Few stories make probability as exciting as the MIT blackjack team. For years, groups of students and graduates from the Massachusetts Institute of Technology and nearby universities used card counting, careful teamwork and serious bankroll management to win money from casinos. The story became a bestselling book and a Hollywood film. Strip away the drama and what is left is a superb lesson in applied statistics.
Why blackjack can be beaten
Most casino games are independent trials: a roulette wheel has no memory. Blackjack is different. Cards are dealt from a shoe, and the cards already played are gone until the next shuffle. That means the odds change as the shoe is dealt. When many low cards have come out, the remaining shoe is rich in tens and aces, which favours the player: more blackjacks (paid at 3 to 2), more successful double-downs, and more dealer busts, because the dealer must hit stiff hands.
The mathematician Edward Thorp showed this rigorously with computer simulations and published the method in his 1962 book Beat the Dealer. It built on a 1956 paper by four US Army engineers, Baldwin, Cantey, Maisel and McDermott, who had worked out a near-optimal basic strategy using desk calculators.
How the Hi-Lo count works
The most widely taught system, Hi-Lo, gives every card a value:
| Cards | Count value | Why |
|---|---|---|
| 2, 3, 4, 5, 6 | +1 | Low cards leaving the shoe help the player |
| 7, 8, 9 | 0 | Neutral |
| 10, J, Q, K, A | −1 | High cards leaving the shoe hurt the player |
The player keeps a running total as cards appear. Because a +6 count means more in a nearly finished single deck than at the start of a six-deck shoe, counters convert it to a true count by dividing by the number of decks remaining. As a rule of thumb, each point of true count shifts the edge about half a percentage point toward the player. A basic strategy player starts at roughly −0.5 percent, so at a true count of about +2 the game is roughly even, and above that the player has the edge.
The counter bets the minimum when the count is low and raises the bet when it is high. The skill is not memory, as films suggest; it is fast, accurate arithmetic under pressure and the discipline to vary bets without attracting attention.
From a January course to a team
In January 1979, during MIT's Independent Activities Period, when students take short non-credit courses, a class on gambling strategy brought interested students together. Some of them formed a team. Bill Kaplan, a Harvard MBA who had already run a successful blackjack team in Las Vegas, agreed to manage and fund an MIT group from 1980, bringing strict training, testing and record-keeping. Over the following two decades, various MIT-linked teams operated in Las Vegas, Atlantic City and elsewhere, with players coming and going.
Teamwork: spotters and big players
The teams used an approach pioneered in the 1970s by the professional blackjack player Al Francesco and popularised by Ken Uston. Spotters sat at different tables, betting the minimum and counting. When a shoe turned favourable, a spotter signalled a big player, who walked in and bet heavily only while the count was high. To the casino the big player looked like a lucky high roller who never seemed to follow the shoe. Pooling the bankroll across many players also smoothed out variance, which is the key to surviving as a counter.
The statistics that really matter
A counter's edge is small, often around 1 percent of money bet, and the swings are huge. A player can be behind after thousands of hands even while playing perfectly. That is why the team's real secrets were statistical:
- Sample size. Results only converge on the expected edge after a very large number of hands, which is why teams kept careful records and judged players on accuracy, not short-term results.
- Risk of ruin. The probability of losing the whole bankroll before the edge shows up. Bigger bankrolls and smaller bets relative to them reduce it.
- The Kelly criterion. A formula published in 1956 by John Kelly, a Bell Labs scientist, for sizing bets in proportion to your edge. Thorp applied it to blackjack and later to investing. Bet too much, even with an edge, and you can still go broke.
Legal, but not welcome
Counting cards in your head is not illegal in the United States; using a device to do it is a crime in Nevada. But casinos are private businesses. In Nevada they can ask advantage players to stop playing blackjack or leave. In New Jersey, a 1982 state Supreme Court case brought by Ken Uston established that Atlantic City casinos cannot simply bar counters, so they use countermeasures instead: shuffling earlier, limiting bet sizes, and using continuous shuffling machines that make counting useless.
Book, film and myth
Ben Mezrich's 2003 book Bringing Down the House, based largely on the experiences of former team member Jeff Ma, made the MIT team famous, and the 2008 film 21 took further liberties. Former members have said the book dramatised and compressed events, so treat it as entertainment rather than history.
Update (August 2026): Card counting still works in theory, but it is much harder in practice than in the team's heyday: more games pay 6 to 5 on a blackjack, continuous shufflers are common, and casinos share information on advantage players. Online blackjack typically reshuffles after every hand, so counting does not apply.
What students can take from it
The MIT team is the perfect case study for a statistics class: conditional probability, expected value, variance, sample size and optimal bet sizing in one real story. Several former members went on to careers in finance, technology and sports analytics, a path we follow in careers built on gambling maths. For the basics of why the house wins every other game, start with house edge explained.