A Professor Bets on Jai Alai -- With Computer Models
By BROCK READ
The sport of jai alai might not have a wide following, but its complex scoring system "has interesting mathematical properties that just beg the techno-geek to try to exploit it," writes Steven Skiena, a professor of computer science at the State University of New York at Stony Brook. In a new book, Mr. Skiena does just that -- using computer models to develop a system for betting on jai alai that he put into practice.
In its 200 pages, Calculated Bets: Computers, Gambling, and Mathematical Modeling to Win (Cambridge University Press and the Mathematical Association of America, 2001) covers a lot of ground. The book explains jai alai's rules, regulations, and betting options; the mathematics behind the sport; the computing technology that Mr. Skiena and three student assistants used to create betting models; and the results of the project. Mr. Skiena also offers asides about topics such as data analysis, elegant computer programming, and why computer scientists hate Microsoft. (More information about the book is available online at Mr. Skiena's Web site.)
Jai alai is played live as a betting attraction in Connecticut and Florida, and its scoring system reflects its gambling accessibility. Games involve eight participants: Two players start by competing for a point, while the others wait in line. The winner of each point stays out on the court, while the loser returns to the back of the queue. A game ends when one player accrues seven points; then, win, place, and show positions are awarded.
Using a computer program that simulated jai-alai matches by randomly generating a winner for each point, Mr. Skiena learned that, if all players were of equal skill, some bets would be much safer than others. A trifecta -- a bet in which the win, place, and show players must be chosen correctly and in order -- pays off substantially more often if it includes the 1, 4, and 2 players than if it is a 5-8-7 bet, for example.
By comparing data from his simulation with records of actual matches, Mr. Skiena learned that his initial simulations only told part of the story. In order to factor in the skills of individual players and doubles teams, Mr. Skiena programmed his computer to download and process game scores posted daily on the Web sites of three leading jai-alai courts.
He also had to take into account the system used by jai-alai betting houses -- parimutuel wagering, in which all of the money bet on a game is pooled and split among the winners. To make money, Mr. Skiena had to teach his program to take into account the effect that its bets would have on each game's pool and to avoid commonplace bets that would yield disappointing payoffs.
Ultimately, Mr. Skiena writes, "it was time to put my money where my mouth was." In the first six months he ran it, his program earned him more than $550; in a later two-week period, it netted nearly $1,000. (Because his winnings came from a research project, Mr. Skiena donated all of the money to a university charity.) Mr. Skiena engineered his transactions completely by computers, which canvassed jai-alai courts' Web pages for schedules and dialed a telephone service to place bets.
His work on jai alai was "sort of a left-field thing," says Mr. Skiena, whose "honest research" is in computational biology. But Calculated Bets links its gambling models to the mathematical and computational concepts that underlie them. Throughout the book, Mr. Skiena digresses from his main story to offer explanations of predictive computer modeling, the difference between probability and statistics, dangers that computer scientists face in generating random numbers, and language-recognition techniques that computers use. He also includes a glossary and an appendix of recommended mathematical-modeling projects.
Calculated Bets is not a textbook, says Mr. Skiena, but its emphasis on applied computing and mathematics could make it a supplement for modeling and statistics courses.
Keith Devlin, executive director of the Center for the Study of Language and Information at Stanford University, agrees. The book "is almost a case study of how you would take mathematical reasoning and put it to work in a particular domain," Mr. Devlin says. "You can see the love and inspiration that [Mr. Skiena] has."
Feedback has come from numerous venues, according to Mr. Skiena. Sports and statistics enthusiasts, gambling-industry insiders, and jai-alai devotees have all contacted him, almost all with positive comments. Some members of the "beleaguered minority" of jai-alai fans have issued a mild reproach, though: "They're always complaining I didn't say jai alai was beautiful enough," Mr. Skiena says.
Mr. Skiena stopped betting on jai-alai matches some time ago, but his program still runs, sending him recommended bets that he checks daily. While he enjoys writing for a popular audience, he does not see any more gambling models in his future: "I feel that we've proven we could beat jai alai."