website statistics
Starting in the 2011 season, the NFL returned the kickoff line to the 35. Due to ever increasing kick distances, the result was as predictable as it was intended--lots of touchbacks.

Watching the Sunday night game this week, I was struck by how the Bears returner struggled to get past the 20. The Bears began their first four drives at their own 6, 15, 20, and 19. Their returner was bringing the ball out from deep in the end zone, and I wondered if it were simpler, safer, and better in the long run to just take the touchback. Does the risk of penalties, short returns, and turnovers outweigh the potential reward of a good return? 

Based on the Bears first four returns, no. But later in the game, their strategy paid off with a touchdown return. One lucky break does not make a strategy, however, so I looked at all the numbers, and I think I know the answer.

The strange thing about this issue is that prior to 2011, returners would routinely accept touchbacks in situations they now regularly run out. And now, coverage personnel start out five yards closer. It doesn't make much sense that returners now run out deep kicks more often. If anything, they should now be more inclined to accept a touchback than they were prior to 2011 given the same kick depth.

The first thing to note is that all the numbers presented here are subject to selection bias. In other words, returners don't run deep kickoffs out of the end zone willy nilly. They make a decision based on whether they believe it's favorable to run it out based on the hang time of the kick, the situation in the game, and their coaches' guidance. But that's ok, because we can still evaluate whether the current typical return policy is sound, or whether return squads would be better off taking touchbacks more often, up to and including always.

Here is how often returners currently bring out kick offs based on kick depth. All data is from 2011 through week 10 of the 2014 season and is restricted to non-onside kicks from the 35.

The chart below plots the average return yard line based on the depth of the kick. Negative numbers indicate the kickoff's depth into the endzone. Predictably, the deeper the kick, the shorter the result of the play.

At first glance you might think it's always a good idea to bring out the kick. The average yard line of the return is always beyond the 20. But that's a plot of the average return, given that the returner chose to run it out. This result suggests returners are generally making the right decision. Unfortunately, things are more complicated. 

What about the risk of a turnover? A turnover returned for a touchdown? A penalty? Just a few years ago, we would have produced the chart above, done some handwaving to dismiss the consideration of penalties, scores, and turnovers, and called it a day. But now we've got models like Expected Points (EP) which can factor virtually every consideration into the analysis.

Here's the average EP as a result of all kick returns. In the version of EP I used for this analysis, a 1st and 10 at the 20 is 0.34 EP for the offense. But touchbacks can have penalties, so the average net EP for a touchback is 0.33. Anything above that means that bringing out the kick was probably a good idea. If returns from, say, 6 yards deep in the endzone result in an EP value below that, it means returners should generally be more inclined to accept the touchback. Where EP values are above 0.33, returners should be generally more inclined to run kicks out of the end zone.

As we might expect, resulting EP values increase steadily as the depth of the kick decreases. What surprised me is that EP values for returns are greater than for touchbacks for every possible kick depth. Only at 9 yards deep into the end zone does it appear a touchback has greater average value, and there are only 86 such returns since 2011 so that is very possibly due to statistical noise.

This does not mean returners should be running out every kick that goes 8 yards deep into the end zone. It means that given the current typical return tactics, returners could probably be a little more aggressive than they are now and still break even. It also means that returners (and their special teams coaches) were probably far too conservative prior to 2011.

The bottom line is no, returners should not always accept the touchback. The current balance is very close to the break-even point, and returners could stand to be slightly more aggressive. My recommendation is that teams seeking high variance (losing or underdog teams) should lean toward bringing out more kicks. Teams that seek low variance (winning or heavily favored teams) should lean toward accepting the touchback.

Leave your comments

Post comment as a guest


Comments (18)

Load Previous Comments
  • Guest - Phil

    This doesn't surprise me at all--if you have a good kick returner he should always take it out. Worst case scenario is of course a turnover, but those are infrequent. Next worse case is a penalty or absolutely horrid return that backs you up to the 10. But think about it: is there an enormous difference between the 10 and the 20? And that's the worst case, generally you see "bad" returns end up around the 15 or so, which is similar enough to the 20 to warrant taking the chance you get a big return.

  • Guest - MGL

    I’m not sure this is telling us anything at all. I mean, whether the average is positive or negative, we have no idea what percentage of players are making a mistake by running it out or not running it out. That is the important thing right. Or to estimate what the optimal result SHOULD be (not easy) and compare it to the actual result.

    We have to assume that some players are making a mistake and others are not, in both directions, either running it out when they shouldn’t or not running it out when they should.

    But the overall result doesn’t really tell us anything much other than it must be quite a bit higher than the BE point in order for everyone, or almost everyone, to be acting correctly.

    It is like advancing on the bases in baseball. If everyone is acting optimally, the overall result should be much higher than the break even point, since every single person and situation has to be some number above the BE point and some players and situations are a lot higher than the BE point.

    So, if we look at the results that this author did, we can conclude the following based on the results:

    1) Overall result is negative: Lots of players are not acting optimally, specifically too many players are running out when they shouldn’t. We have no idea how many players are not running out when they should.

    2) Same as #1 except probably fewer are running out that should not be. But some are. Again, no idea how many are not running out but should.

    3) Result is a little above BE, as the author found. We have no idea if this is optimal play by everyone. Of course it likely is never going to be. In this case, there are likely mistakes on both sides or perhaps not enough running it out who should be but few running it out who shouldn’t. But again, we have no idea really of the number of mistakes in both directions.

    You would have to do a detailed analysis using results of each player along with hang time at the least in order to model optimal play and then see how that compares to actual play in order to get SOME idea as to how optimal or not players are doing this.

    I’m not seeing this result as telling us much of anything other than perhaps being able to compare teams. I mean, if the overall result were less than the BE point, that would mean that players were making horrible decisions.

    Finally. the author’s suggestion that players are not being aggressive enough because the result is a little above BE is completely wrong for the reasons I explained above (Optimal decisions by every player should aggregate to well above the BE point).

  • Guest - Dave

    This seems particularly susceptible to huge changes in non-baseline data which, when coupled with the relatively small sample size, makes it difficult to accept a strong conclusion either way.

  • Guest - Ryan

    I don't think your reasoning is correct. What this data is showing is similar to run vs. pass data, that shows that teams should be doing the riskier thing more often. There's an optimal balance - for runs and passes, teams should pass slightly more often, but still run (the safer option) at times to "keep the defense honest." While there is, in effect, no defense for a KR touchback, the idea is the same - teams are on a whole choosing the conservative option slightly too often.

    No need to bring hang time or individual players' tendencies or skills into the equation; yes, sometimes returners will make mistakes both ways, but if returning yields a better average EPA than a TB, that suggests there are times when returners can make riskier choices (as an underdog, playing from behind, etc.) in order to balance out the equilibrium.

  • Guest - Ryan

    @MGL again,
    Never mind - rereading what you wrote I guess I kind of agree with you, that you'd need stronger data to really say people should be running the ball out of the endzone more. The fact that it's so close to the break even point suggests that it's possible (yet impossible to know for sure based on the data) that KRs are performing at perfectly optimal level; and since it is greater than the break even point, it's better than always taking a knee. I'd be curious to know what the average EPA would be if EVERY kick was returned (i.e. if there were no TBs by rule). My guess is it would be fairly close either way, that there's not a huge difference between "should vs. shouldn't" bring it out.

  • Guest - Tom Malone

    Should we also consider the added risk of injury to players on the returning team? Not sure if it's a large enough factor to consider.

  • Guest - SlackerInc

    I am really glad to see this post. I too have noticed a recent trend to take more balls out of the end zone. And I have certainly seen a lot more cases of returners giving stuffed up to 10 or 15 yard line or sometimes worse. So I am a bit surprised by this conclusion. I still think there are a lot of cases where situationally, seeing what the coverage looks like, these returners should be more cautious. But that is outside the scope of this type of analysis.

    Another apparent trend I have been noticing, that maybe you can analyze next, is an increased frequency of taking fair catches inside the 10 yard line. I have to think that the odds of the other team downing the ball at the 1 yard line are low enough that you are better off giving the ball a chance to bounce into the end zone and create a touchback, but maybe I'm wrong.

  • Guest - Bob

    I am sure all of this data is accurate and it may be better for an NFL returner to take the ball out, but in Cleveland, where we have no true NFL players, we somehow find a way to A)get stuffed inside the 15 everytime or B)commit a holding penalty or C) turn the ball over or D) get a player injured or E) some combination of any/all of those. So I scream at the television anytime one of our players runs it out because it never turns out positive....even the times we catch it right at the goal line. I would be interested to see Browns-only stats. Take a knee if you're a Brown.

  • Guest - Len

    This is a great piece of research, there are so many other factors that I think go into this that others have already mentioned that may not be significant such as the risk of penalty, risk of player injury on kick returns, risk of fumble. In a real life situation in a tight game some teams may be more tempted to take a risk while others may be less tempted as they may have faith in their offense and defense or even their own punting team to pin the opponents offense deep. An average field position of 3 yards to me isn't a huge statistical difference in the course of a game. The risk/reward assessment will depend on a lot of matchup factors and idiosyncratic human factors such as emotion and ego. A universal strategy would never be employed by an NFL team, even if the numbers may say it should be. I'd like to see numbers for each team for each year although it may not constitute a significant sample size.