Take risk out of the game
Kaplan is a walking oxymoron: the risk-averse gambler. He won't play until he has taken every possible measure to ensure he will win. At cards and in business, he never bets above his conservative comfort level, and he refuses to indulge reckless speculators. "No one should purchase any units who cannot afford the loss of his entire investment," reads a warning on his prospectus for the 1992 limited partnership agreement.
So Kaplan looks for opportunities where math and research -- rather than skill or luck -- determine the outcome. Blackjack is one such opportunity. "Everyone thinks you can't win at blackjack, because if you can, how could those casinos make so much money?" says Kaplan. "But if you analyze the game properly, you can come up with a strategy that gives you a positive expectation."
The basis of Kaplan's strategy, which he learned from Beat the Dealer, a 1962 book by Edward Thorp, involved assigning number values to high and low cards as they are played. Aces, face cards, and tens are worth -1; two through six are worth +1; and seven, eight, and nine are worth 0. Card counters keep a running total in their heads. The higher the total, the fewer aces and face cards that have been played, and the more likely a rich vein of such cards is imminent. Because face cards and aces favor the player, higher counts elicit higher bets. Players using this system have a 1 percent to 2 percent advantage, but results swing wildly in the short term. (Kaplan once lost 20 hands in a row, a one-in-a-million probability.) So it usually takes many, many games -- 500 to 1,000 hours of play -- to achieve the desired return.
Team play, which Kaplan learned in Las Vegas from Blackjack Hall of Famer Ken Uston, allows for far more time at the table. It also allows some players to act as "spotters" -- monitoring the game and betting conservatively, then signaling discreetly to "big players" who join the action when a deck is poised to spit high cards. Big players always bet high, so they don't tip off casino personnel by changing their behavior just as the cards get hot.
Kaplan relied heavily on his university's computers -- mainframes with punch cards in those days -- to run every possible hand and calculate the optimal player response. Blackjack, in this context, becomes an if-then proposition. "There was no flexibility about what players did at the table," says Kaplan.
On any given Saturday, Kaplan might have had teams plying Las Vegas, Foxwoods in Connecticut, New Orleans, Montreal, and Lake Tahoe. Every six hours, team members would meet at designated locations in their respective cities and calculate how much they were up or down; then they would call an 800 number and leave their results for Kaplan. Kaplan would determine whether the teams should, as a group, raise or lower their bets and by how much. "Say we started with a million dollars, and now everyone has played, and we're down $100,000," says Kaplan. "Now we have nine-tenths of our capital base. Before we were betting $1,000; now we're going to bet $900. You are always betting in proportion to your capital. If you lose money, you lower your bet. So theoretically, your risk of ruin is always zero."
Risks in Kaplan's other ventures are less clear-cut. "It's not like blackjack, where you can program the whole game into a computer," he says. Still, in real estate he reserves his big bets for sealed-bid auctions and sales of bank-owned properties, where information is limited and so his research and deep knowledge of the local market give him an edge. At FreshAddress, he is scrupulous about financial exposure -- never assuming debt, reducing costs by licensing data as needed, and using resellers who receive a percentage rather than salaries. Kaplan expects FreshAddress, like blackjack, will pay off handsomely if he is prepared to wait.
Train. Test. Test. Test.
What do blackjack players and air traffic controllers have in common? They can't afford to make mistakes.
Perfection, of course, is a lot to ask. But Kaplan demanded it from his players, because the difference between making money and losing it was paper-thin. So he devised a series of staged tests, or "checkouts," everyone had to pass. Newcomers -- usually referred by a team member or intrigued by a blackjack class held on campus during winter break -- would show up at the MIT classroom that served as a practice space. First, they would learn basic strategy. Then, the checkout: two straight hours of play without a single mistake. Pass, and the aspiring player would learn to count cards, followed by a second two-hour checkout. "Every few rounds, we'd ask them, 'What's the count?" says Kaplan. "If they were off more than once, or by more than one, they failed. They also had to make the right bets. The pressure would mount. Everyone would crowd around. They'd yell, 'He's never going to make it!' People are calling out numbers. They're doing everything they can to throw him off."
The final checkout -- three two-hour sessions -- took place at a casino. Again: no mistakes allowed. Blow it, and the player would have to practice more and try again. Pass it, and he could gamble with the team's money.
Sort of. Because just because someone could play perfectly didn't mean she always did play perfectly. Skills dulled over the week as people turned their attention to jobs and homework. So on Thursdays, the day before teams dispersed for a weekend of gambling, members would gather in the classroom for yet another checkout -- this one lasting 45 minutes. "Occasionally, people who failed Thursday night would say they'd practice on the plane," says Kaplan. "Fine. But if you fly all the way to Las Vegas, and the trip manager checks you out and you still fail, you are not going to play."
