Three cups deep: the watchtower

I began this series six weeks ago, as a way to scratch two different itches simultaneously: 1) creating a unique way to preview upcoming Lions games, and 2) exploring the interplay of offensive and defensive systems.  It seems like this is something that almost completely ignored by the greater football-watching public; what happens when one specific offense meets one specific defense.  Even when delving into the dark heart of football knowledge, I don’t see as much consideration given to this aspect of the game. For example, when I read an explanation of an offensive system, it's often chalked up in a defenseless vacuum, or matched against a vanilla defense:

Chris Brown – Smart Football

In my mere two decades of consciously watching football, many offensive and defensive systems have fallen in and out of favor--sometimes in direct response to one another.  Famously in Detroit, we saw the June Jones/Mouse Davis “Run n’ Shoot”.  Its pure, original form was shorn of much of its effectiveness by the rise of the Lebeau/Capers zone blitz–based defenses.  The R&S often employed a rollout or half-roll, and the RB would be used either to pass protect, or to chip and then release to a screen.  Traditional blitzes would be picked up by the RB, and LBs in zone coverage would assume the RB was blocking—and disregard him.

As Chris Brown of Smart Football explains, the zone blitz threw that for a loop:

Back in his days with Texas A&M, Bob Davie was an innovator. Against run and shoot teams like the University of Houston, he would run his 3-4 defense, blitz his outside LBs (thus forcing the RB to stay in and block), and drop off defensive linemen and interior linebackers so he could still play zone with six to eight defenders. As a result the R&S's protection and formation scheme broke down. They blocked with six, had the running back on a bad matchup with a good OLB, faced an unblocked rusher, but the defense still had 6-8 guys in coverage, so the R&S's "hot reads" and breakoffs did not work either.

This is pure scheme-on-scheme interaction; it's nothing but a numbers game.  Without the zone blitz, there's a threat of one more receiver than defenders.  With the zone blitz, there's one more rusher than can be blocked, and no open options.  The zone blitz team will have “a good day” on defense, and the R&S team will have “a tough go of it” on offense, almost entirely due to the interaction between the two schemes.  

Many game previews are either a position-group-versus-position-group breakdown, or a laundry list of mitigating/extenuating circumstances that might alter hinder or augment each team’s typical performance to date.  I was not satisfied doing that type of breakdown; there are plenty of sites that already do it well.  Further, I was curious: was there a way to analyze—and, possibly, predict—how the Lions’ Xs-and-Os would interact with their upcoming opponent’s?

I am not a football coach, and I can't throw 80 hours a week into this stuff; I can't "gameplan" for upcoming opponents, replicating the Lions’ coordinators’ work as they do it.  However, I CAN analyze historical statistics to capture the end results of previous interactions between the schemes, and then project them forward.  Using each team’s annual per-game averages for points scored (/allowed), and per-play averages for passing and rushing gained (/allowed), I can see when there’s a statistically significant deviation.  I can also factor in the end-of-year ordinal rank for each scheme’s scoring offense and defense, which gives us an idea of overall skill level.

In the inaugural installment of the Watchtower, I first looked at Scott Linehan versus Saints DC Gregg Williams.  I concluded:

Given equal talent and execution, Gregg Williams’s attacking 3-4 defense will disproportionately disrupt Scott Linehan’s balanced offense. Given superior talent and execution, and/or excellent pass protection, Scott Linehan’s balanced offense disproportionately gives Gregg Williams’ attacking 3-4 defense fits.

Given that the Saints' D has been suspect for several seasons in a row, and the Lions’ bevy of young talent on offense, it seemed rational to expect the Lions to score "above expectations"—but, then I had no idea what those expectations should be. 

The Lions did score 27 points against the Saints--more than they had all year in 2008--but 7 of those were from a defensive TD.  Moreover, the Lions had an unusually strong performance from their return game, shortening the field and making it easier on the offense.  Removed from the “first game as a pro” context, Stafford’s 16/27, 205 yard, 3 INT performance could only be considered awful.  Ditto for Kevin Smith’s 15-carries-for-20-yards showing.  Frankly, the Lions were lucky to put up 20 in this game.  In this case, my initial analysis was right—the Lions’ equal-or-lesser talent should underwhelm against Williams’ defense, and it did, on a per-play basis.

