On any given play in the NFL, a coach is faced with one of four basic decisions: run a passing play, run a running play, attempt a field goal, or punt the ball to the other team. Whether it’s business, poker, or deciding the most fuel-efficient route to take to the grocery store, we hope to make decisions that maximize our expected value (EV) or average long term return. Generally speaking, it’s best to analyze the decision within a vacuum, based on as much hard information (data) available to you at the time of the decision, free from things like tradition or results-oriented thinking based past decisions’ outcomes.
The NFL is a widely exposed institution where it’s become clear that many decisions are not being made based on an analytical thought process, but rather based on a history of “the right way to play the game.” It’s almost become second nature within casual sports fandom in America to say: “Well, it’s fourth down so you have to punt here” without considering the consequences of all options. This line of thinking seems to be based on a thought process in football coaching born from an era before we even had computers. NFL coaches are now equipped with the data (historical play records) to maximize their EV (winning probability %) based on choosing one of their four available options (run, pass, kick, punt). We can calculate the success rate of the average pass play, the expected point value of a field goal based on distance from the goal (hint: it’s almost never a full 3-point expected value), and a team’s change in win probability based giving possession back to your opponent via the punt.
In the first week of the 2015 NFL season, New York Giants coach Tom Coughlin was tasked with one of these decisions in a high-leverage situation of a pressure-packed NFL season opener vs. the Dallas Cowboys. Let’s examine.
The situation: The Giants lead the Cowboys 23-20 with 1:43 to play in the game. 3rd down and goal-to-go from the 1-yard-line. Dallas has 0 timeouts. Per the Advanced Football Analytics Win Probability Calculator, the Giants had a 94% win probability at this time.
3rd Down: Let’s eliminate the field goal and punt options from the discussion and focus on our two basic 3rd-down options: run or pass.
Option 1, Run: In 2014, just over 57% of rushing plays from the 1-yard-line resulted in a touchdown – a much publicized fact following the Seattle Seahawks supposed play-calling blunder in Super Bowl XLIX. A touchdown here essentially ends the game, as it gives the Giants a 9-point lead. If the run fails, the Giants can run the clock down to under a minute
Option 2, Pass: Interestingly, pass plays from the 1-yard line had the same success rate as the run in 2014, at just over 57%. However, a failed attempt uses practically no time from the clock and would leave 1:40 remaining.
While both options give the Giants a 57% chance at immediate victory, a failed run has the added benefit of leaving the Cowboys 40 fewer seconds to counter attack. A failed run decreases the Giants win probability to 90% going into 4th down, while a failed pass drops the Giants win probability to 82%. While this is still a dominant position to be in, an 8% drop in win probability is certainly significant. The clear option here is a running play.
4th Down: Assuming the Giants chose the worst of the two options, let’s now put them at the 1-yard line with 1:40 remaining with two options: kick a field goal or go for the touchdown.
Option 1, Field Goal: For argument, let’s assume a 100% field goal success rate from this distance and an ensuing kick off to the 20-yard line. The Cowboys would trail by 6 points with 1:30 to go in the game. Conventional thinking would say that the Giants should take field goal, as it will increase their lead from 3 points to 6 points. Following a field goal, the Giants now have an 84% win probability.
Option 2, Go for the Touchdown: 57% chance of immediate victory (from our previous 3rd down example). If the Giants fail on the touchdown, this now gives the Cowboys the ball on their own 1-yard line with 1:35 remaining and a 10% win probability, putting the Giants aggregate win probability at 95.7% (.57 + .43(.9)).
The clear option here is to go for the touchdown on 4th down.
So what did Tom Coughlin and the New York Giants decide to do? On 3rd down they ran an unsuccessful pass play, using virtually no game clock in the process. On 4th down, they opted to kick the field goal to go up by 6 points. Tony Romo promptly marched the Cowboys promptly marched down the field in 1:27, throwing a touchdown pass and captured the 27-26 victory over the Giants.
While the pass play on 3rd down was certainly unconventional, the Giant’s decision to kick the field followed the traditional NFL 4th-down thought process ingrained from a past era. Per Forbes, the New York Giants are valued at $2.1 billion. Tom Coughlin’s 2015 salary is $7 million – by not using a data-based approach in his decision making, this single set of decisions could end up costing the Giants at least that much in expected value. Just as with any business decision, it would have greatly benefitted the Giants to invest the time into using a data-focused approach to the game plan.
By Grant Klein
Posted by admin