Can Cepheus play an essentially perfect game of poker?
Posted by: Timothy Tibbetts on 02/18/2015 12:07 PM
[
Comments
]
A while back we posted a story about Cepheus, a computer program that is supposed to play poker better than any human being. This is being done as a test from the University Of Alberta, Canada. Being a poker player, I had to try it out. I took a while to even have it working online and when it was on line, took 15 minutes to play as it only lets 5 players at a time with 100 hands.
There may be science behind it but it took me about 5-10 hands to realize how to beat the computer. Here is what I spotted that let me adapt my style of play:
The sounds or lack thereof The game needs sounds. Poker has sounds. A dealer talks the game out. Players talk. Chips shuffle. The game has NO sounds. There should be a sound specific to call, raise, re-raise, and fold. IMHO, this makes the game a little harder to play and puts the human player at a disadvantage.
The colors man The first thing I noticed was how the layout was lacking in any basic sounds or colors. Would it kill a college to grab some sounds so that you have a sound effect for raise, reraise and when you click? While they want to say this is a science project, it’s also important for a player to spot the different colors and suits of a card to watch for hands like a flush. There are different color cards for each suit but a poker player is used to seeing the Ace of spades being a black spade, the ace of hearts being a red heart and so on. Once again, this makes the game a little harder to play and puts the human player at a disadvantage.
Raises too much While people will fold pre-flop, they do so less heads-up. Toss in the fact that its limit hold’em and I play everything. I won a ton of hands playing rags against a pre-flop raise.
Bluff happy I intentionally played a number of hands just to see what the computer had. He bluffs and is willing to play weak hands to the end. You know, like any poker AI you have ever seen.
Very aggressive This covers both the above mentioned steps. While the computer can be very aggressive, it is willing to fold following a few re-raises. Quite a few times following these folds, the computer seemed more willing to fold pre-flop the next hand.
Conclusion Overall, the AI does work well if limit hold’em was an aggressive game but it’s not. I don’t think it has “solved the problem” as they claim. I doubt it ever will; poker just has too many variables on both sides of the table. There are plenty of single player games you can try that play against a computer and they are about as good. Then again free online poker lets you play against real people anytime.
I went down a few points in the first 5 hands but by hand 50 I was ahead 550 points. Now consider that I’m rusty and rarely have played limit hold’em. A good AI needs to learn to adapt to its competitors gameplay to figure out how to win or at least make it competitive. Then we would have a game. Otherwise, in the case of Cepheus all you have to do is adapt to the computer’s gameplay to win.
Learn more or try it yourself at http://poker-play.srv.ualberta.ca/.
There may be science behind it but it took me about 5-10 hands to realize how to beat the computer. Here is what I spotted that let me adapt my style of play:
The sounds or lack thereof The game needs sounds. Poker has sounds. A dealer talks the game out. Players talk. Chips shuffle. The game has NO sounds. There should be a sound specific to call, raise, re-raise, and fold. IMHO, this makes the game a little harder to play and puts the human player at a disadvantage.
The colors man The first thing I noticed was how the layout was lacking in any basic sounds or colors. Would it kill a college to grab some sounds so that you have a sound effect for raise, reraise and when you click? While they want to say this is a science project, it’s also important for a player to spot the different colors and suits of a card to watch for hands like a flush. There are different color cards for each suit but a poker player is used to seeing the Ace of spades being a black spade, the ace of hearts being a red heart and so on. Once again, this makes the game a little harder to play and puts the human player at a disadvantage.
Raises too much While people will fold pre-flop, they do so less heads-up. Toss in the fact that its limit hold’em and I play everything. I won a ton of hands playing rags against a pre-flop raise.
Bluff happy I intentionally played a number of hands just to see what the computer had. He bluffs and is willing to play weak hands to the end. You know, like any poker AI you have ever seen.
Very aggressive This covers both the above mentioned steps. While the computer can be very aggressive, it is willing to fold following a few re-raises. Quite a few times following these folds, the computer seemed more willing to fold pre-flop the next hand.
Conclusion Overall, the AI does work well if limit hold’em was an aggressive game but it’s not. I don’t think it has “solved the problem” as they claim. I doubt it ever will; poker just has too many variables on both sides of the table. There are plenty of single player games you can try that play against a computer and they are about as good. Then again free online poker lets you play against real people anytime.
I went down a few points in the first 5 hands but by hand 50 I was ahead 550 points. Now consider that I’m rusty and rarely have played limit hold’em. A good AI needs to learn to adapt to its competitors gameplay to figure out how to win or at least make it competitive. Then we would have a game. Otherwise, in the case of Cepheus all you have to do is adapt to the computer’s gameplay to win.
Learn more or try it yourself at http://poker-play.srv.ualberta.ca/.
Comments
