Proposals to change kickoffs would substantially change the game--we know that--but in what way? The following post will try to see what changes in the kickoff game, through a lens called Expected Points Added.
I'm taking my data from advancednflstats.com, and they are a fantastic site. I love it, and you should too. One of the concepts they use to measure performance is EPA - that is, the Expected Points Added of a single play. A lot of plays obviously don't directly result in points but are important. How do we compare the importance of a tackle for loss to a first down? Is it better to get an interception on your 30 or to get a sack on their 20 to force a 2nd-and-20?For example, a sack is worth about 1.7 points, on average. Most plays obviously do not add a single point, but fractions of a point. Obviously, every situation is worth different amounts of points, taking into account downs remaining, field position, point differential, and time left, which is why they use an enormous database and regression analysis to calculate the EPA and WPA (Win Probability Added, which is what it sounds like--how much each play adds to the likelihood of the win). WPA is mostly useful for in-game specific situations and EPA is a pretty decent measure of general strategies. The good thing about EPA here is that almost all kickoffs are the same, and the database is larger and more useful. It's fairly easy to evaluate kickoffs and their relative importance.
Here's the data for current kickoffs and field position. Billy Cundiff leads to league in touchback percentage with a whopping 50% (40 of 79 Kickoffs). The next best is Stephen Gostkowski, at 35.7%. For those who are curious, Longwell is at 4.6%. A touchback is worth about 0.8 Expected Points (That is, the amount of points a team is expected to score from the 20 yard line is 0.1, and the amount of points from an average non-touchback return is 0.9 points).
The spike at the 60 is from Out-Of-Bounds kicks.
The rule changes would change the offense's expected points to (if I read the graph below correctly) to 0.8, instead 0.1. The median kickoff return is to the 27 (which I think is a better evaluator than the mean, because it does some correction for endzones decreasing the average and the out-of-bounds skewing the average). The problem with determining what the future avg or median return will be is of course the changes in kickoff formation procedure and the fact that you can't simply subtract 5 yards because there is also a corresponding decrease in the amount that the coverage team has to run. The other problem with using the graph above to determine expected points is a) the lack of normal distribution and b) the small bump between the 20 and 35 (which I imagine comes from the kicker tackling the breakouts... that also explains why people who make it past the 20 typically make it to the endzone, unless your name is Dan Connolly).
My intuition is that one would have to expand the graph to account for kicker position and the fact that runners in coverage get a headstart over returners instead of shifting the peak 5 yards. This is not my data, so I didn't expand the graph to any degree that I thought would be useful. As it is, I think that (and the data is out there, I just haven't checked) that changing the kickoff position to the 35 instead of the 30 would reduce the median return by 3 yards, which is an admittedly arbitrary calculation, especially with the restrictions to coverage formations. Also, the changed incentive structure for people receiving the ball within inches of the endzone would unpredictably change the data in ways that I don't know how to account for.
If the median return changed to the 24, like I "predict," then most returns would give the offense 76 yards to go... which is a one yard difference from a touchback. The EPA of a touchback is virtually 0.... if anything it would be negative under my quick calculation, because you would rather have someone make it to the 24 (and hit them in the process) then given a spot at the 25 (I know I haven't priced in the discomfort of the uncertainty of a touchdown for a normal kickoff return, but it's probably balanced out by the fact that a) you get to hit someone and b) you might be able to keep them inside the 10).
The average kicker booted it 64 yards before the return team got their mitts on the pigskin, which means the average kick will put the ball on the 1-yard line. The data cited includes every single kick from every kicker, which means it's skewed one way by kickoff specialists (like Billy Cundiff) and another way by kickers who have been let go or at least relieved of their duties as kickoff kickers. They probably don't balance equally, and I imagine there are more kicks in the dataset from KOS than from kickers (a quick glance confirms that - I would think that Cundiff's 79 kicks singlehandedly outnumber the relieved kickers), but I'll let that go. Those 5 yards are therefore very significant, to move the placement of the ball from the 6-yard line to the 1-yard line, and substantially reduce the importance of kickoff specialists.
It's very odd that touchbacks could actually be (marginally) bad under this new rule system. Obviously in close games near the end of the game the team in the lead would obviously want to reduce variability and have the touchback rather than the return, but most game situations (and particularly the beginning of the halves) would seem to dictate avoiding the touchback if possible. I would also rather have a touchback in an overtime situation, given the high degree variance in that situation.
Of course, while I dismissed its influence when calculating the data, the proposed changes in the coverage formations probably has significant effects. Outlawing wedge blocks in the NFL increased the return average by 0.4 yards.
So those are my thoughts.