By coincidence, "D" and "J" were also the initials of the first person I was blackjack teammates with back in 1997, when I left my position as a professor of mathematics at Ohio University and ventured into the world of blackjack card counting. To avoid confusion, I’ll call my old teammate "DJ2." I met DJ2 in 1983 when I first joined the mathematics faculty at Ohio University. At the time, DJ2 was an un-tenured assistant professor of Computer Science who didn’t like students very much. He was also knock-your-socks-off brilliant. He was exceptionally strong at strategy games and was one of my earliest chess (and drinking) buddies. He only lasted two short years in academia. In 1985, DJ2 sold almost everything he owned, packed the rest into his new red sports car, and drove off to Silicon Valley in California. He landed in the right place at the right time and thrived in the early days of the high-tech boom.

Fast-forward 11 years. In late 1996, I got a call from DJ2 out of the blue. It was after midnight, and it was clear that he was very drunk. But he couldn’t stop talking about how he had proven this mathematical theorem that could be applied to horse racing. He explained that if he had enough data from enough races, together with the real-time odds at the tracks (and a lot of other variables), he could give odds that were much more accurate than the posted odds. Anytime a horse was paying substantially more than its true odds, money could be made by betting on that horse. The idea was to wager a fraction of the bankroll on each horse in proportion to the edge his model predicted, i.e., Kelly betting. In a given race, there could be multiple horses on which to make a bet, or none at all. It was all about the model and the data. With his unique access to the data-Highway in Silicon Valley, he was well positioned to clean up.

As he conducted his mathematical research, DJ2 had many of his questions answered in early online message boards (the newsgroup rec.math.statistics, for example). DJ2 then explained to me that he had been contacted by someone, who had heard about his research second-hand from the newsgroup threads. The person (and I’m probably not remembering the details correctly) either wanted to buy the system from DJ2 or wanted to pay DJ2 to not use it, I don’t recall which it was. But, either way, I do recall that the fee they wanted to pay DJ2 was $90,000. I then learned that a syndicate in Hong Kong had been using a similar model for some time.

Here are some links that discuss the Hong Kong syndicate and use of a mathematical model for horseracing: