Fixing Football Kickoff Strategy
Exploring an idea to radically change the dominant strategy for football kickoffs
American football as much as any sport tends to be very traditional and slow to change. For example it has taken two decades to move the needle on David Romer’s very strong case that teams should go for it on fourth down A LOT more than they otherwise did and still do.
While there are rational reasons teams do not follow Romer’s advice, those are all outside of the scope of win maximization falling generally into the category of selfish motivations (there is no “t-e-a-m” in “conflict of interest”).
I will admit that saying this can “fix” football kickoff strategy makes it sound broken, which it isn’t in a literal sense. But considered from the point of view of a clearly dominant strategy, one that is better than the current strategy in almost all cases, current convention would be considered broken.
Last year a friend and I were talking about this epic high school football comeback (down 17 points with 1:10 left to play), and we got to thinking about why teams don’t kick onside more often than they already do. This is especially puzzling since unexpected onside kicks (usually those coming well before a team has to try it at the end of a close game) work about 45% of the time. What follows is the strategy we devised to theoretically give the kicking team a big advantage basically putting them on the offensive rather than passively giving the ball back to the opponent.
If a team is going to attempt to recover their own kickoff (kick “onside”), they usually try to get the ball once it has gone the requisite (by rule) ten yards past the line of the kick by kicking it softly or in a way that creates a crazy bounce. Basically they are introducing volatility into the path the ball takes. The advantage right out of the gate for the kicking team is they know which way they are going to kick the ball to attempt recovery (left, right, or straight ahead and with or without a lot of force).
However, a few years ago at least to my noticing teams also started to kick “deep onside” attempting to outrun the receiving team to the ball as it bounced far down field. I “noticed” this when my team, OU, under head coach Lincoln Riley repeatedly refused to put a deep man back before an obvious onside attempt. The echos of my screaming disagreement with this obviously flawed thinking can still be heard in Gaylord Memorial Stadium to this day. And it was especially appalling to see this same mistake repeated even after Texas nearly recovered a deep onside kick by outrunning the backpedaling OU players.
Putting these together my friend and I realized kicking teams have several options with which to work. Namely, they could:
Kick a short onside attempt to the left, right, or center doing so a couple of ways hard or soft.
Kick a deep onside attempt to the left, right, or center doing so as well in a variety of styles.
Kick a regular kick attempting to get the ball to the end zone for a touchback or directly to an opponent including a high kick likely to be fair caught.
Current strategy is to do option 3 almost always aside from end-of-game situations. Adding the other options might give an advantage to the kicking team if they could expect a successful onside recovery often enough. While the 45% figure cited above is likely extremely higher than a team employing this strategy could expect, the chances do not have to be anywhere close to that good. Before I get to some math, let’s consider some game theory.
If your opponent has become well known for very often attempting onside kicks, you would likely need to respond with a change in your typical kick receiving strategy. Namely you would likely need to put a lot more players close to the kick lest the kicking team have a clear recovery advantage for short kicks. This change is necessary but unfortunate for you the receiving team because now the kicking team’s ability to recover deep onside is much improved and your ability to set up a return is much hindered—all your blockers are too far down field. This new equilibrium has very likely already moved the advantage toward the kicking team.
If you can recover onside attempts well as the receiving team, this is mitigated. So what can the kicking team do to maximize their advantage? It starts with them knowing what is coming in the kick. That is step one. Step two is using the same formation every time to minimize the chance the receiving team knows what is coming before the ball is contacted. Here is the setup and execution:
Kicking team makes a determination of what the kick will be.
Short onside (left, right, or center)
Deep onside (left, right, or center)
Deep traditional
Team breaks from the sideline to set up as always with five players on each side of the ball lined up about ten yards off the line of the kick.
Kicker sets the play into motion beginning to approach the ball to kick. At that point all five players on each side begin running. They know where they are headed, but the receiving team doesn’t yet.
