The Call on the Field

The call on the field: in NFL sports betting, we commonly hear that teams coming off bye weeks have a better performance due to the extra rest. Knowing whether this is a myth or not could have a huge impact on your wallet. In this week’s Under the Hood analysis, we will use historical NFL game data to determine if statistics can back the claim.

Approach & Data

We are going to compare the Points For (PF) and Points Against (PA) for games that occured after a team’s bye week to games that occured after a non-bye week. First we need to get some game data. At Hooded Rhino, we have a dataset of all NFL games since the 2000 season. This dataset contains 7978 observations. An observation is a single team’s performance during a game and whether the games was after a bye week or not.

      Team          Season          Week          Points.For    Points.Against      Game.Type   
 NE     : 269   Min.   :2000   Min.   : 1.000   Min.   : 0.00   Min.   : 0.00   After Bye: 534  
 IND    : 264   1st Qu.:2003   1st Qu.: 5.000   1st Qu.:14.00   1st Qu.:14.00   Regular  :7444  
 BAL    : 263   Median :2007   Median :10.000   Median :21.00   Median :21.00                   
 PHI    : 260   Mean   :2007   Mean   : 9.529   Mean   :21.68   Mean   :21.68                   
 GB     : 259   3rd Qu.:2011   3rd Qu.:14.000   3rd Qu.:28.00   3rd Qu.:28.00                   
 PIT    : 259   Max.   :2014   Max.   :21.000   Max.   :62.00   Max.   :62.00                   
 (Other):6404

Normal vs Bye Week Scores

First we will look at the distribution of the Points For (PF) earned by all teams on normal weeks and games after bye weeks. In the plot below, we can see that the distributions are very similar between the two groups. The bye weeks have a slightly higher mean PF with 22.67 and a mean of normal games of 21.6. It looks like on average, teams perform better by 1 point when coming off a bye.

By performing a simple t-test, we can see if there is a difference between the two groups of scores

Null and Alternate Hypotheses

H0: There is not a difference between scores of teams coming off bye weeks and normal weeks in the NFL
(\(\mu\)bye = \(\mu\)normal)

H1: There is a difference between scores of teams coming off bye weeks and normal weeks in the NFL
(\(\mu\)bye <> \(\mu\)normal)

95% T-Test

    Welch Two Sample t-test

data:  byes$Points.For and normals$Points.For
t = 2.3444, df = 616.308, p-value = 0.01938
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
 0.173300 1.961973
sample estimates:
mean of x mean of y 
 22.67228  21.60465

With a 95% confidence interval, we can see that there is a signifcant difference between the two means (p-value = 0.0193763). We reject the null hypothesis, accept the alternate hypothesis, and conclude there is a difference between the two groups of games. However, I don’t know about you, but I don’t think that a 1-point increase after bye weeks is much help when placing wagers. I would like to see if we can do better.

Points Against Comparison

Before we drill down into the 1-point advantage, let’s first check out the Points Against to see if there is any difference.

95% T-Test (Points Against)

    Welch Two Sample t-test

data:  byes$Points.Against and normals$Points.Against
t = -1.8825, df = 615.005, p-value = 0.06024
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
 -1.76346831  0.03728208
sample estimates:
mean of x mean of y 
 20.87079  21.73388

It appears that there is NOT a signifcant difference between the two means (p-value = 0.0602384). We accept the null hypothesis, and we can continue with the offensive side of the ball.

Early or Late Bye Weeks

It may be helpful to see if early or late season bye weeks have any affect on the offensive scores. For this analysis, we will add another column. If a bye week occurs in the first 8 weeks of the season, we will call it an EARLY bye week. Otherwise, we will call it a LATE bye week.

From this plot, we can see that early season bye weeks don’t have much impact on the offensive scores, but there seems to be some difference on the late season bye weeks. Let’s see the impact with another t-test


    Welch Two Sample t-test

data:  Late.Byes$Points.For and Late.Regular$Points.For
t = 3.0846, df = 303.188, p-value = 0.002226
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
 0.7251522 3.2806128
sample estimates:
mean of x mean of y 
 23.58582  21.58294

From this t-test, we can see that the difference in means is about 2 points and the p-value of 0.0022259 is small. That is getting a little more interesting: a two-point average increase in offense production after a late-season bye week.

Win Percentage

Before we make our call, I think we need to look at win percentage. Do teams win more often after bye weeks? This could help with money-line bets. I will two more columns to indicate if the game was a win or loss, and I will calculate the win percentages. In this analysis, ties are considered losses.

  Game.Type       Win
1 After Bye 0.5393258
2   Regular 0.4965073

    Welch Two Sample t-test

data:  byes$Win and normals$Win
t = 1.9154, df = 612.348, p-value = 0.0559
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
 -0.001082413  0.086719590
sample estimates:
mean of x mean of y 
0.5393258 0.4965073

It appears that there is a 4% increase in winning percentage after a bye week. However, the p-value of 0.0559037 the t-test is right on the border. We can’t conclude there is a statistically significant change to the winning percentage.

Win Percentage by Team

One last check. Let’s see if some teams do better after a bye week.

In this plot, you can see that teams like Baltimore, Philadelphia and Pittsburg have better winning percentages after bye weeks, and teams like Carolina and Cincy have worse performances. On the other hand, teams like Arizona and Buffalo are the same.

When we perform a t-test for each team, only one team shows a statistically significant increase or decrease in performance on games after a bye week - Philadelphia

[1] "PHI"

    Welch Two Sample t-test

data:  g$Win by g$Game.Type
t = 3.8072, df = 23.404, p-value = 0.0008857
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
 0.1437838 0.4852336
sample estimates:
mean in group After Bye   mean in group Regular 
              0.8888889               0.5743802

Conslusion

After futher review, the call on the field STANDS. There is a statistically significant difference between regular season games and games after bye weeks.

However, it should be noted that the difference is small. Overall, the average gain in Offense is about 1 point, and 2 points when the bye week occurs in the 2nd half of the season. In addition, some teams have historically made the most of their bye weeks, while other teams have failed to capitalize on the rest.

© Hooded Rhino, 2015