surprised
at all if both bots had a minus score. [And I wouldn't be
surprised if many OTB world class human chess players would
have a minus score!]


POKER
-----
On the other hand, although PokiBot might be able to play
really well in HU LHE, I don't think it would crush
significantly weaker poker players in NL HE, especially in
the context of a ring game, as well as a world-class NL HE
player. It's simply because it's often how much you win on
a hand that makes a big difference and a world-class NL HE
player would win more than a bot (as they currently stand in
NL HE) in a ring game.

[Knockwurst]

[ QUOTE ]
[ QUOTE ]
Back to the more interesting question -- Can a program eventually beat the best poker players in the world? I sure would like to see you weigh in on that one.

[/ QUOTE ]

Definitely not. No neural net's going to come close to an optimal human brain for pattern recognition and psychology. That's what will always set the human poker player apart.

HU, yeah, this program could crush 99% of opponents. But at a full table, not so much.

[/ QUOTE ]

Your assessment of heads up is interesting. In saying 99% of opponents, does that mean the hypothetical program could beat many of the best players in the world? At the very least it could be programmed to play a non-exploitable strategy so that the best would have no advantage over it, don't you think?

I don't see why a similar program could not be devised for a ring game taking into account each participants propensities by their previous play in the game, and then determine the best play when taking every players possible holdings in account. Admittedly, this is far into the future.

I guess one difficulty would be observational cues and psychological factors. Though you could program the computer so that if someone was beat within the last x amount of hands by an improbable turn of the cards and that player subsequently bet or raised, then y% of the time that player had a weak holding.

[FortunaMaximus]

Well, yeah, you can never really gauge incomplete information to a degree of precision that can be sustained indefinitely.

Note: I said 99% of players, not 99% of the world's best. That's the 1%.

Depends on how far into the future we're talking. A century, nah, I don't think so. A millennia, sure. Even then, what would be the point? By then, any Turing-capable AI will self-trigger causality. But that's another discussion.

[Knockwurst]

[ QUOTE ]
CHESS
-----

Yes, there were some strange occurences in the game where
Kasparov missed a fantastic draw against Deep Blue on the
Black side of the Ruy Lopez (e.g., how could a "bot"
overlook that a king move, although natural to a human
player, was a blunder that would lead to a repetition of
position?) although I still have some doubts. If I had to
choose whether if IBM had put in a human player at some
points of the match (or overruled the move that Deep Blue
chose) or if they did not, unfortunately, I think they did
for this game for at least two moves!

[/ QUOTE ]

The hand of Joel Benjamin may or may not have played a part in the match. But that's really incidental, because all that means is that there was a bug or problem in the program that did not allow it to see/calculate that variation, which could be addressed by improving the program or the hardware. There was nothing inherent about the positions that I'm aware of that allowed a human to find the best move where a computer could not. Though on that day that computer did not.


[ QUOTE ]
I totally agree that if you take two best "bots" and the eight best correspondence players of the world (now) to
play a double round-robin, I am very confident that a bot
would not win the tournament. Also, I wouldn't be surprised
at all if both bots had a minus score. [And I wouldn't be
surprised if many OTB world class human chess players would
have a minus score!]

[/ QUOTE ]

I don't see why a computer like deep blue running for as long as it takes for a correspondance player to make a move wouldn't out calculate and beat the correspondance player. Chess is utlimately calculating various analysis trees, opening theory and end game theory are just short cuts. On what basis would a human have any advantage?