Wednesday, August 28, 2002

Vegas, Baby. Vegas. A fabulous article at Wired about a group of super-genius card sharps from MIT that actually beat the house at Vegas.
For six years in the 1990s, Lewis was a principal member of the MIT Blackjack Team, an infamous cabal of hyper-geniuses and anarchistic whiz kids who devised a method of card counting that took the gaming world completely by surprise. Funded, in part, by shadowy investors and trained in mock casinos set up in classrooms, dingy apartments, and underground warehouses across Boston, Lewis and his gang used their smarts to give themselves an incredible advantage at the only truly beatable game in the pit. A baby-faced card-counting team possessed with impressive mathematical skills - here was a novelty that turned blackjack into an arbitrage opportunity. Their system was so successful, it took nearly two years before the casinos began to catch on - engaging in a cat-and-mouse war with the well-trained MIT conspirators.
They played the age old game of card-counting - if you remeber the Vegas scene from Rain Man, you saw a ham-fisted dramatisation. It sounds like there were two keys to the scam. One was working as a team to trade signals and track cards so as to make it look like the winner wasn't counting cards (counting cards gets you thrown out in Vegas) and worked against the typical Vegas security profile of card counters
The real genius of the MIT scheme was how it turned the casinos' own profiling techniques against them, using stereotypes to camouflage the big money bets.

The MIT team thrived by choosing [big players] who fit the casino mold of the young, foolish, and wealthy. Primarily nonwhite, either Asian or Middle Eastern, these were the kids the casinos were accustomed to seeing bet a thousand bucks a hand. Like many on the team, Kevin Lewis was part Asian, and could pass as the child of a rich Chinese or Japanese executive. "When you're recruiting, you don't recruit white kids. They look conspicuous. Asian kids, Greek kids, dark skin fits in better with lots of money in the casinos. White 20-year-olds with $2 million bankrolls stand out," explains Andrew Tay, one of Lewis' teammates. "A geeky Asian kid with $100,000 in his wallet didn't raise any eyebrows."
Plus, they used sophistcated statistical techniques.
Two of the tricks that became a staple of the MIT system, shuffle tracking and ace tracking, exploit a concept called the nonrandom shuffle. Because of time constraints, blackjack dealers cannot achieve completely random redistributions during the shuffle. This means that certain packets of cards remain close enough together to be "tracked" through the deck. By watching a group of low cards, for example, it's possible to cut the deck (players assist the dealer by placing the cut card into the shuffled stack) in such a way that some low cards never have to be played. Likewise, a good shuffle tracker can "predict" a string of high cards and raise his bet even before the count goes positive.
But perhaps most importantly the did an enormous amount of simulation ahead of time.
After passing the Spotter test, Lewis moved on to the [big player (BP)] exam. Called to a table midplay, a BP has to take the running count and convert it into the more accurate "true count," by estimating how many cards are still left in the shoe. That's because a count of plus 10 - a ratio of 10 extra high cards to low left to be played - has a much higher value when there is only one deck left in the shoe, as opposed to six. Once the true count is established, a BP has to determine the proper bet. On the test, Lewis was asked to make highly complex decisions - such as when to split pairs against certain counts - while Martinez and Rosa graded his play from across the room.
In time - but only after years and millions in winnings - the casinos caught on.

The article is an excerpt from an upcoming book on the subject. I'm betting it will be quite a good read.