Skip to primary content

Algorithm | Aug 01 2004

The game is five-card draw. You compete against three opponents. Each is represented by a photograph in one of the corners of the screen. You chose these opponents from a gallery of candidates. The program provides a brief bio of each, with an assessment of his or her skill level.

Selecting at random, you choose a beginner, an average player, and an expert. The expert, Marv, reminds you of your mother’s ex-boss, the man who fired her. He is a shrewd-looking middle-aged man in a suit. The beginner, Janet, is an equally shrewd-looking thirty-something woman in a suit. You notice, because it is difficult not to, how sexy she is. Actually Marv is sexy too, as is the average-level player you chose, Jim, a youngish guy in a suit.

All the players wear suits, including the ones you pass over, and all are sexy.

Each player begins with five hundred dollars. There’s a fifteen dollar maximum bet in the first round, twenty-five in the second, with two raises permitted per round. Stacks of chips rest near the players’ photos. When a player makes a bet, an appropriate number of chips disappears from that player’s stack and re-appears in the center of the screen, accompanied by a chips-hitting-the-table sound. Play proceeds around the table. The program flashes each player’s plays and prompts you when it’s your turn.

Right from the start, Janet plays like a beginner and squanders her money. You feel bad for her: she bet on nearly every hand, no matter how terrible—a classic beginner’s mistake. You play as you usually play in a new game—conservatively, feeling people out. Only you soon realize that there’s nothing there to feel. The players have different styles, that much is clear, but there’s no sense of a personality or thought process behind their decisions.

When Janet loses her final bet, you and Jim have won most of her money. Surprisingly, Marv has played aggressively, frequently raising in the first round. The tactic has cost him; he’s down to less than three hundred dollars. Soon, though, his fortunes turn. He and Jim compete for nearly every pot, each repeatedly raising and re-raising while holding nothing. Marv wins the majority of these battles and quickly jumps into the lead.

A big hand follows in which everyone remains in for the maximum number of raises. You win with three kings. Marv loses with a queen high two-pair. Jim loses with absolutely nothing—an insane, kamikaze bluff. This nearly finishes Jim, but then he wins a few small pots to get back over a hundred dollars. The final blow is struck by Marv. It is another confrontation of nothing versus nothing. Marv’s nothing wins this time, and now it’s just the two of you.

You sit up in your chair.

When your mother was fired, she was given a single afternoon to clear out her desk. A security guard sat with her to ensure she didn’t sabotage the computer system or do god knows what. She had worked at the company for seventeen years. The boss who fired her was her sixth. He was hired to cut costs, and that’s what he did. Your mother, now asleep in her bedroom down the hall, was a cost he cut.

Things go badly at first. Marv is aggressive and wins a long series of hands in which you fold after his first or second bet. Soon you’re down to less than five hundred dollars, the amount you started with. This means that Marv has over fifteen hundred. You’ve been banking on Marv’s recklessness, but it’s not working. The amount you win in the big hands, the ones in which Marv stays in with nothing, is less than what Marv is winning in the larger number of small hands. You need to change tactics.

You pause the program to think it through.

The approach you settle on is not something you would ever try at a real poker table. It comes to you when think back and realize that once Marv has made a big bet, he always remains in until the end, provided you haven’t preceded Marv’s bet with one of your own. For this reason you decide to rely exclusively on what is called sandbagging, which is the act of passing with a good hand in order to lure others into misevaluating your hand and thus betting more than they should. Once they do this, you raise the maximum. It’s a kind of trap, and when it works, your opponent feels compelled to call your raise, because of how much he or she has already bet on the hand.

The plan works to perfection. Each time you’re dealt a strong hand, you bet nothing, waiting for Marv to bet first. When Marv does this, which he does regardless of which cards he holds, you raise him the maximum, and now he’s yours. A real poker player would quickly realize what you were up to and stop falling for it. But Marv is not a real poker player; he’s an algorithm, and a rather predictable one at that.

The final hand is typical. You’re dealt two queens—a strong hand in a two-person game of five-card draw. Still, you choose to pass. Marv bets ten dollars and you raise him fifteen. Marv raises you another five and you call. Marv takes two cards; you take three, none of which help you. Marv bets twenty dollars and you raise him twenty-five. Marv calls with his last twenty-five. There’s now a hundred and ninety dollars in the pot.

Marv has a jack high nothing. He called a twenty-five dollar raise in the last round with a jack high nothing. Your pair of queens win, although a single queen would have sufficed.

A message appears on the screen:

Game over. Would you like to play another?

You click NO and close the program. Down the hall your mother calls out something in her sleep. Probably she was doing this the whole time you were playing—she does it a lot—but you were too engrossed in the game to hear her.