website statistics

Originally published at the Harvard Sports Analytics Collective 

Last month, we tried to assess the probability that front offices from each team would fire their coach at the end of a season. While the methodology is laid out in the last post, we can quick summarize here: we found 14 quantitative variables related to team performance—including wins, career win percentage of the coach, and whether or not he made the playoffs that year—that significantly influenced the probability that a coach would lose his job in the upcoming offseason.

 

 

Parameter

Estimate

P-Value

Intercept

3.656

<.0001

Win Percentage

-8.1205

<.0001

First Year Coach

-2.532

<.0001

Second Year Coach

-1.257

<.0001

Z-Score of Career Win %

-.427

.0003

Defensive Rank of Points

-.0664

.0009

Change in Defensive SRS

-.1106

.0019

Strength of Schedule

-.2069

.0037

Playoff Appearance

.1836

.0144

Change in Offensive Yards

.0314

.0145

Playoff Wins

-2.4111

.0166

Change in Turnover Rank

-.0243

.0194

Third Year Coach

-.7235

.0219

Career Super Bowl Wins

-.5462

.0397

 A limitation of our model is that it takes year-end figures as inputs, which makes it hard to predict midseason coaching changes. In this post, we worked around this shortcoming by downloading Neil Paine’s Week 11 ELO scores and using 538’s formula that converts two teams’ ELO scores to their respective win probabilities. With a reasonable estimate of the outcome of each game from now until the end of the season, we ran a Monte Carlo simulation that outputs the distribution of each team’s 2014 win totals. We ran 1,000 trials and using the 0.025 and 0.975 quantiles found a 95% confidence interval for each team’s win totals. Plugging in the bounds of this confidence interval into our firing model, we translated this 95% confidence interval for wins into one for the firing probabilities for each team’s head coach.


Coach

Team

Probability Mean

(95% Confidence Interval)

Gus Bradley

Jacksonville Jaguars

.551 (.238, .803)

Rex Ryan

New York Jets

.524 (.218, .803)

Marvin Lewis

Cincinnati Bengals

.481 (.091,.646)

Ron Rivera

Carolina Panthers

.420 (.126,.728)

Mike Smith

Atlanta Falcons

.330 (.105, .713)

 

Bradley still holds the greatest chance of getting fired when the end of Week 17 rolls around. However, his chance of getting fired has dropped from 64% to 56%, probability due to the fact that the Jags were able to squeak out one win against the Browns after our initial post. Ryan’s choice to go with Vick got the team a win over the Steelers in Week 10, but did not do much in terms of dropping his chance of getting fired; in fact, it has increased since October, going from 40% to 52%.

After a super hot start, the Bengals having gone 3-3-1 in their last seven games, reducing Marvin Lewis’ chances at retaining his job to a 50-50 proposition. Ron Rivera has led his teams to three wins in a weak division, but may yet dig himself out of the hole as his Panthers are just one game out of the division

Two notable movers are Lovie Smith and Marc Trestman. Smith has a 32% chance of getting fired, which is pretty high considering the grace period first year coaches receive, but a 2-8 record hardly inspires confidence from the Bucs brass. Marc Trestman has a 6.2% chance of leaving Chicago this winter, which is really small, given their embarrassing performances against New England and Green Bay. The number is so low because the Bears, despite the two blowout losses, have actually improved on defense – going from -7.1 on DSRS to -4.1.

With only six weeks left in the season, these coaches might want to implement some new strategies (like the ones Jeff Fisher used against Seattle in Week 7, or make some drastic personnel changes lest they spend the winter looking for new jobs.