MIT blackjack team: the real history behind the legend

MIT runs an Independent Activities Period (IAP) every January, when students can take short, informal courses. In January 1979 one of those courses covered casino gambling, and a handful of students came out of it convinced that blackjack card counting, published by Edward Thorp in 1962, could be run as a disciplined operation.

The early group had more enthusiasm than structure. That changed in 1980 when the players met Bill Kaplan, a Harvard MBA who had run a successful counting team in Las Vegas after graduating. Kaplan agreed to manage and fund the MIT players on the condition that they adopt his methods: rigorous training, written records of every session, and a strict split between investors and players.

J.P. Massar, one of the original students, is usually named alongside Kaplan as a founding figure. Over the following decade the personnel changed constantly. Several distinct teams and partnerships used the "MIT" label, and some members were never MIT students at all.

This page is part of the famous gamblers series in the Casino knowledge topic. The counting method itself is explained on the card counting page; the focus here is how a team turned it into a business.

The team's real innovation was organisational rather than mathematical.

Investors and bankroll

Outside investors put money into a bank for a fixed period. Players drew on that bank for trips. At the end of a bank's life, profits were split between investors and players according to a formula agreed in advance, often with players paid partly by hours played and partly by results.

Training and checkouts

A new player had to pass "checkouts" before playing with team money: counting down decks at speed, playing perfect basic strategy under distraction, converting the running count to a true count accurately, and handling bet signals. Failing a checkout meant more practice, not a trip.

Record keeping

Every session was logged: hours, hands, bets and result. The logs let managers compare actual results with expected results. A player running far below expectation over a long sample could be retrained or investigated; a small sample of losses was accepted as ordinary variance.

Why pooling matters

Pooling does two things. First, it lets each player bet in proportion to the whole bank instead of their own savings. Second, it combines many players' hands into one result, so the bank's swings are smaller relative to its size than any individual's would be. The same edge, spread across more hands, reaches its expected value faster. That is the law of large numbers applied deliberately.

A worked pooling example

Imagine five counters with $20,000 each. Using half-Kelly sizing, a top bet at a +1.5% edge is about 0.6% of bankroll (1.5% ÷ 1.3 variance ÷ 2), so each can bet only about $115 at a good count, because each has to survive their own losing streaks alone. Pool the money into a $100,000 bank and each player can bet about five times as much at the same count, because a downswing at one table is often offset by an upswing at another. Expected profit per hour roughly scales with bet size, so the pooled team earns several times what the five solo players would, while the bank's risk of ruin stays in the same range. The cost is trust: every player has to count honestly, report honestly, and follow the bet ramp exactly. That is why the checkouts and logs mattered as much as the maths.

The team's table method was borrowed from earlier teams, most notably the "big player" approach developed in the 1970s by Al Francesco and made famous by Ken Uston.

How a signal worked

A spotter who saw the true count rise past a set threshold signalled the big player, often with a body position and then a code word that carried the count. The big player sat down, bet heavily while the count stayed high, and left when it fell. To surveillance staff, the spotter never varied their bet and the big player looked like a wealthy gambler who happened to sit at a good moment.

The maths

With the Hi-Lo count, low cards (2–6) are +1 and tens and aces are −1. The running count is divided by decks remaining to get the true count. A common rule of thumb is that each point of true count adds about 0.5% to the player's edge. In a six-deck game that starts around −0.5% for the player:

• True count +1: about 0%
• True count +3: about +1.0%
• True count +5: about +2.0%

A big player betting $2,000 at a true count of +4 (about +1.5%) has an expected value of about $30 a hand. The spotter's $10 bets at negative counts cost a few cents each. The team captured the profitable hands and avoided most of the bad ones, which no solo counter can do without spreading bets in a way surveillance notices.

In 1992 several veterans formed a limited partnership called Strategic Investments to run team play on a larger scale. It recruited and trained a new generation of players and sent teams to casinos across the United States and abroad.

The larger the operation, the easier it was to spot. Casinos shared photographs through private agencies such as Griffin Investigations, and players connected to the team were increasingly recognised, backed off (told to stop playing blackjack) or barred. Some reported being detained or questioned. Strategic Investments wound down within a few years, and later groups using the MIT name operated on a smaller scale.

Were they profitable?

The team's own accounts and later interviews describe strong returns for investors over many banks, with some banks losing. No audited public figures exist, so the widely quoted totals in the millions should be read as approximate. What is well supported is that the method had a real edge and that the main threat to it was being barred, not losing money at the table.

Why it ended

• Surveillance and shared photo files meant a known face could be recognised across many casinos.
• Casinos added decks, cut penetration and introduced continuous shufflers in some games.
• Managing large numbers of players and investors created internal friction.

Ben Mezrich's *Bringing Down the House* (2003) told the story through a student called Kevin Lewis, based largely on former team member Jeff Ma. It was a bestseller and was adapted as the film *21* (2008), starring Jim Sturgess, Kevin Spacey and Laurence Fishburne. Jeff Ma consulted on the film.

Mezrich acknowledged that he used composite characters, changed names and compressed timelines. Later reporting and some former members disputed several scenes and the overall shape of the story.

The book is a good read and gets the core technique right. It is a poor source for dates, amounts and who did what.

The edge was small; the discipline was the product

A top counter plays with an edge around 1%. The team's advantage came from turning that small edge into a repeatable process: training, records, pooled money and role separation. Without the discipline, the same edge is easily lost to errors.

Variance is expected, not a failure

A 1% edge with a per-hand standard deviation of about 1.15 bets needs tens of thousands of hands before results reliably track expectation. Losing banks happened. The logs let managers tell variance from mistakes. See variance in gambling.

The casino controls the game

Card counting is legal in most places, but casinos can refuse to deal to you. The team's run ended because of backoffs and shared surveillance, not because the maths stopped working. For the legal side, see is card counting illegal.

It only works where cards have memory

Team play depends on a shoe dealt without replacement. On any game where each round is independent, a spotter has nothing to spot.

PVPspinArena runs three player-vs-player games in USDC or ETH on Base: Jackpot, Coinflip and Roulette. None of them has a shoe. Every round is generated from fresh committed seeds, so no previous result tells you anything about the next one, and there is nothing for a spotter to count.

• Roulette uses a 15-slot wheel: 7 Purple and 7 Silver pay 2x, 1 Green pays 14x. Every bet returns 14/15, about 93.33%, a 6.67% house edge, whatever happened on earlier spins.
• Coinflip is a 50/50 between two players, with the winner taking the pot minus any fee shown before entry.
• In Jackpot, your chance of winning equals your share of the pot.

What does carry over from the team's playbook is record keeping and a fixed budget. Any settled round can be checked on the fairness page. Play is 18+ only, and limits are on the responsible gambling page.