When you can’t stop, steal

Hey Siri, what’s the recovery rate of a surprise onside kick in CFB?

Finally, some high entertainment value from the Saturday night games. While the Miami-Clemson game got out of hand eventually, the ‘Bama-Ole Miss game was close down to almost the last play.

Some were critical of Ole Miss, in a tied game, attempting a surprise onside kick with 11:31 to go in the 4th. I personally loved it, and while I don’t have the hard stats to back it up like I would going for it on 4th down, I’m going to try to explain my argument as data-driven as I can.

First, we should just acknowledge that Alabama’s win probability didn’t move one bit after recovering that onside kick. So that play in particular was inconsequential. Now, once they scored on that resulting drive their WP went up 9% to 89%, however they were already in the mid to high 70s for most of the game, despite a tie score. I’m guessing this is because it’s Alabama and Nick Saban, and ESPN thinks they win close games against weaker opponents most of the time. The point I’m trying to make here is that despite the tied score, Ole Miss were essentially losing this game.

Second, just the reality of the game. Alabama had scored touchdowns against Ole Miss on 5 straight drives, and 6-of-8 total drives thus far in the game. 2 of these drives were 85 yards, and they would go on to have a 90 yard TD drive in the drive after the onside kick recovery drive. So field position was not the issue. Ole Miss were not stopping Alabama, and quite frankly Alabama were not stopping Ole Miss, who had scored on their last four drives, minus a 1-minute drive to end the 1st half. And the situation was doubtful to improve for either team as defenses tend to get worn out after giving up several hundred yards.

Third, and finally, is the sheer probability of recovering a surprise onside kick. Now, clearly I don’t have the data on this from college, and surprise onside kicks happen so infrequently that there probably isn’t enough good data to come up with a solid number. However, I can presume it is significantly higher than a normal onside kick where 11 guys on the receiving team are waiting for your 11 guys, vs. 4 or 5 guys in-range reacting to your 11 guys who know the plan. I actually stumbled upon an interesting article by Brian Burke from 11 years ago that talks about this in the NFL. The first point is probably obvious: teams most often attempt onside kicks when the game is essentially out of reach. The below graphs are credit Brian Burke and his article.

Teams use onside kicks most often when win probability is extremely low. Credit Brian Burke.

His second graph is more interesting and less immediately knowable by watching games. The recovery rate based on a team’s win probability.

Onside kicks are much more successful when the game has not been definitively decided yet. Credit Brian Burke.

Now this is NFL pre-onside kick rule changes (so similar to how college rules still are today). The 19% recovery rate when a team has WP of ~1% is pretty much in line with what the league averages were back then. This makes sense since, as the first plot shows, most onside kick attempts are made out of desperation. But as soon as you have even a 20%-30% chance of winning (Ole-Miss’s exact scenario), the recovery rates of onside kicks shoot way up to 60%! Regardless of formation, we can consider these attempts as “surprise” onside kicks, because it is inherently surprising for a team to do an onside kick when they are still well in the game, or if there’s a lot of time on the clock.

So is it worth it? A 60% recovery rate sounds great, but what does a 40% failure rate do to you on defense with a short field? Well, based on EPA, Brian determined that a successful onside kick was worth 1.2 expected points, a failed onside kick was -2.1 expected points, and your normal kickoff is already worth -0.7 expected points. So a failed onside kick is 1.4 expected points worse than the normal kickoff. Using some basic algebra, he calculated that you would need a 42% success rate for your onside kicks for it to be worth it in expected points. So, given the plot above, we’d say that it was definitely worth a shot because on average, the 60% recovery rate will net you positive expected points.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s