On the defensive side of the ball, my analysis of prior Sean Payton/Gunther Cunningham matchups led me to conclude:

Given equal talent, Cunningham’s hyperaggressive 4-3 is extremely effective against Payton’s pass-heavy offense, but only if that aggression leads to mistakes and turnovers—otherwise, the holes in the defense will be exploited. Effective quarterback play may neutralize the defensive advantage.

. . . would you call 26/34 for 356 yards and 6 TDs “effective”?  Good, because so would I.  Mike Bell, of all people, gashed the space underneath the traumatized Lions’ secondary for 143 yards on 28 carries. 

Though there's historical evidence that a decent Cunningham defense, when successful, is disproportionately disruptive of a Payton offense, the talent gulf between the Lions' D and the Saints' O is enormous. Unless the Lions generate three or more turnovers, I don't see their defense having any kind of success in slowing the Saints down. Therefore, the most probable outcome of this game is a shootout that the Lions lose. There is a chance that the Lions' defense disrupts the passing game early, and that the Lions score on their first two possessions, thereby allowing the defense to safely turn up the heat--and the offense to put it in the cooler. However, the offense will have to overcome a systemic disadvantage with talent, and the defense will have to overcome a significant talent gap with a perfectly-executed gameplan.

On the balance, this prediction was roughly accurate: the Lions defense didn't have any kind of success in slowing the Saints' offense down, the offensive gameplan wasn't perfectly executed, and when the Lions got down early the gameplan went out the window anyway. I think the data was here for an accurate conclusion, but A) I didn't take it far enough and B) we had no idea what to expect from the Lions, or their offense.

In the second Watchtower piece, I looked at Scott Linehan’s track record against Vikes’ DC Leslie Frazier.  I saw some strong patterns emerge, and concluded:

Given greater or equal talent, Scott Linehan’s balanced offense significantly outperforms its averages when facing a Dungy-style Tampa 2--especially with the run. Given lesser talent, Linehan’s offense meets or mildly outperforms expectations against a T2. However, a disproportionate amount of sacks and turnovers seem to be created by a Tampa 2 when facing a Linehan offense.

The Lions did indeed turn in a better overall offensive performance against the Vikings, despite scoring fewer points.  Stafford completed 18/30 (60%) of his passes, for 152 yards, his first TD, and 2 picks.  Kevin Smith turned in a much more Silent Bob-like performance, grinding out 83 yards on 24 carries against one of the most feared run defenses in football.  However, those two picks, 2 sacks that killed the opening drive of the second half, and a horribly-timed Smith fumble snuffed out the Lions’ chances of shocking the world with a home-opener upset of the (currently undefeated) Vikes. 

On defense, I analyzed Brad Childress’ track record against Gunther Cunningham, and found that:

Given equal or lesser talent and execution, Gunther Cunningham’s hyperaggressive 4-3 disproportionately disrupts Brad Childress’s conservative Walsh-style offense, especially in the running game. However, a very effective deep passing game can stretch the defense, reduce QB pressure, and produce points.

Now the Vikings never got the deep pass working. However, after being shut out (!) in the first half, Childress and Favre adjusted to the surprisingly effective Lions defense in another way . . . In fact, let me just quote NFL.com's Bucky Brooks:

The Vikings turned to a quick passing game to thwart the Lions’ blitz-heavy tactics. Brett Favre spent the early portion of the game under siege, but the decision to use an assortment of bubble screens, quick slants and underneath routes allowed the Vikings to move the ball against Detroit’s extensive use of Cover-0 (all-out blitz) in key situations. While the Lions mixed in some two-deep coverage, the host of quick throws helped Minnesota move the chains and eventually enabled the Vikings to give the ball to Adrian Peterson in the late stages of the game to run out the clock.

Yeah, that pretty much covers it.  My prediction for the game as a whole was:

Therefore, the most likely result of this game is a closely contested, medium-to-low scoring slugfest, with a lot of turnovers and penalties. It is slightly more likely that Minnesota’s talent overcomes Detroit’s systemic advantages, but this will be a volatile game in Detroit’s home opener.

