Skip to main content

Are You Getting Excited/Scared Because of NFL Preseason Performance? Don’t.

This weekend, the NFL preseason kicks off in Canton with the Colts and Packers. Also, the Olympics, International Champions Cup, and the Indiana State Fair are happening. So, it’s a busy time in sports and artery destruction. So definitely it’s going to be entertaining, and maybe it will impact your health in 20 years (I can dig some football, futbol, sports only on TV every four years, and random fried things). But, as you see your favorite NFL team founder, excel, or be somewhere in the middle over the next month, how much should you draw from that in terms of the next few months of your life?

Some may say very little because it’s 4 (or 5, for the Hall of Fame Game teams) games, mostly featuring back ups. And of course, the 2008 Detroit Lions managed to win every preseason game and lose every regular season game. But that is hardly a deep empirical analysis. Of course, with the limited sample size, a strict, actual winning percentage is a bad measure of team performance. Winning half your preseason games and winning 3 out of 4 is a rather similar performance, much more similar than regular season records with the same winning percentages, 8-8 and 12-4.

So, how can the performances of a team in the preseason and regular season? One tool, originating in baseball and adapted to football by Football Outsiders, is the Pythagorean expectation, which uses the points scored by and against a team to predict the record (in this case, I’m using winning percentage because they can be comparable independent of number of games played) of a team. Since there is a greater variation in those values between teams in preseason, it should allow a more realistic range of potential results. And, because the values are also very easily able to be calculated for both the preseason and regular season, it should allow a large number of data points to be compared. For football, the equation is as follows:
Pythagorean Expectation=((Points for)^2.37)/((Points for)^2.37+(Points against)^2.37)  
So, how does the preseason stack up as a predictor? Using the NFL’s data from the 2006-2015 seasons, this comparison can be easily performed (well, aside from a lot of copying and pasting from NFL.com). Not particularly well, actually. Well there is a slight correlation (a Pearson coefficient of .261), but as can be seen below, it is much more of noise than a pattern.

There are a mix of teams that excelled in the preseason and tanked in the regular season (2008 Lions), did poorly in the preseason and well in the regular season (2011 San Francisco 49er’s), and did roughly the same in both (2013 Seattle Seahawks).

So if your favorite team is doing very poorly or very well, it means nothing, right? Not exactly. There is a significant (Pearson of -.684) between a team’s preseason Pythagorean expectation and the difference between the preseason and regular season record. As shown in the graph below, this effectively means that teams that do either very well or very poorly in the preseason are likely to not be as big of a success or failure in the regular season.


Teams, on average, trend to being pretty average. So if your favorite team absolutely faceplants during the preseason, don’t be alarmed. They will probably not be terrible, just most likely mediocre. And potentially a playoff team. Which is better than the alternative. Sorry Lions’ fans.

Comments

Post a Comment

Popular posts from this blog

2017 Weekly Boiler Stat Summaries: Week 13, v. IU

This is the first time the Bucket has been home since I was a junior at Purdue – the first year of Darrell Hazell’s tenure in West Lafayette. I am not nearly a good enough writer to capture my happiness in this moment. Purdue spent my entire post-college life (and half of college) being a joke nationally with very little to be proud of on Saturday afternoons in the fall. In just one year the narrative has changed. The Boilers won five games against FBS competition (a winning percentage of 11.36%) and three conference games (9.09%) during the 48-game Hazell era; in a single season, Brohm has won six FBS games, including four in conference, for a 50.00% winning percentage overall and 44.44% in conference. And with 6.8 second order wins (wins expected based on per play success rate) per S&P+, one may say that Purdue underachieved in the win column compared to how the Boilers played. Jeff Brohm has done something incredible this year, and capped it off with a win against IU – finally b...
Post a Comment
Read more

Making a Spreadsheet, Checking It Twice, Gonna Find Out When It's Socially Acceptable to Listen To Christmas Music

It’s that time of year again. And by that time, I mean the time when you aren’t sure if it’s too early to hear Christmas music yet. Is it OK to walk into a certain coffee shop chain in early November and grab a seasonal beverage that can barely call itself coffee? Is it too early to hum a carol when I’ll be waiting to get on a plane the day before Thanksgiving? While driving to your favorite Black Friday activity (the correct answer is skeet shooting or watching football), should the speakers be spreading holiday cheer? Or do you need to wait until you break out the advent calendars to put on a festive record? To answer this age old question, I’m trusting the fine Google searches of my fellow Americans. Using Google Trends data from 2012-2016, I’m going to compare each week of the years’ average searches for the term “Christmas music” as a percentage of the maximum to determine when Americans start listening to Christmas music [1]. Figure 1: Christmas Graph, Oh Christmas Graph ...
Post a Comment
Read more

This Year's Bowl Season Was As Bad For The Big Ten As It Looked.

After this year's bowl games, the narrative that developed among the sports media was the Big Ten, after what seemed like a historically strong regular season, proved to be weak after a terrible bowl performance. But how bad was that bowl performance? To find out, I compared the actual performance of the conferences to a set of 10,000 simulations of this year's bowl games, excluding this Monday's National Championship Game. To create the simulations, I used log5 and Pythagorean expectations, as I did previously for the NFL and MLB . However, because of the variability in strength of schedule in college football, using actual points scored for the Pythagorean expectations would be inaccurate. Instead, I'm using the Football Outsider's S&P+ ratings of offense and defense  (referenced on 1/4/2017), which adjust for strength of schedule and account for efficiency as well as performance.  To rate the performance of each conference, I found the percentile of t...
Post a Comment
Read more