Regression Analysis in Sports Betting Systems

Sport betting is a form of wagering on the outcomes of traditional probability games such as cards, dice, or roulette as well as on the outcomes of sporting events such as football or baseball. Betting results are resolved at the conclusion of the event and generally requires that neither of the parties involved in the wager has any influence on the event in question. A private citizen wagering on the outcome of a football match between Manchester United and Chelsea would be considered a form of sports betting, while the team owner making that same bet would not be considered a form of sports betting.

Sport betting has been popular for centuries. The earliest forms of contemporary sport betting revolved around cock fights in the late 17th century, while wagering on horse races became highly popular between the 19th to 20th centuries. In 1960, television gave birth to a new era in the history of betting on team-based sports, and the longstanding 10% tax on sports betting was eliminated in 1974, leading to even more popularity. In 1990s, the advent of the Internet facilitated online sports betting, creating an increasing need for sophisticated statistical tracking, like regression analysis, to develop winning strategies for wagering.

Traditional gambling and sports betting: Are they same?

There are several key differences between traditional gambling and sports betting. In traditional gambling, the likelihood of any event can be calculated accurately even though the number of possible results may differ (for example: the odds of picking a specific card out of a standard deck is always one in 52). However, the likely result of a sporting event cannot be calculated with such precision. While the odds of winning in traditional gambling are derived from a known probability, sports betting relies on parsing a large number of variables, usually with more than one potential outcome.

Sports gambling: Profitable or not

In 2011, the gross gaming revenue of the global gambling industry was $368.4 billion, with 8.60% of this total earned from wagers made over the Internet. Online gambling includes all Internet-based portals that provide lotteries, poker, casino games, bingo, and sports betting. Sports betting remains most popular form of online wagering, representing 43% of gambling revenue earned from Internet sources, with a total market of $13.66 billion.

Figure: Worldwide gross gambling revenue

**source: Statistical Methodology for Profitable Sports Gambling

Methods used in sports betting system

Various methods can be used to generate a sports betting system, although most experts agree that the most widely used method is regression analysis. Regression analysis can be used to establish the important factors and variables which will influence the overall outcome of a sporting event. Multivariate linear regression, logistic regression, and multiple regression analysis can all be used to calculate the probability of any outcome, and since determining the outcome of a sporting event requires analyzing a high number of variables, regression analysis provides a suitable framework for defining and assigning a value to these variables. For example: A multivariate linear test on American football games was conducted by NFL. The result showed that the most important variable – the variable with the highest influence over the outcome of the match – was “passing efficiency”. Recent movies and bestseller titles like Moneyball have delved into the world of statistical analysis, driving increased interest in the use of regression analysis for sports betting.

Logistic regression analysis

Logistic regression is a forecasting technique that provides a probability percentage for a given variable. For example, if one wants to calculate the probability of a team winning the 59th game of the season, they would analyze the last 58 games to obtain the team’s point differential or margin of victory (MV or MOV). Margin of victory is a statistical term which indicates difference between the number of points scored by the winning team and the number of points scored by the losing team. A smaller MV represents a close match, and by using statistical software like SPSS, the following equation can provide the percentage chance that the team will win, based on MV scores:

(e is known as euler’s number, roughly 2.72).

A percentage chance of winning can also be determined in Microsoft Excel by using this equation:

1 / (1 + EXP(-(-0.0039+0.1272*[MOV])))

Multiple regression analysis in sports betting

Multiple regression systems are widely considered the most reliable modern sports betting system. This core of MRA is built on a timeless logical assumption: “what’s past is prologue”. This means that one must know the past to know the future. To create a multiple regression betting system, one must have reliable data regarding past information of the players and teams, meaning that trustworthy historical data is crucial to building an effective multiple regression system.

An example of using a multiple regression system in sports betting

A sports bettor will wager on the final match between Team A & Team B.

Regression #1: Bettor finds that Team A won the regular series against Team B by 3-1 during the first match of the year.

Regression#2: Bettor finds that Team B crushed Team A in a recent playoff match.

Regression#3: One player of Team A is Player X, and Player X has never won against Team B.

Since both teams have scored a victory, bettor determines that the key variable is the presence of Player X, and decides that Team B will win the match. Thus, by using multiple regression analysis, bettor is able to analyze the events of the past and extrapolate the most probable future.

To utilize multiple regression methodology in a betting system, one needs to posses consistent and reliable data on the past performance of both teams and players (“Multiple Regressions”:2013). Without an extensive and dependable source of historical data, the bettor will not be able to regress into the past to determine probable outcomes of future events (“Multiple Regressions”:2013).

To develop a multiple regression system, mining data from an online sports book that can offer accurate historical sports data in a format that is easily accessible and actionable is highly recommended. These sports books also provide step by step rules for implementing regression analysis techniques in sports betting.

Note that regression analysis methodology is also employed by most casinos in an effort to generate probabilities that favor the house – for similar reasons, sports books use regression analysis to provide sports betting enthusiasts with the same advantage. While we all know that no future event can be predicted with 100% accuracy, a comprehensive regression analysis system can be used by sports boo developers  to calculate probabilities that are highly reliable.

Problem of using regression analysis in sports betting

There is one glaring problem in using regression analysis to predict outcomes of sporting events: the differentiation between correlation and causation. Regression analysis is effective at identifying a correlation between events, but cannot properly identify whether one event is caused by another. For example: regression analysis can be used to show that every time Team A loses, player X does not score a goal. However, regression analysis cannot be used to conclude that Player X not scoring a goal is the cause of Team A losing the match.

In other words, regression analysis can be used to determine probable future performance based on defined past outcomes, but is unable to define causes for past outcomes. Ultimately, the effectiveness of any multiple regression system relies entirely on the proper selection and comparison of variables.

Other betting systems

In addition to multiple regression analysis, there are two other commonly used wagering methodologies: the arbitrage betting system and the use of statistical anomalies. Arbitrage betting is designed to generate profit without taking a loss (“Multiple Regressions and Statistical Anomalies”:2012), and in most cases the result of sports event is not considered. Naturally, profits are not guaranteed, but arbitrage is a straightforward strategy that can easily be learned by novice bettors.

When implementing a strategy around statistical anomalies, the bettor seeks to gain a competitive advantage by diverging from seemingly sound predictions by introducing variables that are often overlooked by other forms of betting systems. Using this tactic successfully requires a careful study of both teams and players, as well as a variety of incidental variables, such as weather, crowd sizes, health conditions, or injuries. By using this methodology, the bettor is attempting to determine how individual players and teams deal with anomalous situations not generally encountered during a match (“Multiple Regressions and Statistical Anomalies”:2012).

Conclusion

While regression analyses can help a bettor identify and define the variables that may affect the outcome of any given match, determining which variables to measure and compare is the central challenge in building a winning regression system. Therefore, regression analysis in sports betting is based upon not only a comparison of reliable past data with future events, but in deciding which variables may potentially alter the probabilities of those future events.

References:

1. Margin of victory.(2013).Retrieved May 25,2013 from http://www.sportingcharts.com/

2. Sportsbetting-Past & present.(2013). Retrieved May 24, 2013 from http://www.thesluggers.com/

3. Multiple Regressions.(2012). Retrieved May 23, 2013 from http://www.casinoreview.org/

4. Multiple Regressions and Statistical Anomalies on Sports Betting.(2012).Retrieved May 23, 2013 fromhttp://www.777sport.co.uk/

5. Moya F.E. “Statistical Methodology for Profitable Sports Gambling”.