By Anthony Zonfrelli
Every week, NFL coaches list their players with injuries as either âDoubtful,â âQuestionable,â or âProbableâ to play that weekend. What many people donât know is that these statuses are supposed to indicate a specific probability of playing. Under NFL rules, these labels mean that the designated player has a 25 percent, 50 percent, and 75 percent chance, respectively, of playing that Sunday, which raises the question: are NFL Coaches assigning their injured players these statuses with any accuracy?
Using every injury status update for each team through the 17 weeks of the 2012 season, I went back to chapter one of the statistics textbook and used one-sample t-tests to examine the coachesâ accuracies. I only used data from the one season to (essentially) hold coaching staffs constant.
The status breakdown for the league went like this:
Injury Status Assignment |
Number of Players Listed |
Expected Fraction of Players to Play |
Fraction of Players that Played |
Doubtful |
167 |
0.25 |
0.27 |
Questionable |
3146 |
0.50 |
0.62 |
Probable |
1986 |
0.75 |
0.89Â |
With a p-value of 0.63, there isnât evidence to suggest that the true mean of âDoubtfulâ play rates is not 25 percent. However, the play rates for âQuestionableâ and âProbableâ are both substantially and significantly higher than their implied values. This means that coaches tend to overestimate the severity of week-to-week injuries when they are not serious. Therefore, players labeled as either âQuestionableâ or âProbableâ have a higher chance of playing than their statuses indicate.
This trend may exist because of the subjectivity of the terms âQuestionableâ and âProbable.â In everyday speech, these terms mean different things to different people in different situations. If somebody says that something will âprobablyâ happen, then they more than likely assign the event nearly a 100 percent chance of occurring, but thatâs not what it is supposed to mean in the NFL Injury Report.
If we use the sample means as our best estimates of the true means for âQuestionableâ and âProbableâ listed players, then they really have a 62 percent and 89 percent chance of playing, respectively. Keep that in mind when you are setting your fantasy lineup.
Some of the more interesting findings, though, happen just from eyeing the differences between teams in the sample. For example, while Bill Belichick is often criticized for abusing the injury report, the Patriots only ranked 8th and 6th in the league in terms of highest âQuestionableâ and âProbableâ play rates last year. Actually, Buffalo was first in both categories (T-1st in âQuestionableâ play rate). In fact, 98 percent of the 81 Bills listed as âProbableâ last year saw playing time that week. The Bills might not even be aware of this rule.
The Green Bay Packers, though, nearly hit all three categories on the money â they had marks of 25 percent, 49 percent, and 75 percent. They would have been perfectly accurate if just one more wimp listed as âQuestionableâ had played. Also, with a perfectly legal and accurate 75 percent play rate, Green Bay had the lowest play rate for âProbableâ-listed players, which is to say that players listed as âProbableâ really have at least a 75 percent chance of playing. Similarly, at 49 percent, the Packers were tied for the second lowest âQuestionableâ play rate; New Orleans curiously only played 36 percent of their âQuestionableâ players.
The doubtful designation used to be near-zero. See http://usatoday30.usatoday.com/sports/football/nfl/2007-11-22-injury-report-cover_N.htm
and http://online.wsj.com/news/articles/SB10001424053111904900904576554640692835656
Did the NFL crack down on this recently?
I also don’t quite understand the rules. If the chance of playing is closer to 100% than 75%, should the player be probable? If the chance of playing is closer to 0% than 25% should the player be doubtful?
Interesting stuff! It always seemed like this was happening, and now we know to what degree. Also good for fantasy decisions, perhaps đ
Out of curiosity, what were the observations for the t-test? Each week? Did you compare the number that played to the number that “should-have-played” in a matched t-test? My results from a Chi-square test show the same results. Just curious.
Thanks again!
this known by most that do follow injury reports (well at least in las vegas). Belichick has gained his reputation over the years, go back a dozen year, See what else you can come up with as I wonder if coaches or perhaps even organizations are more ‘truthful’ than others.
this would be quite a grind especially tracking coaches that change teams.
Perhaps if you notified the commissioner of the NFL, something might be done, but I doubt it. They can always say week-to-week or day-to-day as we all know coaches want to make their opponent prepare for everything, so any advantage helps even in the form of misinformation or late information.
There are no penalties for late information on coaches or organizations, just a lot of hand wringing by fans and bettors alike.
I’m unsure if this process could be any better.