Predicting the outcome of sports events is easy, right?  Just ask around, everyone has an opinion they feel pretty confident about.  Funny how so many people can feel so confident about something, but still be wrong.  It’s no secret the majority of people don’t make money when they go to the sportsbook.  If they did, they would quit their jobs and bet on the games for a living.  The simple fact that they get up and go to work every morning should be an indicator that they can’t consistently win.  But what about the guys who do it for a living, the ones that claim to be professional handicappers, what do their records look like?

The explosion of technology in the last two decades has benefited the handicapping industry.  Powerful software and statistical packages have given another tool to handicappers as they develop their systems.  More and more successful handicappers have been moving away from alleged ‘inside’ information and instead basing their picks on statistical analysis.  While some have argued that professional sports are too much of a random walk to be predicted by formulas, the results are startling.

The book that showed the public the type of hidden value that can be seen in statistics was Moneyball: The Art of Winning an Unfair Game by Michael Lewis.  Moneyball is the story of how the 2002 Oakland Athletics amassed a mountain of wins while having the smallest player payroll in all of Major League Baseball.  By focusing on an analytical, evidence-based, statistical approach they were able to become more successful than ever before.  But the Athletics didn’t just go to the normal box scores to get their statistics; they took a unique look at the game when it came to the numbers, sabermetrics.

Sabermetrics is a specialized analysis of baseball using objective in-game data.  What sabermetrics gave us was a way of seeing the game that nobody thought of before.  They created their own statistics, things like Wins Shares, Runs Created and Total Player Rating.  An age old adage is; Statistics lie and liars use statistics.  However, the truth is numbers can never lie, they are only numbers.  The skill lies in knowing what numbers to evaluate, and how to evaluate them.  By objectively looking at the game the Athletics were able to find were value was going unrecognized by traditional stats and because of that, being undervalued.

The same principles that allowed the Athletics to find valuable, underrated players in Major League Baseball apply to sports betting systems.  Sports are events which have many variables interacting together.  Identifying the right ones and analyzing them correctly becomes critical if you expect to beat the line on game day.

Moneyball started a trend that spread around to every major sport.  Statisticians began to analyze football and basketball, professional and college, the same way others had been doing in baseball with sabermetrics.  Jeff Sagarin is one of the more famous sports statisticians of recent years.  His rating systems have been regularly published in the USA Today since 1985.  One of the systems he developed, called “Predictor”, compares teams and declares a winner according to victory of margin.  His systems are among the most famous and Predictor is one of the few that will allow bettors to predict a game according to the spread instead of “straight up”.  Most systems only predict who will win a game, usually by providing a percentage, or likelihood of a team winning.  Sagarin has one of these too, he calls it Elo Chess, and it is assumed to be based on the Elo rating system used among chess players.  It’s a system that ranks players only according to wins and losses completely disregarding margin of victory.  However, prominent people in the sports industry have begun to stand up and take notice of the statistical revolution.  No longer do guys like Sagarin have to deal with their formulas being relegated to academic articles and the back pages of sports sections in newspapers.

Dallas Mavericks owner Mark Cuban is always trying to be on the cutting edge.  When he purchased the team he immediately began to redesign the locker room, travel facilities and anything else he thought would help the team towards its ultimate goal: winning the NBA championship.  By always looking for an edge it didn’t take long for Cuban to hear about the value statistics could provide to his team.  Cuban hired Sagarin, and his former classmate at MIT Wayne L. Winston, for their advice.  The pair developed a system called Winval which the Mavericks used to help them determine lineups and who to sign in free agency.  What was once strictly the realm of scouts and insiders has spilled over into statistical metrics we can break down and analyze.

Brian Burke has quickly become one of the more prominent statisticians for his advanced NFL insights.  He is a former Navy F/A-18 carrier pilot turned applied mathematics guru.  Brian started a website called Advanced NFL Stats.  On it he analyzes many small situations in games that have potentially large effects on the outcome.  However, he doesn’t just recognize situations, he looks at them through the bigger picture of what they add to winning the game, and what the other potential options would have been in that same situation.
In one article asks the question of whether or not coaches should go for the 2-point conversion instead taking the extra point.  In a style that is similar to that of Sabermetrics, Burke arrives at some data that challenges long held beliefs.  Yet, as interesting as determining the optimal run-pass balance is for an NFL team, that is not what Burkes greatest contribution is.
One of his main goals when beginning statistical analysis of the NFL was to determine a formula which could predict the outcome of games.  Even though Burke has a BS in aerospace engineering, it isn’t as easy as it sounds.  Many of the formulas that people use have the quality of being good retrospectively, not predictively.  It may not seem like much, but it is the difference between being able to explain why something happened, and what is going to happen next. 

