Football Spreads

20
Spread Football Analysis Tim Hoolihan [email protected] @thoolihan

description

Presentation from 11/19/2014 at Cleveland R User Group on using R to improve and track spread NFL picks throughout the season.

Transcript of Football Spreads

Page 1: Football Spreads

Spread Football

AnalysisTim Hoolihan

[email protected]

@thoolihan

Page 2: Football Spreads

What Is A Spread?

• Point Differential

• Added to favored team to determine adjusted score

• Attempts to get 50% of betters on each side of the bet*

• Different than trying to be accurate. For example: large fan-base skew

• Rule of Thumb: Home Team starts with -3

Page 3: Football Spreads

Goals

• Come back from a horrible start

in my league

• Pet project in R that motivates

learning more

Page 4: Football Spreads

Track

• Google Spreadsheet

• Download as CSV

• Blank copy you can edit:

http://bit.ly/1xLvg64

Page 5: Football Spreads

Functions For Game Types

Page 6: Football Spreads

…continued

Page 7: Football Spreads

League Trends

• rmarkdown

• convert to html

Page 8: Football Spreads

Home vs Away

Favorite vs Underdog

• First meaningful insight

• It appears away favorites are a

better pick

• Why?

Page 9: Football Spreads

Testing The Math

Page 10: Football Spreads

Spread Performance By

Team

Page 11: Football Spreads

My Trends

Page 12: Football Spreads

Upcoming Games

Page 13: Football Spreads

My Results & Progress

Page 14: Football Spreads

Residuals

Page 15: Football Spreads

Packages

• rmarkdown

• scales

• dplyr

• ggplot2

• gridExtra

Page 16: Football Spreads

dplyr

• Workshop exposure

• chaining, functional

• Linq in .Net

• Closures (JavaScript, Ruby)

• Domain Specific Language - like (see Residuals.R)

Page 17: Football Spreads

ggplot2

• Nice, but challenging

• qplot vs ggplot

Page 18: Football Spreads

Have I Learned Anything?

• Yes, a lot more R

• League position improved from

basement to basement stairs

• Don’t pick home underdogs

Page 19: Football Spreads

What Comes Next?

• Calculate Correct, other columns

• Clustering by spread size

• Model training (machine learning… caret?)

• My performance vs recommendation

• Break out web pages further (league, my performance,

next week)

• Shiny?

Page 20: Football Spreads

Your Turn

• https://github.com/thoolihan/FootballPicks

• Google Docs http://bit.ly/1xLvg64

[email protected]