Using Sabermetrics to Project 2017 NFL Team Wins

Earlier this offseason I looked at which teams got lucky in 2016 from a sabermetrics perspective regarding won-loss record. Using each team’s Pythagorean win percentage, we can use their 2017 schedules to give an estimate of wins for each team in 2017.

Calculating Expected Team Wins

Pythagorean wins can be estimated using each team’s Pythagorean win percentage from the year prior. This is done via the Log5 formula. The formula states that if Team A has a Pythagorean win percentage of P­and Team B of P­B, then the probability of Team A winning (denoted W­A) is as follows:

W­A = (P­A – P­A * P­B) / [P­A+P­B – (2 * P­A * P­B)]

These win percentages are how the team would play at a neutral site. I’ve made adjustments so that if a team is at home, their Pythagorean win percentage is increased slightly while the visiting team’s Pythagorean win percentage is reduced slightly. The exact amount of these adjustments is such that it gives the home team an extra 2.7 point boost. From there, we can calculate the expected wins for each team over the course of the 2017 season. It should be noted that it’s obvious that this is strictly a formulaic approach to calculating wins. The formula knows nothing of player development, personnel changes, scheme changes, or coaching changes that might influence the true win percentage of a team. However, since NFL wins are a net-zero sum (there have to be 256 wins every year1), then in the long run, these are true averages that we expect from each team, should the offseason with all the changes be randomized many, many times over. To find Pythagorean Win percentages from 2016, check out this link. Let’s dive into the 2017 projected team wins.

2017 Pythagorean Team Wins
  1. Counting ties as half a win for each team.  (back)

Subscribe to the best value in fantasy sports

You're all out of free reads for now and subscribing is the only way to make sure you don't ever miss an article.

By RotoDoc | @RotoDoc | Archive

Comments   Add comment