Tuesday, February 2, 2016

Testing Home Field Advantage by Team in the NFL


Testing Home Field Advantage by Team in the NFL


Issue:  Is it possible to use team specific data for only one season to test for home field advantage in the NFL?

Discussion:   Sample size is an obvious potential problem.   With only eight games at home and eight games on the road one extra home win will alter the difference between home and road win percentage by quite a bit.  But a difference between proportions based on a difference of only one or two “successes” is not likely to be significant.

Sample size is important.   Let’s illustrate the issue with data from the New England Patriots during the 2015 regular season.

Question:   The New England Patriots had a 7-1 record at home and a 5-3 record on the road during the 2015 regular season.   Was the New England home win proportion significantly higher than the road-win proportion?

Answer:   This question was addressed with the chi-square test on differences in proportions.  I developed a calculator for this test in a previous post.   The previous post was applied to a single major league baseball team over a 162 game season.




Will this procedure provide useful results for one team over the shorter football schedule?

The chis-square test calculation for the New England Patriots is presented below.



Calculation of chi square test for difference in home versus road win percentages
Team
New England Patriots 2015 Regular Season
Observed Win/Home
7
Observed Lose/Home
1
Observed Win Away
5
Observed Lose Away
3
Total Home
8
Total Away
8
Total Observations
16
Pooled Win Prob
0.7500
Pooled Loss Prob
0.2500
Expected Win/Home
6.00
Expected Loss/Home
2.00
Expected Win/Away
6.00
Expected Loss/Away
2.00
(OWH-EWH)2/EWH
0.17
(OLH-ELH)2/ELH
0.50
(OWA-EWA)2/EWA
0.17
(OLA-ELA)2/ELA
0.50
Chi Square
1.33
P-value
0.248
Diff.  Prob Win Home- Prob Win Away
0.2500


The New England Patriots had the best overall win percentage of all AFC teams during the 2015 and were only one game above 500 on the road.  Despite this stark differential, the New England Patriots home win percentage is not statistically significant than its road win percentage.


The NFL regular season sample size is too small to test for team-specific differences in home field advantage based on one season.

It may be possible to combine data from more than one season but this has to be done with care because team make up, coaching staffs, philosophies, and performance change over time.   We can discuss the possibility of pooling data across seasons in a future post.

Go back for home field advantage testing problems:
http://www.dailymathproblem.com/p/home-field-hypothesis-testing-problems.html


No comments:

Post a Comment