A Deeper Look at Projecting the Philadelphia Eagles’ 2013 Season
by Jerome's Friend
Last week, on vacation, there was a moment when I stood at the edge of the Atlantic Ocean and started to think about the Philadelphia Eagles (note to wife: I thought about you more!). It seemed to me a nice metaphor for this new Eagles team and the upcoming season. As the team heads into the season’s first preseason game, we’re not yet knee-deep in Chip Kelly’s tenure. I don’t even know if we’re ankle deep. Instead, we’re probably still toeing the water, testing the temperature, not knowing exactly what lies beneath the deeper waters ahead. But it’s really hard to not get caught in the beauty of the view, or get excited about moving forward.
If you discount torn anterior cruciate ligaments and racial slurs, there is still a lot of excitement. My own excitement for this Eagles’ season was fueled by the Eagles chapter in the 2013 Football Outsiders Almanac (it serves as a great companion to BGN’s Eagles Almanac… pick up both). According to their one million(!) simulation runs, the Eagles have a 10% chance of achieving 0-4 wins, a 35% chance of achieving 5-7 wins, a 42% chance of achieving 8-10 wins, and a 13% chance of 11 or more wins. These simulations are based on Football Outsiders’ native Defense-Adjusted Values Over Average (DVOA) statistic, as well as values assigning weight to factors such as home-field advantage, weather, dome, coaching, and other variables, applied to each game of the 2013 seasons (again, one million times!). I will forever be an optimist, someone who views the glass as half full, but even I find it hard to accept that the Eagles have almost an 80% chance of getting between 5 and 10 wins, or a less than 10% chance of equaling their 2012 win total.
Much of any optimism obviously stems from the presence of Chip Kelly, NFL revolutionary. However, new coaches in the NFL do not necessarily have the positive impact we expect to see, especially during their first season. According to a Football Freakonomics study, there are three primary reasons for this:
- Regression to the mean: teams that have done very badly for a long time are more likely to win a bit more in the future, whether they get a new coach or not. Sadly, the opposite is also true for winning teams.
- Most former NFL Coaches of the Year are eventually fired. Did they suddenly forget how to coach? Did their brilliant strategies evaporate? Or, more likely, was their former winning a consequence of a lot of factors that went well beyond coaching?
- It is hard in general to satisfactorily measure leadership – whether we’re talking about a football coach, a CEO, or the President of the United States – but a variety of empirical research shows that an institution’s top man or woman is seldom as influential as we think. It’s a natural inclination to pin a lot of blame (or, occasionally, glory) on the figurehead. But just as the President doesn’t actually have much control over the economy, a football coach has limited control over his team’s outcome.
If Chip Kelly’s impact on the Eagles’ success during his first season may be minimized, what exactly could the season look like? I conducted my own simulation to find out. Unlike Football Outsiders (FO), the basis for my model is Pythagorean expectation. The correlation for FO’s DVOA to wins for the following year is .374 while the correlation for Pythagorean wins (expectation) to wins for the following year is .324, just as adequately significant.
My model takes the Pythagorean wins for the Eagles and their opponent each week, randomizes those values within the standard error value (2.6 wins), and applies a random “external factor” value to each team. I included the external factor value to randomly, and aggregately, account for things like luck, home field advantage, day vs. night games, win/loss streaks, weather, game planning, the impact of a certain new head coach, etc. (This is similar to, but probably different from what FO has done.) I ran the simulation 1,000 times (sorry, couldn’t do a million!), so the external factor value changes for each team in each simulated game, each simulated season. For each game, the products of the modified Pythagorean wins (based on standard error) and external factor value for each team are compared, and the team with the higher product is awarded a win.
Over 1,000 simulations, the Eagles reached five wins 21.7% of the time, six wins 20% of the time, seven wins 15.7% of the time, and four wins 14.7% of the time. Further, we can look at the areas under the curve to determine the likelihood the Eagles will fall within a range of wins:
According to this graph, the Eagles have a 56.9% chance of reaching 4 to 6 wins, a 50.8% chance of reaching 5 to 7 wins, a 36.5% chance of reaching 6 to 8 wins, and a 21.7% chance of falling within the 7 to 9 win range. More importantly, in my model, the odds that the Eagles will achieve 5 to 10 wins drops to 61% from Football Outsiders’ 77%. To me, this seems to be a more likely result (but I’m admittedly biased).
At any rate, based on their schedule and Pythagorean expectation, the Eagles do seem likely to improve in 2013. In the end, projecting wins is really irrelevant. Which result you choose to believe is up to you. At this point though, it may be better to just toe the water, walk coolly forward, and admire the view in front of us.
Here is a summary of how the Eagles performed in my simulations against each 2013 opponent: