clock menu more-arrow no yes mobile

Filed under:

Statistical In-Ferentz, Week 6: Mirror, Mirror

Statferlogo_medium

Did anyone else get the feeling watching Iowa play Penn State that you were seeing a man fight his doppelgänger? The defensive style, the absurdly conservative strategy, the overall game plan -- it was as if some transporter accident had separated the same football mind into two bodies and then the doubles were forced to battle to the death. It was a fascinating kind of defensive football duel that, depending on your view, either represented the apotheosis of real football or its death by slow suffocation. 

The similarities started on defense. I re-watched all of Penn State's defensive plays against Iowa's offense, and found the same thought kept going through my head: "Oh, this is what it must have been like to play against Norm Parker's defenses the past few years." I'm sure there are subtle differences between the two that I'm missing, but the fundamental approach both coordinators take seems to be the same: they both trust their defensive lines to control the run game without much help, play lots of fairly deep zones for much of the game, and try to limit big plays. The challenge both defenses seem to offer opposing offenses is the same: is the quarterback disciplined enough to march down the field five yards a time without screwing up?

On Saturday, Penn State relied almost exclusively on three and four deep zones, and while they did blitz a fair deal, they always left zone coverage behind. The Nittany Lions were absolutely committed to not letting Iowa get any kind of deep passing game going, and were successful in that regard. You can see that clearly by looking at James Vandenberg's passing chart by area of the field:

Vandenpass_medium

Vandenberg went 0/3 in the deep areas of the field (15+ yards past the line of scrimmage), which is an indication both of how well-covered those areas were and of how effective Penn State was pressuring Vandenberg to get rid of the ball quickly. Vandenberg was most efficient in the short zones, going 8/12 for 47 yards, but that's exactly the kind of small change the zone is designed to give up.  The middle zones were the real horror show: Vandenberg went 9/19 with two interceptions in the middle area of the field. These were the areas where Vandenberg was forced to make difficult throws between linebackers and safeties, and he and his receivers clearly had some problems connecting there. Urban Meyer raved about Vandenberg's ability to move a linebacker with his eyes on a particular play, but the replay of that play emphasized just what kind of mental gymnastics Penn State's zone required of Vandenberg all day. There was none of Pitt's linebacker-man-coverage-on-KMM strategy here, just lots and lots of blue shirts sitting in wait for him to make a mistake.

The one area where Tom Bradley's defense looked nothing like Parker's was in his use of the blitz. Penn State blitzed a lot, on 19 out of Iowa's 64 times (30%), and even though Parker sent a few blitzes (Hyde and Bernstine most notably), it was nowhere near that number. Part of that was the result of Iowa being down and needing to pass (10 of the 19 PSU blitzes came after the 8:08 mark in the fourth quarter, when Iowa was down 10), but part of it was a recognition by Bradley that Iowa's response to the blitz, both in protection and quarterback recognition, was not very good. The most significant play of the game came on a blitz: the strip-sack of Vandenberg early in the fourth quarter, when Iowa was still down three points and driving in Penn State territory. And Vandenberg's numbers against the blitz were gruesome all day: 2-10, 33 yards, fumble, interception, four sacks. And 25 of those yards came on the meaningless 25-yard pass to McNutt on 4th and 40 at the end of the game. Iowa wasn't burning up the field against Penn State's defense to begin with, but Bradley's use of blitzes on obvious passing down was a real stroke of genius and caused Iowa no end of trouble. Expect other teams to watch the tape of this game carefully.

Notes on Strategy

In his excellent write-up, Ross criticized some of the more controversial coaching decisions of the game and wrote "I don't even need statistics to back me up." Come on, Ross, everyone needs statistics!  But it's true, the errors were pretty egregious. I saw three big ones: 1) punting from the Penn State 33 in the first quarter, 2) settling for a field goal from the Penn State 5 in the first quarter, and 3) punting from the Penn State 41 in the second quarter.  You could probably add Iowa only semi-trying to score before half-time, but I didn't have time to look at that closely. On the other three decisions, I applied the same WP analysis as I did in an earlier post*, what I found was that on all three decisions, a generic NFL team would have been better off trying for the fourth down conversion if they felt they could make it about 25%-30% of the time (specifically 23.5% on #1, 31.8% on #2, 28.6% on #3). And that analysis assumes that the punts in question would land exactly where they did, which is a pretty favorable assumption (they very well could have been touchbacks). It's possible, I suppose, that Iowa's offense simply wasn't capable of converting even 30% of the time in those situations (4th and 8, 4th and 4, 4th and 5), but I sincerely hope not. If that is the case -- if Iowa's offense can't get five yards one out of every three times -- then it's not really surprising that the team only scored three points; it's hard to imagine an offense sustaining any kind of drive down the field with that kind of haphazard success.  
* I didn't want to bog things down here, so I put the technical details in an endnote if you're interested.

Luckily for Iowa, Joe Paterno matched Kirk Ferentz conservative decision for conservative decision. Specifically he settled for field goals from the Iowa two and three yard-lines. By my analysis, if Penn State could have converted either fourth down about 40% of the time or more, going for it would have been the better decision. And given Penn State's success running the ball against Iowa, it seems likely that they could have converted on those plays. The first decision to kick the field goal was especially strange, coming as it did on the heels of a smart  fake field goal call on 4th and 4 from the 17. Why 4th and 4 from the 17 was a good gamble but fourth and 2 from the two was a bad gamble is not clear to me.
 
It's fun to point out the ultra-conservative foibles of Paterno and Ferentz, but I can see a way that their thinking makes a certain kind of sense, assuming one thing: that each coach assumes he has a very good, very consistent defense. Here's an analogy that helped me think about it: imagine that the outcome of a football game is determined by rolling two dice, one for offense and one for defense. If the number a team rolls for offense is greater than the one it rolls for defense, that team wins. Now suppose a coach has the following six-sided dice to choose from:

Defense: 2,2,2,2,2,2
Safe Offense: 3,3,3,3,3,3
Risky Offense: 1,1,1,6,6,6

In this scenario, even though the average roll for the risky offense would be higher than the safe offense, it makes no sense to ever choose the risky offense. Why? Because using the safe offense guarantees a win 100% of the time, while the risky offense gives only a 50% chance of winning. This is what I imagine overall strategy might look like to a coach with a really good defense. If he knows that the defense can reliably hold the opposition in the 0-10 point range, it is perfectly logical for him to avoid high variability results on offense, even if riskier strategies would produce higher point totals on average. The high quality and low variability of the defense means that the coach can afford to rely on less risky offensive strategies and still win -- after all, winning by one is as good as winning by 50.

This is pretty simplified, and who knows if coaches actually think this way, but I think some kind of mentality akin to this is required to explain things like kicking field goals from your opponent's two yard-line and punting from the 33. From the perspective of the coach with the good defense, it's not about leaving points on the field, it's about taking the possibility of bad outcomes off the field. And in certain situations, every coach quite reasonably adopts this mentality.  For example, say the New England Patriots had the ball with two minutes left and a seven point lead: they probably wouldn't keep passing, even though they would probably score more points that way. They would run, because that would reduce the chance that something disastrous would happen, like a pick-six. The Ferentz/Paterno low-risk mentality seems like a way of extending the endgame logic to the full length of the game: if the coach knows with some confidence that his defense can win the game with minimal offensive help, then figuring out a way for his offense to score 40 points becomes less critical, and limiting downside risk becomes paramount.*

Of course, making this strategy works requires accurate knowledge regarding whether your defense is that good and is that consistent, and that your "safe" offense is that safe. If the "defense" die read 4,4,4,4,4,4 instead of 2,2,2,2,2,2, suddenly the safe offense would go from a 100% winner to a 100% loser and the risky offense would represent your only chance of winning. Penn State did, in fact, seem to have a very good defense on Saturday, and their low-risk strategy worked out (although it was an uncomfortably close-run thing until that strip-sack). On Iowa's side of the ball, though, it's an open question whether Iowa is playing with the "good defense" die anymore. Compared to past years, it doesn't look like it. And if that's the case, then the kind of punt and defend, settle for field goals strategy doesn't make as much sense. Especially against teams that are more daring than Penn State, Iowa may have to start thinking of rolling sixes instead of threes. That's not a pleasant situation to be in -- there's always the possibility of rolling a one and losing big -- but that's life as an underdog.

* I actually kind of hate this way of thinking about football and don't think it makes much sense outside of certain fourth quarter situations. The reason why is that it forces the coach to assume that he knows more about how the rest of a game will turn out than he in fact does. You may think 17 points will be enough to win the game, but what if something strange happens? What if your defense has a bad second half, or your safe offense suddenly fumbles the ball away? Suddenly you might find yourself needing 24 points and wishing you had just picked the plays with the best overall reward, not the ones with the least downside. I don't know -- what do you think?

WP Analysis Endnote:

The general idea here is to look at a decision point (e.g. going for it on fourth down vs. punting), calculate the benefits of each branch from that point in terms of win probability (WP values from the Advanced NFL Stats calculator) of all the outcomes (e.g. converting, not converting and punting), and adjust the probability of converting (p) and not converting (1-p) until the two sides are equal. This gives you the "break-even" probability where the expected value of both options is the same. It's a way of judging just how confident you would need to be in, say, converting a fourth and 5, to be indifferent between going for it and punting:

p*WP Change from Converting + (1-p)*WP Change from NOT Converting = WP Change from Punting

In all these cases, I simplified things by just assuming that the right side of the equation (e.g. landing a punt at the 10 yard line, making the field goal from the 2 yard line) was a 100% guaranteed affair. That's not totally accurate -- the punt could be returned for a touchdown, the field goal could be blocked or just miss -- but at least this way we're giving the conservative option the most optimistic result. If the conservative option doesn't make much sense even assuming things go perfectly, then that's all the more reason to go with the more risky play. Also, one last caveat: the Advanced NFL Stats model is based on NFL games, so these numbers only roughly apply to college football. I still think it's a good way to think through the costs and benefits of the situations, but take the precise figures with a grain of salt.

Iowa Decision #1: 4th and 8 from Penn State 33, game tied 0-0, 10:40 left in first quarter [Iowa punted]
* WP increase if Iowa converts 4th down: +.10 WP [.53 to .63]
* WP decrease if Iowa fails to convert and turns the ball over on downs at the 33: -.07 WP [.53 to .46]
* WP result of pinning PSU at 10 yard-line: -.03 WP [.53 to .50]
* WP increase if Iowa makes 50-yd. FG: +.06 WP [.53 to .59]
* WP decrease if Iowa misses FG: -.07 WP [.53 to .46]

* Break-even probability [go for fourth-down conversion]: 23.5%

* Break-even probability [kick field goal]: 53.8%

Iowa Decision #2) 4th and 4 from Penn State 5, Iowa down 3-0, 10:51 left in first quarter [Iowa kicked FG]
* Benefit of converting fourth down and touchdown*: +.15 WP [.37.47 to .62]
* Cost of failing to convert: -.07 WP [.47 to .40]
* Benefit of kicking field goal: .00 WP

* Break-even probability [go for TD]: 31.8%

Iowa Decision #3: 4th and 5 from Penn State 38, Iowa down 6-3, 6:45 left in third quarter [Iowa punted]

* Benefit of converting 4th down: +.10 WP [.39 to .49]
* Cost of failing to convert: -.04 WP [.39 to .35]
* Benefit of pinning PSU at 4 yard-line: +.00 WP

* Break-even probability [go for 4th-down conversion]: 28.6%

Penn State Decision #1) 4th and Goal at Iowa 2, 1:36 left in first quarter, game tied 0-0 [Penn State kicked field goal]

* Benefit of converting 4th down and touchdown: +.12 WP [.59 to .71]
* Cost of failing to convert: -.09 WP [.59 to .50]
* Benefit of kicking FG: +.00 WP

* Break-even probability [go for TD]: 42.9%

Penn State Decision #2) 4th and Goal at Iowa 3, 4:59 left in second quarter, game tied 3-3 [Penn State kicked field goal]

* Benefit of converting 4th down and touchdown: +.12 WP [.59 to .71]
* Cost of failing to convert: -.09 WP [.59 to .50]
* Benefit of kicking FG: +.00 WP

* Break-even probability: 42.9%