Burke gives a pretty rigorous explanation about his methodology and how he went through the trial and error process.  He began with the idea of a basic logistic regression model:
Y = B0 + B1X1 + B2X2 + …BnXn + ei
It seems pretty basic, and actually it is.  There are numerous statistical analysis programs available that can run a regression formula on statistics from games.  Yet, the data that the software gives back to you won’t get you very far.  What separates good statisticians from everyone else is their ability to correctly interpret the data, identify and weaknesses and make the necessary adjustments for the next model.  This happens to be one of the areas where Burke stands head and shoulders above most others in the field.

Burke claims a very impressive winning percentage with his formula, 69.8%.  However, it must be noted, his predictions are done straight up, no point spread involved.  Unlike most people in the field, Burke gives us his formula after he ran all the numbers.


VARIABLE

COEFFICIENT

STD ERROR

T-STAT

SLOPE at mean

constant

-0.26

1.36

-0.19

AHOME

0.74

0.09

8.29

0.19

AOPASS

0.45

0.07

6.56

0.11

AORUN

0.27

0.1

2.65

0.07

ADPASS

-0.54

0.09

-5.9

-0.13

ADRUN

-0.21

0.11

-1.87

-0.05

AOINTRATE

-15.9

6.26

-2.54

-3.98

ADINTRATE

17.68

5.16

3.43

4.42

AOFUMRATE

-20.5

7.79

-2.63

-5.12

APENRATE

-1.49

0.72

-2.07

-0.37

BOPASS

-0.45

0.07

-6.54

-0.11

BORUN

-0.27

0.1

-2.64

-0.07

BDPASS

0.53

0.09

5.83

0.13

BDRUN

0.2

0.11

1.79

0.05

BOINTRATE

15.71

6.26

2.51

3.93

BDINTRATE

-18.95

5.16

-3.67

-4.74

BOFUMRATE

21.01

7.79

2.7

5.25

BPENRATE

1.47

0.72

2.04

0.37


How is it that sports betting markets can be beat?  Most people assume that if the casino is willing to offer you the bet, they must have an advantage, right?  Of course they do, it’s called the juice.  On average bets at the sportsbook pay out -110.  That means that if you win your bet you will win $100 for every $110 you wager.  That extra $10 is the juice and it’s how casinos can guarantee themselves a profitable venture.

Imagine a two person scenario:  I place a $110 wager on Team A.  You place a $110 wager on Team B.  The casino doesn’t care who wins, they take $100 from the losing wager and pay the winner, keeping the $10 juice.  This gives them an effective profit of 4.54% of all wagers taken in (10/220). 

This is a very simple explanation and more rigorous analysis is given in a variety of articles.  Some systems have even been granted patents by the United States Patent Office.  However, our example does illustrate that the casino does not care who wins.  Sports betting is a pari-mutuel betting exchange and as such it responds to the money wagered in real time.  A casino needs a balanced book to make a profit and it accomplishes this through line movements and changing the amount of the juice.

Once we understand how a casino makes money through their sports book, we can see the opportunity for value in a system.  Both the statistician and the casino are involved in sports betting to make money.  The casino does so through the juice, the bettor through winning enough of his bets.  The casino develops a mathematical system that analyzes the bets as they come in and alters the line and juice to encourage a balanced book.  The applied mathematician however devises a system that measures the true value of teams.  When there is a big enough difference between the two, there is value.  Why would there be a big enough difference to matter?  Because the casino sets their book according to what the general public will do, not the informed mathematician.  In essence, to have a successful system you only need to consistently be able to beat the general public.  Then both you and the casino win.   
This brings us to two questions: First, if the average bettor doesn’t have an advantage, why do they continue to bet?  The answer has two parts, first is enjoyment.  Many people gamble with the hope to make money, but not an expectation.  They take a vacation to Las Vegas and play the slots, blackjack and craps knowing that the casino has a built-in advantage with all those games.  However, they have fun while they do it.  They aren’t there with the single focus to make money, they’re there to enjoy themselves.  Wagering on the game is no different; suddenly the game becomes much more interesting when you have $100 on the outcome.

