# Cornell Football: By the Numbers

Updated: October 25, 2016

The Big Red are off to a good start this year, holding a 3-3 record after the first six games. The team looks to have improved compared to the last couple seasons and is playing strong football. Their success is due to a multitude of reasons, but especially an improved defense.

Several of the most important plays this year have been turnovers caused by this strong unit. The defense has already amassed 14 takeaways, picking off opposing quarterbacks 10 times and recovering four fumbles. These turnovers have led directly to scoring opportunities and have kept the opposing offense off the field.

Turnovers are an important aspect of the game and can be good statistical markers for how well a team is playing. Taking a look back at the Big Red’s turnover stats from previous seasons can help to understand how this year’s team is doing.

The statistics being used date from 1970 to the present and represent the team’s yearly totals. The 46 seasons were plotted on two different scatter plots with the current partial season shown as a red dot.

This first chart compares turnover differential with winning percentage. The scatter plot shows somewhat of a positive correlation, and a linear regression of the data demonstrates that there is a weak positive correlation. This means that the higher the turnover differential, the more wins the team is expected to achieve.

However, along with the regression line is an R2 value, which is meant to show how much of the dependent variable (winning percentage) is explained by the independent variable (turnover differential). An R2 value of 1 would suggest that all of the data were explained by the variable, while a value of 0 would indicate no connection.

The R2 value here is not very high and only about 0.19. At first glance, this number seems to suggest that there is hardly a connection between the two variables, but the data needs to be taken in context. In a football game, there are over 100 plays and countless formations. There is a lot that goes into a win, and the fact that turnovers explain 19% of the team’s winning percentage is actually significant.

It’s also important to evaluate the data in a different manner – through a statistical significance t-test. This test will show if there is a connection between the two variables. The test assumes the data is unrelated and random. Then, it calculates the probability that the data would appear as it does if the situation was truly random.

Chart 1 at the end of the article represents the outputs of the test. It demonstrated that there is a significance level of 0.003, which means there is only a 0.3% chance that the variables are unrelated and a 99.7% chance that they are related. There is a near certain chance that turnover differential can help determine a team’s winning percentage.

However, this simple calculation doesn’t fairly account for blowout games. One day, the team could register a large turnover differential, yet no matter how many turnovers they get, it still only counts as one loss or one win. To make up for this issue, I compared turnover differential with the respective season’s points differential.

The data indicates an even stronger correlation than the previous scatter plot. After doing the regression, the R2 value comes out to 0.28, which, again, shows that turnover differential explains 28% of the team’s points differential. Also, the significance test shows that these two variables are even more closely related. Chart 2 illustrates that these two variables are 100% related (rounded to the third decimal).

There is a clear connection between the turnover differential of the team and its ability to win games. This year, the Big Red sit at a turnover differential of +1, which is not incredible, but is still the program’s 18th best in the last 46 years. There are four games left to play, which leaves a lot of room for improvement. This season, Cornell has a -7 points differential and a 0.5 winning percentage, putting the team in about the middle of the pack historically.

Ultimately, these models cannot truly predict how many points the team will score in a specific game or season, but they can display the importance of maintaining possession of the ball and causing turnovers. The team is in a good position this year, and if they stay mindful of turnovers, the Big Red will continue to have success on the field.