The NFL season still seems young. There is still a lot of questions and optimism and a bunch of what-ifs that are still to be decided. But in reality, four games is 25% of a team’s season. So what can we take away from this 4-game sample size? Let’s take a mathematical approach (ah! run!) to see how predictive four games of the NFL season can actually be.
Here is the process that I used. Using Bet Labs, I was able to find the win percentage and pythagorean win percentage for each NFL team going into their 5th game:
I can then compare the win percentages and pythagorean win percentages after four games to those same percentages at the end of the season. We’ll use the correlation function to measure on the -1 to 1 scale. What that means is that correlation can range from -1, which means inversely correlated, to +1, which means 100% positively correlated. A measurement of zero means there is no correlation in the sample at all.
Here are the correlation results:
|4 Game Win %||4 Game Pythag %|
|Full Season Win %||.60||.63|
|Full Season Pythag %||.54||.66|
At .66, the highest correlation is between the 4-game pythagorean record and the full season pythagorean record. This is usually a better indicator of team success than their win-loss record but it does have some flaws. It does not account for strength of schedule. The pythagorean record could be improved by adjusting for the opponent. So while it’s not perfect, it’s still a great predictor for a team’s record. If you look at the correlation at the end of the season between a team’s win % and their pythagorean %, the coefficient is .93 which is very high.
Here are the top ten teams in pythagorean win percentage this season (excluding Eagles and Packers who haven’t played four games yet):
- Vikings 79.2%
- Broncos 78.7%
- Steelers 67.1%
- Patriots 66.2%
- Cowboys 65.5%
- Bills 64.2%
- Falcons 61.8%
- Ravens 59.0%
- Cardinals 58.2%
- Chargers 56.7%
Arizona and San Diego both find themselves in the top ten despite having 1-3 records. Arizona’s is inflated due to a giant blowout of the Buccaneers, but the Chargers should definitely have 3 or 4 wins this season and they may be undervalued going forward based on their record. (Although losing a player to an ACL tear each week isn’t going to help.)
The Vikings and Broncos are in the top tier and both are doing it with their defenses. Both teams were favorites to win their conference in our latest simulation results.
Pythagorean win percentage is a fun statistic I like to use throughout the NFL season to find teams that are overperforming or underperforming their current record and Bet Labs has a few filters you can use to do research at any point during the season. If you are interested in doing some of this research on your own, click this link to take advantage of our 6-day trial.