If executed consistently, they receiving team will not be able to tell initially if the players running right at them are planning on engaging them, as they would be if the ball was a short kick to that side, or planning on running past them to try to run down a deep kick. If they just go to block them thinking the ball is going deep, they will likely sacrifice their ability to recover a ball kicked into their area. If they just attempt to recover a ball and ignore them, they will sacrifice their ability to hinder a deep onside kick. Damned if you do, damned if you don’t.
The downside for the kicking team is giving the opponent the ball on average farther down field than traditionally otherwise. How big that difference is comes down to the average starting position in each case (onside strategy versus traditional strategy) and the value of the starting position. The math to determine this tradeoff is a comparison of net expected-points added (EPA).
Here is a good explainer on EPA including some estimates of it for various starting positions on a drive in the NFL. Notice in the graph under the section “Application of Expected Points” that starting on the opponent’s 45 yard line looks to be worth about twice the EPA as starting on a team’s own 25 yard line (about 3 versus about 1.5). I will refine these estimates below.
The important variables to consider are:
Probability of recovery by the kicking team [pK]
Average starting position for the kicking team if they maintain possession of the ball (aka, recover an onside kick) [Kyard]
Average starting position for the receiving team if they obtain possession of the ball [Ryard]
EPA for various starting positions [EPA]
The most straightforward way to put this together would be as follows:
pK * EPA|Kyard - (1 - pK) * EPA|Ryard = Value of Strategy
Running the above for each strategy (onside and traditional) yields the net EPA comparison. To the degree that the result is higher for the onside strategy, that strategy would be said to be an improvement over the traditional strategy. If it is high enough, we might say it “dominates” the other.
Here is a hypothetical comparison using an approximation of the EPA from the above-linked chart. “OS” stands for onside strategy while “TS” stands for traditional strategy.
Comparison of OS and TS1
OS: pK = 25% <times> EPA ~= 3.5 given Kyard = 40 <minus> (1-pK) = 75% <times> EPA~= 2.6 given Ryard = 55
TS: pK = 1% <times> EPA ~= 5.4 given Kyard = 10 <minus> (1-pK) = 99% <times> EPA~= 1.6 given Ryard = 70
thus,
.25*3.5-.75*2.6 versus .01*5.4-.99*1.6
or
(-1.08 for OS) versus (-1.53 for TS)
Given these assumptions, OS is perhaps dominantly greater (less negative) than TS.
Range of Comparisons
Taking the calculation for TS above as the given baseline, we can compare a range of possible combinations of recovery probabilities and EPAs given various starting yard lines.
Given the assumptions in the table especially the consistent starting position for the receiving team, it looks like the interesting decision between the two strategies begins once the onside kick probability of success (from the kicking team’s perspective) approaches 25-30%.
No starting position can rescue the onside strategy if the recovery rate for the kicking team is very low (like 10-15%). At the same time no starting position will doom the onside strategy if the recovery rate is high (like 45% or more).
I think this establishes in principle that my onside strategy change to traditional football thinking is worthy of consideration. Now if we can just get a coach to try it . . .
P.S. Obviously there would be nuance to the actual implementation of the strategy as one would have to consider opponent by opponent what the EPA comparison would actually be (e.g., the variables would change drastically from the average or in direct comparison to Alabama or Vanderbilt being the opponent). Additionally, the determination of the kick mix (i.e., how many kicks of each kind are attempted) would likely change game to game, opponent to opponent.
Notice that I do not assume the starting field position is the same for recovery by the kicking team and recovery by the receiving team. In fact I put the kicking team closer to their target end zone (by five yards) when they recover an onside kick as compared to the receiving team getting possession after the kick. While one can certainly argue with the values I have chosen, it is very likely the two average recovery points are different. For one the kicking team can choose how often to kick deep rather than short. Another factor that would imply a lower average starting position (closer to their goal) for the receiving team is the fact they can advance the ball upon recovery. The kicking team cannot. Notice in the table below how I arrived at my rough assumptions on starting position.
