Monday, December 23, 2013

NFL Coaches on the Hot Seat - A final look at our model

There's only one week left in the NFL regular season, and while a good number of coaches are focused on getting their teams into the playoffs, there are another set of coaches who are left to sit and wonder whether or not they'll still have a job after the season is over.

Those are the coaches I focused on when developing the logistic regression model behind my NFL Coaches Hot Seat Index. And even though there's one week left of action (plus tonight's MNF game), I felt as though the results this week would really be enough to give us a final outlook on who might be staying and who might be going.

To refresh - a brief explanation on the model itself. I was always disappointed with the NFL coach hot seat rumor mill that would creep up over the second half of each NFL season. Not because I don't like speculation, but because it was never really based on anything concrete, anything objective. With that in mind, I gathered historical data on NFL coaching performance going back several decades, and ran some analysis to find statistically significant factors in when coaches get fired.

The two factors I found, and the ones that found their way into my predictive model, were point differential and win change from prior year.

I ran the first analysis after Week 7 of this season, when team's had enough action to give some initial impressions. But now, with most teams having played 15 of 16 games, it's safe to say we have a much clearer picture.

The chart below has our updated data. The model estimates the odds (out of 100%) that a coach will be fired, given his team point differential and expected win change. The win change number is still an estimate (as per Football Outsiders) given more games to play.

This chart is also a little different from the prior version, as I've included not just the Week 15 results, but also the Week 7 results. This will allow us to see which coaches really improved over the second half, and if any got noticeably worse.

Lastly, I included a 'Hot Seat Zone' rating, which is just a basic red-yellow-green color coding based on some admittedly arbitrary cutoffs. If you're a 'Green' rating, you have a less than 10% chance of being fired (some would argue none of these coaches would ever get fired, but it has happened - most recently with Lovie Smith of last years' Bears team. Smith's team recorded a point differential of about +100 and improved by 2 wins, but he was still let go.)

If you're in the 'Yellow' zone, your odds of being fired fall somewhere between 10% and 40%. Certainly not as safe as the Green Zone. Last year, three of 15 coaches in this zone got fired (that's 20% for those scoring at home. Chan Gailey, Pat Shurmur, and Norv Turner)

Last but not least, the 'Red' Zone - which is just about as hot as you can get. Last season, four of six coaches in this zone were canned (Romeo Crennel, Mike Mularkey, Andy Reid, and Ken Whisenhunt).

So where have our coaches shaken out this year?

You'll notice only five coaches fall in the 'Red' Zone this year, but with two exceptions. Gus Bradley is up there, but as noted in the table, he'll likely earn a pass as a first-year head coach. So he (and other new coaches) are excluded from our ratings. Were Gus in his second season, his odds of being fired would be a sweat-inducing 45%, although it's important to note that Gus and the Jags also improved his odds the most of any coach over the second half of the season from a 76% likelihood after Week 7.

You'll also notice that the 'Red' Zone has already seen its first victim, Texans' coach Gary Kubiak. Heart condition be damned, the Texans let Kubiak go earlier this year - when his odds of being fired were already over 50%. I left the stats as is, because it didn't make sense to keep predicting an event that's already come and gone.

Besides Kubiak - who has the most reason for concern. Mike Shanahan of the Redskins, Mike Smith of the Falcons, and Leslie Frazier of the Vikings. These names aren't all that surprising, although other coaches like Tom Coughlin and Greg Schiano may have climbed their way out of danger, these other coaches continued to see poor performance over the last 8 weeks.

Now remember, this is a projection of what might happen to these coaches after the season, based only on historical data. This doesn't include the fact that Mike Smith got a vote of confidence from his owner, or that everyone in Detroit might be sick of Jim Schwartz.

But by and large, these numbers should give us a pretty good indicator. And as an Eagles fan, it's nice to see Chip Kelly showing a big improvement over the second half. After last night's Bears game, I'd doubt you'd find any Philadelphian would would put his odds over 7%, most would tell me I'm nuts. But then, if the Eagles lose in Dallas next week, I'm sure I'll hear from folks who'd want it at 100%.

No comments: