Clock management is certainly a crucial task that the personnel of an NFL team has to work on and be prepared for. However, it should be obvious even to the casual fan that the teams are all using pretty much the same *guidelines *for clock management. In this article, I am not going to judge these guidelines — most of them appear to be ok anyway. However, I am going to describe how I would choose to use my time outs in an of course purely analytic way. To do this we will need an in-game win probability model, i.e., a model to provide win probability for a team given the current state of the game. Of course, I am going to use the one I developed and presented here in this blog, i.e., iWinRNFL.

The general idea of using an in-game win probability model is to calculate the probability added by saving a timeout for the end of the game instead of taking it earlier to avoid a delay of game penalty (e.g., say during the third quarter). Of course, a detailed treatment of clock management is not easy enough to be treated within a blog post, but as I said above my goal is to provide an analytics point of view on how to evaluate options.

Let’s assume that your team is visiting and is down 5 points on a first and 10 at the 50 at the beginning of the fourth quarter. You still have all of your timeouts and so does your opponent. The play clock is running down. What are you going to do? Burn a timeout to avoid the 5-yard penalty or take the penalty and keep a timeout. A rational coach would certainly measure up the options. Is the difference in the win probability between 1st and 10 at the 50 and 1st and 15 at your own 45, (much) higher as compared to having 3 timeouts instead of 2? This is what it basically boils down to. However, coaches almost instinctively they will gladly take the timeout. Let’s see whether this is the best option that you have.

Let’s first see how the number of timeouts left impacts the win probability. For a 1st and 10 on the 50, with 14 minutes left on the clock the following gives the win probability as a function of the score differential (negative means the visiting team is ahead).

As we can see having all of your time-outs gives you consistently the best chances. Obviously for large score differentials the differences are not very important but for close games the differences are distinct. For example, for a tied game the total win probability decline is 1.6% per timeout. Simply put if we take a timeout at this point we reduce our win probability by 1.6%. What if we take the penalty? If we take the penalty we will go from a 1st and 10 on the 50 to a 1st and 15 on our own 45. This translates to a total win probability decline of 0.13%!! The following figure shows for the specific situation we have at hand (midfield, start of fourth quarter etc.) the total win probability decline when calling a timeout instead and when taking the false start penalty as a function of the score differential.

As we can see the total win probability reduction is always higher if the team burns the timeout (in this scenario)! One can work out different scenarios and create *cheat sheets *for different situations. For example, for a third and short, taking the timeout might actual reduce the win probability less as compared to taking the penalty and having a third and long instead. Of course, someone could argue that a 1 or 2% probability is not important. So I will leave that to anyone that thinks that way. However, I truly believe that a more data-driven approach to clock management is certainly going to be beneficial for the teams. They might end up with the same game plan they have today, but this will validate further their choices.