By now everyone and their mom (I’m serious, my mom called after the game to question it), is talking about the fake punt early on in Monday’s game. Here is a sampling of the reactions:
EDSBS"12:20: As if on cue, we turn the channel to the Cotton Bowl and Auburn’s blocking a punt. Wait–no. They’re actually faking a punt on a reverse, which Auburn speedily dismantles. Callahan shows his inner schmuck by making that call, putting Auburn on the NU 15 or so."
Big Red Network"You don't run a fake punt on your own 20-yard line. You...just...don't. I’m stunned by the raw arrogance of it. In a game where both defenses are playing well, field position is especially important. Why NU did that, I don't know. And, almost no explanation is likely to be good enough. It absolutely was just like giving the dang game away. Some will say that if the play worked and they got a first down, then it would be genius. Nope. It would still be arrogance. The Greeks had a word for that - hubris. Show too much pride, and the Gods will smite you. Callahan got smote good. And, it cost these kids the game."
Husker Mike"Callahan has made a habit out of trick plays this season, and apparantly this fake punt was all that was left in the bag of tricks. It should have stayed there, as this one was all-wrong. Run from deep in Nebraska territory, there's no guarantee that it would have led to points as we were far from being in a position to capitalize. Simply put, the risk-reward balance was too high on the risk side without that much reward. It was poorly executed, as Dane Todd's pitch to Andrew Shanle was fumbled, giving Auburn even better field position."
Fullback Dane Todd
"I thought it was a great call, a great situation to do it. You take a gamble like that, you're going to get burned once in a while."
Coach Callahan himself
"It was my call. It didn't work, obviously, but nonetheless, it was still early enough in the ball game that if it didn't work and if it faltered, we were still in a good position we felt to come back, but things got discombobulated there. We fumbled the exchange, then lost some critical yardage, so that hurt us. We got behind the eight ball on the short field."
Given the strong reaction to this call, I decided to give it another look and to try and understand it not from an emotional standpoint, but from a rational and logical perspective. First, in the spirit of full disclosure – I had some immediate reservations about the call. My major concern centered on the use of Dane Todd and Andrew Shanle as the key cogs to the execution of the play. These are two players with limited experience in handling the football. I recognize that the presence of other, more “visible” players might have tipped off our intentions to Auburn. However, this is the coaching staff that lined our backup QB at kicker in order to complete a fake FG against Colorado.
Anyway, onto the analysis.
Examining decision-making during the course of a football game hinges on an analysis of statistical probabilities and hints at the concepts of
Game Theory. Despite 18 hours of graduate level statistics, this lies just beyond my expertise. Fortunately, we have the fine folks at
Football Commentary to help us out.
The Football Commentary site has developed the
Dynamic Programming Model. The footballcommentary.com Dynamic Programming Model is intended to provide guidance for certain decisions that arise during a game, such as two-point conversions and going for it on fourth down. The Model is built around the idea that in making decisions, we are trying to maximize our team's probability of winning the game, and the opponents are trying to minimize that probability.
According to the site:
“There are three types of situations, called states, in which the Model explicitly evaluates our probability of winning. The first type of state is when one team or the other has just gained possession. The second type is when a team has just scored a touchdown, but has not yet tried for the extra point (or points). The third type is when a team is about to kick off.
Options to attempt a two-point conversion, to try an onside kick, or to go to a hurry-up offense are modeled explicitly. In addition, making a first down at a particular time, field position, and point differential is equivalent (from the Model's standpoint) to first gaining possession at that same time, field position, and point differential. Therefore, the model will allow us to analyze decisions to go for it on fourth down.”
The Dynamic Programming Model produces a series of tables that are intended to provide guidance regarding when to go for it on fourth down rather than punt. Although they do not cover the decision to fake a punt directly, I contend that we can utilize these charts to examine the astuteness of Callahan’s decision to fake the punt from deep within our territory against Auburn.
The
“Go For It” tables include several variables for guiding the decision-making. These variables include, field position, score, and for first half decisions, whether you will be kicking off or receiving to start the second half. Tables are provided for four different field positions: Our own 5, 20, and 40-yard line, and the opponent's 40-yard line. In addition, the tables are computed under the assumption that a punt nets 40 yards, except when the line of scrimmage is the opponent's 40-yard line, from which we then assume a punt nets 30 yards.
So let’s examine the variables facing Callahan and the Huskers when the crucial decision was made. The game is tied 7-7 with approximately 15 minutes left in the first half, and we face a fourth down near our own 30-yard line. We will also be kicking off to start the second half. Because the 30-yard line falls in between the tables for our own 20-yard line and our own 40-yard line, we will have to do some extrapolating.