The second reason people continue to bet is due to cognitive biases.  A cognitive bias is defined by Wikipedia as “a pattern of deviation in judgment that occurs in particular situations, which may sometimes lead to perceptual distortion, inaccurate judgment, illogical interpretation, or what is broadly called irrationality”. 

This most obvious example of this in the general sports gambler is betting on his favorite team.  Regardless of anything else he wants to see his team win and will be bias towards them.  This can go hand in hand with enjoyment because of the added bonus of winning many when your favorite team wins.  It also occurs in the case of injuries.  While it can be hard to quantify the value one player brings in team sports, many formulas have been created.  The general bettor will usually place an incorrect value on what an injury means to the outcome of a game.

The second question is: If a system can win, why aren’t there more of them out there?  This answer also has two parts.  First, it ain’t easy.  A quick look at the records of even famous handicappers will show you that finding a consistent winner is extremely rare.  Yet, the real reason is there is a limit on the use of a system.  For example: Why don’t hedge fund managers employ statisticians to devise a formula to beat the spread and invest a portion of the fund in that as they do individual stocks?

Sports betting is a pari-mutuel betting exchange and moves in real time dependent on money coming in.  The stock market is another example of a pari-mutuel system.  As more people buy or sell a stock, its price fluctuates on the exchange.  The stock exchange is massive, however with sports betting this places a limit on the amount of value that can be extracted from any one situation.  This effect occurs in direct correlation with the size of the market.  An accepted figure in the industry has a small hedge fund as having under $100 million in assets.  Imagine attempting to walk into a casino and wager even 2%-3% of that on a game, or across multiple games.  Then, think of all the hedge funds that could possibly get involved.  The casinos would likely not even take such a large amount of action, but even if they did all the added “smart” money would move the market to a point of efficiency and all the value would disappear.

Numbers like this don’t apply to many people, over $3.2 billion was wagered on sports in Nevada last year alone.  There are not many out there who could wager money, even on individual games, that could make an effect on that.  However, this brings us to our question of sports betting as an alternative investment vehicle.  Despite all the recent bad news about the stock market, returns are very consistent over the long term.
A chart compiled by simplestockinvesting.com shows the averages for the S&P 500:

 

Price
Change

Dividend
Dist. Rate

Total
Return

Inflation

Real
Price Change

Real
Total Return

1950's

13.2 %

5.4 %

19.3 %

2.2 %

10.7 %

16.7 %

1960's

4.4 %

3.3 %

7.8 %

2.5 %

1.8 %

5.2 %

1970's

1.6 %

4.3 %

5.8 %

7.4 %

-5.4 %

-1.4 %

1980's

12.6 %

4.6 %

17.3 %

5.1 %

7.1 %

11.6 %

1990's

15.3 %

2.7 %

18.1 %

2.9 %

12.0 %

14.7 %

2000's

-2.7 %

1.8 %

-1.0 %

2.5 %

-5.1 %

-3.4 %

1950-2009

7.2 %

3.6 %

11.0 %

3.8 %

3.3 %

7.0 %

      
Figuring the juice at -110 a sports betting system needs to win at 52.38% to break even.  Assuming any system seeks to maximize profits while limiting the potential for ruin we can safely assume they are using a proportional money management system.  The Kelly criterion (or a fractional Kelly strategy) is a proportional strategy shown by numerous authors as the optimum money management strategy for betting. These authors include; Breiman, Hakansson, and Thorp (1975), Bell and Cover (1980), Ethier and Tavare (1983), Finkelstein and Whitley (1981), Friedman(1981), Griffin (1984) and MaClean, Ziemba and Blazenko (1987).  Using a Kelly criterion calculator we can see what are expected rate of return will be with any sports betting system projections.  For example;  assuming a 55% win rate, 7 wagers and 12 consecutive series we see an expected rate of return of 25.94%.  That is much higher than anything we can expect to see from the stock market.

Sports betting systems have been around for some time now.  The most famous may be the Computer Group in the 1980’s led by none other than Billy Walters.  Today a quick glance at a site like The Prediction Tracker shows a list of people who have started systems.  The advancement of technology has allowed almost anyone with an idea the ability to put it out there and see if it stands up.  There will always be people who distrust systems, who say the only way to correctly handicap a game is by watching them all and having an inside knowledge of the game.  However, we can see that isn’t the case anymore.  Any sports betting system has the ability to be good or bad.  The power of any model lies with the person who created it.  Finding a good handicapper isn’t easy, but adding sports betting to your investments portfolio is certainly worth the risk.