Transcript Slides

Betting in Super Bowl match
ups
PRESENTATION TO MIS480/580
GABE HAZLEWOOD
JOSH HOTTENSTEIN
SCOTTIE WANG
JAMES CHEN
MAY 5, 2008
Who did what
2
Literature
Review
Subject
Matter
Expert
Gabe
Hazlewood
X
X
Josh
Hottenstein
X
James
Chen
Scottie Wang
Data
Extraction
Analysis
Statistical
Modeling
X
X
X
X
X
X
X
X
Research Question
3
“Can patterns in historical game performance allow
the bettor to gain a better understanding of what
makes a good bet”
Introduction
4
 Purpose
 Provide bettors with an “angle” that can be used to exploit
certain inefficiencies in NFL betting market
 Objective
 Analyze whether there are any exogenous variables that could
aid in better determining the outcome of a Super Bowl bet
relative to its line
 Usefulness
 Seasoned bettors can add any findings to repertoire for future
use, as it pertains only to a game played once a year
Literary Reviews
5
Walker, Sam. "The Man Who Shook Up Vegas." The Wall Street Journal 5 Jan. 2007. 11
March 2008 <http://online.wsj.com/public/article/SB116796079037267731wjPu4ACcg5J5Qvjh05IYEI_Ooeo_20070112.html>.
1.


Examines success rates of experts in sports betting
Introduces the viewing of betting as an investment rather than a gamble
Gray, Philip K., and Stephen F. Gray. "Testing Market Efficiency: Evidence From The NFL
Sports Betting Market." The Journal of Finance, Vol. 52, No. 4, (Sep., 1997), pp. 1725-1737.
2.


Examines the efficiency of the NFL betting market
Introduces more sophisticated betting strategies (i.e. bets are placed only when there is a relatively high probability of
success)
Gandar, John, Richard Zuber, Thomas O'Brien, and Ben Russo. "Testing Rationality in the
Point Spread Betting Market." The Journal of Finance, Vol. 43, No. 4, (Sep., 1988), pp. 9951008.
3.



Presents empirical tests of market rationality using data from the point spread betting market on NFL games
Examines whether, at any point, a moving line becomes more significant as to the outcome of a bet
Old but NOT outdated
Avery, Christopher, and Judith Chevalier. "Investor Sentiment From Price Paths: The Case
of Football Betting." The Journal of Business, Vol. 72, No. 4, (Oct., 1999), pp. 493-521.
4.


Further examination on previous citation’s findings
Validates that movement of a spread is predictable, and attempting to exploit it yields a very low profit at best
Literary Reviews (cont.)
6
“The Man Who Shook Up Vegas”
 Significant Findings



When betting against a point spread, bettors must win
52.4% of their wagers to make a profit
Experts realize close to 60% winning percentage
Most highly regarded expert is Bob Stoll

Looks for “angles” that predict future results (i.e. team favored
by 7 or more in minor bowl game after losing their last game,
fail to cover spread 77% of the time)
 Use in project
 Only accept findings yielding greater than 52.4% probability;
aim for closer to 60%
 Find “angles” similar to Bob Stoll example; proven effective
Literary Reviews (cont.)
7
“Testing Market Efficiency: Evidence From The NFL
Sports Betting Market”
 Significant Findings



Model indicates that the market overreacts to a team's recent
performance and discounts the overall performance of the team
over the season
Exogenous variables such as rushing/passing yards could be
added to increase the predictive power of the model
Inefficiencies exist, but not all are exploitable
 Use in project



We will use season long stats, taking overall performance into
account
Attempt to find which exogenous variables, if any, will increase
predictive power (angles; consistent with expert methodology)
Look for inefficiency in Super Bowl betting market and if it can be
exploited
Literary Reviews (cont.)
8
“Testing Rationality in the Point Spread Betting
Market”
 Significant Findings

In the NFL, the closing line does not provide a more accurate
forecast than does the opening line; and vice-versa
 Use in project
 Using closing lines, available in our data set, will not
compromise validity of our findings
Literary Reviews (cont.)
9
 NFL spreads are biased predictors of actual results
 Creates inefficiencies
 Certain inefficiencies can be exploited
 Exploit, most profitably, by finding exogenous
variables that provide an “angle”
 Aim for 60% probability, above 52.4% acceptable
 Confidence in data set
Apply to Super Bowl!
Data collection
10
 Data source



Spider data from Databasefootball.com
Collected all game play stats for the 17 regular session games and the
Super Bowl for the last 10 years
Collected betting line and over data for the last 10 Super Bowls
 Collection Technique



Spider data for the site
Load the data into excel workbook
Load work books into respective tools
 Analysis techniques


Tools used SPSS and MathLab
Simple stats, correlation analysis and multi factor statistical
modeling
Simple Stats
11
 Simple Statistics
 Averages of the favorites regular season:
Total
score
Average
Median

28.23529 21.390374 371.973262
28.00
21.00
Time of
Possession
Rush
Attempts
29.5828877
1.317375966
30.00
1.32
377.00
Averages of the underdogs regular season:
Average
Median

First Downs Total Yards
First
Total
Rush
Time of
Total score Downs
Yards
Attempts Possession
23.71134
18.67526
330.7113
30.25258
1.311419
23
19
333.5
208
6.8
Super Bowl averages:
Final Score
Underdog
Average
Median
Favorite
Average
Median
First Downs
Total Yards
Rush Attempts
Time of
Possession
20.18182
20
16.27273
17
314
339
24.45455
22
1.15
1.12
22.90909
23
19.09091
20
352.5455
331
30.27273
33
1.39
1.38
Betting Line Averages
12
Betting
Line
Average
Median
Over
7.636364
7
Actual Over
46.63636
43.09091
48
46
Correlation Analysis
13
 Line to Regular Season Score
Favorite
Underdog
 Over to Regular Season Score
Favorite
Underdog
Complex Statistic Model
14
 Multiple Linear Regression
Factors selected
15
 Average Difference of Each season

Total Yards (X1)-General ability to offense

Time of Possession (X2)-Ability to control the game

Second Half Score (X3)-Ability to adapt and change

Rush Attempts (X4)-How aggressive the team is
 Super Bowl Score (Y)
Regression Process and Result
16
 P-Value for the Favorite Team Analysis
0.0026
0.00558 0.00276 0.0124
Regression Process and Result
17
 Result for Favorite Team
Y=0.129*X1+11.02*X2+1.028*X3+0.792*X4
R Square:0.6969
Conclusion
18
We developed a procedure to help gamblers to make a better
bet:
 Use the Multiple Linear Regression method to calculate the
final estimate result for both the favorite team and
underdog team.
 Calculate the final estimate line and over data.
 Bet when you found the difference is large enough, the
larger difference it is, the larger possibility you will win on
this bet.
Future work and study
19
 Organize some mathematics experts and football experts to
build a model using reasonable and complex method of
Statistical hypothesis testing.
 Using standard deviation to help prediction
 Uncertain factor which would influence the match a lot
such as weather, big event in super bowl team should be
considered in the prediction
Lessons Learned
20
 With the statistical model, we are capable of winning the
profit and the model could be more effective than some of
the expert estimation.
 the gamblers could use our method to exploit certain
inefficiencies in NFL betting market and make profit of
them.