Stat of the Week: WPA and the Packers’ prevent defense

The “Prevent Defense” is a dumb defense that old-school defensive coaches still occasionally use for no good reason. I’m still a bit honked off that Joe Barry used it on Sunday as the Green Bay Packers tried to hold on to an 8-point lead over the Tampa Bay Buccaneers with three minutes left. The prevent is designed to force the offense to complete passes in front of your defense, allowing your defense to make tackles in bounds, and waste clock. It’s also designed to prevent over-the-top bombs by playing most DBs quite deep.

Because of these “features” of the Prevent, it’s easy for receivers to get clean releases and because every DB is giving such a large cushion, there’s usually a large bubble of empty space over the middle. In theory, the prevent attracts these middle throws to keep the clock running. In practice, however, the yardage offenses gain is more valuable than the time wasted unless it’s VERY late in the game.

There are exactly two instances when using the prevent MAY be defensible. The first is when the offense has no timeouts, well under a minute left, and a long way to go. This is one of the few instances where a team may actually run out of time before scoring. The other is when a team is up by multiple scores and your only goals are to run clock, make sure no one gets hurt, get the ball back, and kneel down.

Week 3 in Tampa was not that. This was a one-possession game with 3:04 on the clock (an eternity in NFL time) and 1 timeout remaining. This was TOM BRADY. Yes, the Bucs were starting from their own 11, but the prevent doesn’t really capitalize on an offense being pinned deep, because it allows such easy gains. Using a prevent defense here was defensive coordinator malpractice.

Before we get into the drive, we should establish what we mean by “prevent” and the second play of the drive, where Russell Gage easily gained 15 yards, is a great example.

You can see how far off the Green Bay DBs are playing and how much space the Bucs have to work with underneath. And the defense “works” in allowing completion in the middle of the field with the clock ticking, but it only trades 20 seconds for 15 yards and a first down, moving Tampa out to their own 35. The entire drive consisted of this little equation of whether the yards Green Bay surrendered were worth the time that ran off the clock.

I’ve been railing against overuse of prevent defenses for as long as I can remember watching football, and I frankly cannot believe in this day and age that actual NFL teams will still run it like the Packers did on Sunday. I used to have only my intuition and the philosophical points made above, but now we have a better way to check whether or not this is a good idea: we can use Win Probability Added (WPA).

WPA is the awkward cousin of Expected Points Added (EPA). Both use historical data to model just how helpful a given play was by taking the state of the game before a play compared to after a play, and provide you the value of the play by subtracting one from the other. As an example, when Tampa started this drive from their own 11, they were in a bad spot. According to the EPA model, they were only expected to score an average of 0.3 points and had only a 6% chance of winning the game. The first 9-yard pass to Russell Gage improved their situation, increasing their EP from 0.3 to 0.7 (an EPA on the play of +0.4) and increasing their win probability from about 5.7% to about 6.7%.

The reason we want to use WPA here is because EPA is a largely situation-neutral stat (though it does account for time remaining). It only tells you about how many points you’re likely to generate in the current situation and doesn’t care how much you lead or trail by. WPA, on the other hand, is all about the situation, and as time slowly ticks down in a close game, major shifts in WPA can occur. Most importantly, for our purposes, WPA can tell us whether trading time for yards is a good idea or not. It sees Tampa with a 2^nd and 1 with 2:39 remaining from their own 20 as a better situation than 1^st and 10 with 3:40 left from their own 11. The 25 seconds of clock the Packers “gained” were not as valuable as the 9 yards the Bucs gained.

Tampa ran 14 plays on this drive. (14 plays! That’s so many plays! They were never going to run out of time.) Of those 14, only 5 did not improve their chances to win the game. Of those 5 plays, 3 came at the end of the game when Tampa was close to scoring and when Green Bay could no longer functionally play Prevent because they were backed up against the end zone. Of the two negative plays they actually had while approaching the red zone, the first was a 5-yard pass to Gage with 2:19 remaining on 1^st and 10. This was one of the few plays where prevent actually did work, limiting the Bucs to a short, inefficient gain in the middle of the field and killing 17 seconds of clock. Tampa lost 0.3% WPA on the play, and had Green Bay tightened things up on second down, we might not be having this conversation. Instead, Leonard Fournette got loose for 17 yards, increasing Tapa’s odds of winning by almost 5%.

The second came with 1:34 remaining, where Tampa suffered a 1% downturn in WPA on an incomplete pass. Green Bay was still in prevent, but played slightly more aggressive post-snap, blowing up what I believe was intended to be a bubble screen.

The last 3 negative plays came near the end zone consisted of a quick checkdown to Scotty Miller with 33 seconds remaining when De’Vondre Campbell jarred the ball loose, a running play to Fournette which Green Bay stuffed, and an incompletion to Cameron Brate with 21 seconds left.

The takeaway, though, is that while Tampa was moving down the field, it was never worth it to trade time for yards. Every successful play, save the 5-yarder to Gage, made it more likely that Tampa would win the game. And if not for De’Vondre Campbell — and maybe the Jumbotron — Tampa might have pulled it off.

When the game is on the line, and you need a stop, you can’t be giving this cushion.