Stony Brook Computer Science Professor Designed an Algorithm for Jai Alai Betting System

Steven Skiena, a Computer Science professor at Stony Book University, recently gave an interview on the Odd Lots radio show on Bloomberg about his sports betting system. Computer professors do not normally talk to economic publications about gambling, but Dr. Skiena has a remarkable story. The professor built a model for wagering on the obscure game of jai alai. Now he teaches people how that financial model can be applied to normal business and economic decisions.

Steven Skiena built a team of researchers to study the game’s patterns and develop a successful betting model. Using the system, he and his team grew their bankroll by 500%.

The team used algorhithms and computer models to break the system. They quickly beat the world of jai alai betting. Now, Skiena says he teaches finance students how to use similar methods for algorithmic trading in the stock market.

The Game of Jai Alai

Jai alai is a Basque game which is played on a court which reminds people of racquetball or handball courts, though the courts are immense. Instead of manipulating a ball with a racket or one’s hand, the ball is caught and thrown with a hand-held device known as a “cesta”. The game is a touted by the Basque Government as the “fastest sport in the world”, because the ball has been timed moving at 188 miles per hour. (A golf ball has been timed moving at 208 miles an hour, but players don’t try to catch golf balls.) Players wear helmets to keep the ball from killing them.

Though the game is obscure in most places, Jai alai is played in South Florida, where the game is a favorite sport for gamblers. Jai alai is played in the the Fort Lauderdale area by “mostly Basque players” at stadiums called frontons. That is where the professor first game the game played. Skiena won his first jai alai bet as a 12-year old in the 1970s, betting on a trifecta (win, place, and show). He turned $2 into $120.

Algorithmic Betting

According to Steven Skiena, an algorithm is typically a “programmed procedure for making decisions”. It is a set of rules, with some level of input or procedure which determines that a series of bets are productive without human input. Over a six-month period, his system produced 500% profits.

How Jai Alai Works

Skiena calls jai alai “incredibly exciting” to watch. He says, “The chances of winning changes with every point,” because of the way the game plays out. Each game has 8 teams, with each team playing against all 7 others. Only two teams can play at one time, though. Teams are marked 1 through 8, and the six teams who are not playing wait in line for their turn. The winner wins a point and keeps playing, while the loser goes to the end of the line.

Thus, the game plays out like certain games of billiards in pool halls or pickup basketball on city courts, where a person or team continues to play, until someone beats them. A team which does well can run the table quickly, though if they lose, they go to the back of the line. The system favors the teams which goes first, because they get a chance to accumulate points quickly. This goes for the second time through the line, too, when point values double. Team #1 or Team #2 are the first to go a second time.

Peri-Mutuel Betting Systems

The game plays out for gamblers much the way it would for horse betters. It is a pari-mutuel betting system, with gamblers able to be on options 1 through 8. You can bet to win, place, or show. The winner is the first one to get to 7 points.

In a pari-mutuel system, the sports gambler is betting against the other bettors. All the money is put in a betting pool, the house skims off the top, and the rest bet against each other. This means it’s player-versus-player instead of player-versus-house. Because of this feature, many gamblers will be bad, which gives a smart gambler a chance to win a lot of money, if they find a system which can beat the vigorish or juice. Obviously, most gamblers cannot do that.

Discrete Betting Events

Steven Skiena and his team developed a system to beat the game. Jai alai can be broken down into a series of a discrete events, like baseball. It is a series of single at-bats, instead of continuous action. Horse racing is not a bunch of discrete events — they are continual. By breaking the game into a discrete event, you can begin to chart as many outcomes as you want — Skiena suggests he and his team charted “a million”. Steven Skiena visualizes the problem like a coin-toss, with each base outcome

Skiena says jai alai betting can be seen as a tree process or tree chart. At every point in the tree is two players playing each other. Depending on who wins it, another point on the tree plays out.

Gambling Using Tree Charts

To win, you have to build a model of the players’ skills, to see which players are better than others. Also, you have to build a model of how the public is going to wager. Once you do this, you can start to see betting discrepencies. Also, you can start to see how your bet affects the betting pool itself, to make sure you aren’t betting the same way as everyone else, and thus exchange money between yourselves.

Take one example. One interesting property is it is hard for all the high uniform numbers to do well. It is almost impossible for the teams with high numbers to all win, place, or show. Yet many gamblers wager on the teams with big numbers to win, place, or show, because they have good odds. These are sucker bets, which have almost no chance of winning.

“Calculated Bets” by Steven Skiena

Steven Skiena’s book on the subject, “Calculated Bets”, did not receive a lot of attention from the jai alai betting community. It did gain a lot of attention from stock traders.

The professor says building a system for jai alai betting was harder than he thought it would be, because of the 20% juice taken by the house. He said the market was small enough that he was not able to engage in volume betting. In jai alai, he said the betting community was small enough that you bet against yourself. In trading, the community is big enough that is not an issue.