website statistics

At four minutes? Maybe not.

With 1:35 left to play, PIT led TB 24-20. Facing a 3rd down and 5 from their own 19, the PIT offense chose to run the ball. They were stopped with a 2-yard loss, but because TB had no timeouts left, the game clock wound down to 50 seconds before PIT had to punt. Unfortunately for PIT, the punt was about 10 yards shorter than normal and TB was then able to strike deep into the red zone, setting up the game-winning TD. Should PIT have thrown rather than run, risking stopping the clock or worse--a turnover?

Everyone is familiar with what a two-minute offense is. It's when a team that's down by a score late in the game furiously passes the ball all over the field while using timeouts and the sideline to keep the clock from running out. The four-minute offense is the opposite. It's when a team with the lead late in the game only needs to play keep-away. They don't need to score, only to convert a first down or two to keep the clock running and the ball out of the opponent's hands.


Offenses naturally play a very run-heavy scheme in these situations. Runs keep the clock moving and have a much lesser chance of a turnover. But this security comes with a price, because running not usually what moves the chains, especially when the defense knows what to expect. It's a delicate situation because the leverage can be so high. A simple conversion can put the game away, while an incomplete pass could hand the opponent an extra 40 seconds of game clock--an eternity in the modern NFL.

I was curious whether the four-minute offense was optimal. Should offenses with a 3-point lead begin a low-risk, run-heavy scheme with 10 minutes to play? 5 minutes to play? 2-minutes to play? What about with a 1-point lead? A 4-point lead? What starting point for the four-minute offense would be best given the game situation?

To answer the question I called several teams last year and asked them to wait until specific points in their games to begin their four-minute drill. I could then collect the data from this experiment and get an idea of what strategies work best how often each team went on to win. For some reason, they all hung up on me. I needed a different approach.

I turned to my game simulator, the WOPR, and programmed it to begin its four-minute strategy at 30 second intervals in the 4th quarter, given certain game situations. Comparing how often offenses go on to win based on when they commence their four-minute play selection rather than a 'normal' play selection can tell us the optimal strategy for any of the tested game situations.

The simulator chooses play types and their outcomes based on down, distance, and field position. The distributions are drawn from actual NFL plays since 2008. Each play type has its own distribution of outcomes, including yardage gains and losses, turnovers, and elapsed time. These play selections and their outcomes are also governed by a team mindset, such as two-minute, four-minute, etc. The mindset is based on game situation, which factors in score, time, and field position. 

I systematically changed the rules for when a team switches from a normal mindset to to a four-minute mindset and recorded how often the team went on to win. To keep things simple for now (and to keep from burning out my processor) I reduced the scope of the question to a down-distance-yard line situation. The offense has a 1st and 10 at their own 25. I looked at leads of 1 point through 10 points, and ran each possible scenario 20,000 times.

I expected to see a curve, where the likelihood of winning was lower when the offense starts the four-minute offense too early, peaking at the optimal time, and then decreasing when its too late to start it. That's not what I got, and I was surprised by the results. But in retrospect maybe I shouldn't have been.

The results for a lead of 4 points is shown in the chart below. To understand the chart, each line represents a starting point in the game. For example, the green line represents scenarios where the offense has a 1st down at their own 25 with 180 seconds (3 minutes) left in the game. Along the green line are the data points representing when the offense switched into their four-minute mindset. For example, if they switch into four-minute mode immediately at 3 minutes to play, they win 87.0% of the games. If the same team gets the ball with a 1st and 10 at their own 25 but then waits until 120 seconds (2 minutes) left, they win 88.7% of the games. And so on.



It turns out that, according to the simulation, it doesn't make sense to ever use the four-minute offense when up by four points. These results suggest it would be better to keep playing with a 'normal' mindset all the way to when it's time to kneel out the clock. For example, when the offense starts with the ball at 3 minutes to play and waits until 30 seconds left before going into their four-minute offense they win 90.0% of the time compared to 87.0% of the time when immediately going into it.

I won't bore you with graphs for the other score differences, but they're all very similar.

These results are limited to a single down-distance-yard line situation, but they're likely similar to results we'd get for most others. The only exception I can foresee would be when offenses are in field goal attempt range. The scenarios were also limited to both opponents having all three timeouts, but it's easy to alter any parameter to get relevant results for any scenario we want.

In PIT's situation last Sunday, the simulator is indifferent. Whether or not PIT is in their four-minute mode, they win 86.2% of the time (over 100k iterations).