In the Table labeled
"Own 20 yard line, first half, we will kick off to start the second half" we go to the row corresponding to a lead of 0, and the column corresponding to 15:00 remaining. The Table entry is 0.58. This means that if the probability of picking up the first down exceeds 0.58 we should go for it (or possibly fake it), and otherwise we should punt. In the Table labeled
"Own 40 yard line, first half, we will kick off to start the second half" we go to the row corresponding to a lead of 0, and the column corresponding to 15:00 remaining. The Table entry is 0.51. This means that if the probability of picking up the first down exceeds 0.51 we should go for it (or possibly fake it), and otherwise we should punt. If we extrapolate from those probabilities to fit our situation at our own 29-yard line we get a range somewhere between .51-.58.
In other words, Callahan should only have called the fake punt if he felt it had a probability of success (gaining 1+ yards), of somewhere around 0.55. Given the element of surprise, the need to gain just one yard, and the likelihood of successful execution of the play in the practices leading up to the Cotton Bowl, it becomes easier to see how Callahan came to the decision to call for the fake. The model is telling us that if Callahan felt the fake would work 6/10 times, then calling it at this point, and from this spot on the field would maximize our probability of winning the game.
Now let’s put the decision into greater perspective by contrasting it with other, similar coaching decisions. First, versus USC, the Huskers ran a similar fake punt. At the time, however, that decision was viewed in a much more positive light, despite the fact that it led to no points. In that situation, the game was tied 3-3 with approximately 10 minutes left in the first half, and we faced a fourth down near our own 40-yard line. We would also be kicking off to start the second half of that game. In the Table labeled
"Own 40 yard line, first half, we will kick off to start the second half" we go to the row corresponding to a lead of 0, and the column corresponding to 9:00 remaining. The Table entry is 0.51. This means that if the probability of picking up the first down exceeds 0.51 we should have gone for it (or possibly faked it), and otherwise we should punt.
Notice that the probability is nearly identical to that of the Auburn game. The only difference was that this particular fake punt led to a gain of 28-yards and a first down. The reaction to the two decisions, however, has been vastly different. My hunch is that the difference has nothing to do with probabilities, or game theory, but instead centers solely on the success of one fake and the failure of another. After all, both were designed to catch the opponent off-guard and both came early in the game allowing the team time to come back if the play failed.
For more perspective, consider another coaching decision. When we faced Kansas State, first year head coach Ron Prince also made what was viewed as a high-risk decision concerning a fake punt. In this situation, Kansas State trailed by 7, with approximately 18:00 to go in the first half, and faced a fourth down near their own 9-yard line. Kansas State would also be receiving the second half kickoff. In the Table labeled
"Own 5 yard line, first half, opponents will kick off to start the second half" we go to the row corresponding to a lead of -7, and the column corresponding to 18:00 remaining. The Table entry is 0.70. This means that if the probability of picking up the first down exceeds 0.70 KSU should have gone for it (or possibly faked it), and otherwise they should punt. Here we see that Prince should have felt that the fake had a 70% probability of success, or he had no business calling it. In this case, the fake worked and gained 38-yards, but led to zero points.
We see, therefore, several fake punt situations and probabilities associated with this decision-making process. One might argue, however, that the Dynamic Programming Model is simply a computer simulation, and has no bearing on coaches, who tend to make decisions based on “hunches” or “a feel for the game”. I believe that many coaches would disagree with this assessment.
For instance, in his book
Developing an Offensive Game Plan, Brian Billick writes:
“Too often people have resisted the technological wave of advancement, thinking that a computer is nothing more than a number-crunching, dehumanizing, complicated mechanism – a device intended either for only the most sophisticated ‘hackers’ or for the games of children. Nothing could be further from the truth.”
In addition, even crusty curmudgeon Bill Parcells has moved into the future. In his own book,
Finding a Way to Win, he says,
“If the competition has laptop computers and you’re still using yellow legal pads, it won’t matter how long and hard you work, they’re going to pass you by.”
Coaches are most definitely aware of this type of probability model and are likely using them to guide their decision-making process. Callahan and the Huskers got burnt when poor execution reduced the probability of success for the early fake punt to zero. Auburn jumped on the mistake and Nebraska’s offense could not duplicate its early accomplishments. The end-result was a 3-point loss to a Top 10 team, and another failed attempt to get over the proverbial hump.
Final disclosure – I had no hidden agenda when writing this piece. I am not being paid by Steve Pederson, Bill Callahan or Callahan’s agent. I was not attempting to defend Callahan’s decision, nor was I sure what I would specifically find by examining the Dynamic Programming Model tables. My sole intent, was to examine the data we have at our disposal in order to put this particular decision into a less emotionally-charged perspective. However, you may note that I had previously glanced at the tables in a cursory manner, and was at least superficially aware of a trend indicating that coaches should probably go for it more frequently on fourth down.
Posts
