When Alex Smith travels to Seattle, he has to go up against 67,000 screaming Seahawk fans that make Seattle one of the loudest stadiums in football. When Joe Flacco goes into Pittsburgh, he has to overcome 65,000 Steelers fans clad in black and gold and waving Terrible Towels. And when Matt Schaub plays in Jacksonville he has to, well...people do go to football games in Jacksonville, right?

Either way, all three scenarios have one thing in common. The home field advantage is exactly the same.

Whether you have a sold out stadium full of rambunctious fans, or the stadium is half full, the home field advantage is the same. Although data analysis has shown that home field advantage is real, the reason is still a bit of a mystery. However, there are some interesting theories out there. But hawks and doves aside, an article came out recently from Advanced NFL stats with some analysis about home field advantage on Thursday night NFL games. Specifically, home teams appear to be gaining even more of a home field advantage than normal. The reason being that if you play a game the previous Sunday, it’s a short week to have to prepare and play Thursday. And having to travel on top of having a short week may affect the visiting team more than usual.

So naturally, I decided to use Minitab Statistical Software to break down the numbers and see if there really is more of a home field advantage on Thursday nights. You can get the data I used here.

# Do Home Teams Win More on Thursday Night?

This recent article said of the 55 Thursday night games since 2006 (counting Cowboy and Lion Thanksgiving Day games, but not counting NFL season openers), the home team has won 35 times. This equates to a winning percentage of 63.6%. Overall, home teams win 57% of the time in the NFL. So home teams do win more on Thursday night, but is the difference statistically significant? We can use Minitab’s One Proportion test to find out.

The p-value is 0.196, which isn’t less than 0.05. So we cannot conclude that the proportion of games home teams win on Thursday is significantly greater than the normal value of 57%. But I’m not ready to give up yet! There are plenty of more statistics to look at.

# Do Home Teams Score More Points on Thursday Night?

Instead of just looking at whether the home team wins or loses, let’s look at by how many points they win or lose. The article says home field is worth about 2.5 points. So in these 55 games, is the home team outscoring its opponent by more than 2.5 points? I took the winning margin of the home team for all 55 games (with a negative value when they lost) and performed a 1-Sample t-Test on the data.

We see that, on average, the home team wins by 4.24 points on Thursday games. But just like before, our p-value isn’t below 0.05, so we can’t conclude that this is statistically greater than 2.5.

All right, I’m going to try one more statistic!

# Do Home Teams Cover the Spread More Often on Thursday Night?

The gambler in me wants to see if an edge can be made against Vegas by looking at the betting lines. I went back and collected the spread on all 55 games, then compared it to the winning margin to see if the home team covered the spread or not. I used Minitab’s Tally command to summarize the results.

The home team *has *covered the spread 31 out of 54 times (taking out the 1 push), good for 57.4%. We would expect teams to cover the spread about half of the time. So once again, we can use the One Proportion analysis to determine if this is significantly different form 50%.

And for a *third *time, the data don’t give enough evidence to show that the value of 57.4% is statistically greater than 50%. In each instance we weren’t able to conclude that home teams have a greater advantage on Thursday night than they normally do.

So this means that we can conclude that they have *the* *exact same advantage*, right? Well, no, we can’t.

# We Need More Power!

Power is a statistical hypothesis test’s ability to detect a difference when one truly exists. For example, say somebody hands you a deck of 1,000 playing cards, where 60% are black and 40% are red. But they don’t tell you if there are more black than red cards; it’s up to you to find out. Instead of counting all 1,000 cards (who would want to do *that?*) you take a random sample of 20 cards, and find that 12 of them are black. Perfect, that’s 60%! Except when you perform a One Proportion test, you get a p-value of 0.503, meaning you can’t conclude the proportion of black cards is significantly different than 50%. The problem is that your sample is too small. Minitab’s Power and Sample Size analysis can show it.

Here we see that we’re testing whether the proportion of black cards is greater than 50% and we took a sample of 20 cards. What the power values tells us is that even if the true proportion of black cards is 60% (the **Comparison p**), our test would only have a 22% chance of detecting a difference.

Now we can apply this to our football scenario. Let’s go back to our first question: “Do home teams win more on Thursday night?” We want to test whether the proportion of home teams that win on Thursday night is greater than 57%. Our sample size is 55, and our comparison proportion is 63.6%. What is the power?

The power is 0.25. This means even if the true proportion of games that home teams win on Thursday night *is *actually 63.6%, the test would only have a 25% chance of detecting a difference. So Thursday night teams many indeed have a greater advantage, but we don’t have enough data to detect it. So let’s see how big our sample size would have to be in order to obtain a power of 80%, which is an accepted standard in the statistical world.

Wow, we see that we would need a sample size of 342 games! That’s going to require another 18 seasons, so it’ll be 2030 by the time our test has an adequate amount of power.

However, the Dallas Cowboys and Detroit Lions have been playing games on Thursdays since the 1960s. Including their games (prior to 2006) will give us about 100 more data points. It won’t get us to 342, but it’ll definitely give us some more power. So I’ll include them, and come back in 2 weeks to see if the extra data tells us anything different!