In this case, the general conclusion I drew was spot-on; the Lions lost, 27-13, in a tug-of-war that was closer than the final score suggests. There weren't a "lot" of turnovers--Stafford threw two picks and Favre none; both teams fumbled twice and lost one--but they played a huge part in the game.  I pointed out some of the more specific successes in the following installment: 

• I found that Linehan's offense seems to be unusually successful against Dungy-style Tampa 2 defenses; Detroit was so successful on the ground that the Vikings abandoned their base defense, and put eight men in the box.
• Linehan's teams seem to turn the ball over frequently when facing a Dungy-style defense. I noted that avoiding that would be key to the Lions' chances for victory. Instead, the Lions turned it over three times, and that was arguably the difference in the game.
• I projected that Minnesota's lack of a consistent deep threat would allow Gunther to blitz, thereby disrupting the Childress offense. It did, he did, and it did; the Lions nearly carried a shutout into halftime. In the end, the turnovers--and resultant short fields for the Vikes--were too numerous to overcome.
• It was indeed an ugly game marred by sacks, turnovers, and controversial calls--most notably the phantom Gosder Cherilus "chop block" that negated a long gainer by Megatron.

This analysis isn't anything more than that: analysis. But rather than pretend that Dominic Raiola had some sort of secret Pat Williams kyrptonite in his pocket, I prefer to believe that the interaction of schemes provided an opportunity for Kevin Smith to succeed . . . and I know of no other way to project or predict such interactions. I said as long as I find this analysis to have predictive value, I'd continue--and I do, and so I shall.

Right on, four-weeks-ago me!

In the third installment of the Watchtower, I had great data to work with for the Zorn/Cunningham matchup--four games, two of which occurred in the same year.  On the strength of those numbers, I concluded:

Given equal talent and execution, there is no systemic advantage for either Jim Zorn's WCO or Gunther Cunningham's hyperagressive 4-3. If this week's game follows the above trends, Clinton Portis will have an above-average day on the ground, but the Lions' blizting will disrupt and confound Jason Campbell and the Redskins' ho-hum passing game. Point production by the Redskins should be right at expectations--which, for 2009 so far, means 13.0; they're ranked 31st.

I had Portis and Campbell flip-flopped, but the Redskins’ offense was indeed one-dimensional.  Portis managed only 42 yards on 12 carries, whereas Campbell was 27/41 for 340 yards, 2 TD, and an INT. That 340 number looks huge, but not when you consider the number of points the Redskins scored: 14, or exactly 1 more than I’d predicted.  Nice. 

Unfortunately, the data for Skins' DC Greg Blache was totally specious; Blache’s Bears defenses weren’t anything like this Redskins defense.  Moreover, I fat-fingered one of the average calculations, so some of the data I was working from was flat-out wrong.  It showed in my assertions about the Lions’ O vs. the Redskins’ D:

Regardless of talent or execution, Greg Blache's philosophy of a strong front four and committment to run-stopping disproportionately slows Scott Linehan's balanced offense.

Oops. What actually happened was that the Lions ran all over the Redskins, carrying 36 for 154 yards; a stout 4.28 per-carry average. In the two-and-a-half quarters that Kevin Smith played, he gouged the Redskins for 101 yards on only 16 tries; that's an astounding 6.31 yards per carry. This was the exact opposite of the result that I predicted.

I really, really, really wanted to predict a Lions victory for the game, but I just couldn't bring myself to do it.  I wussed out, and instead predicted that “the team whose quarterback performs the best wins.”  Though I could argue that Matthew Stafford’s 21/36, 241 yard, 1 TD 0 INT performance was somehow “more efficient” or something, I’d be lying--and wrong.  The Lions won—won!—the game with a great running game, and a very good showing from the front seven.

In the the Watchtower article for Week 4, I only had one data point for Bears OC Ron Turner against Guntherball: The Bears at Kansas City, 13 years ago.  Knowing I'd be on shaky statistical ground, I used data from Turner's brother, Norv, against Cunningham's more recent units:

IF we consider Ron and Norv Turner interchangable--and we don't--then given greater, equal, or lesser talent, Gunther Cunningham's hyperagressive 4-3 appears to match expectations versus a Turner Bros. Coryell-style downfield passing offense (albeit while generating very high sack and turnover numbers). That is to say, there is no systemic advantage or disadvantage for either team.

Unfortunately, the Bears scored 48 points--counting a kickoff return TD--which exactly triples their average for every other game this year. Their numbers weren't impressive, partly because their average drive starting position was on the Lions' side of the 50, but suffice it to say that the Bears' offense did just fine. Given the turnovers, special teams, and field position, it's very hard to say how much better than expected the Bears really were. It's even harder to say how much of their success was due to intrinsic X-and-O advantage, and how much was due to victimizing a not-good Lions defense.

In the other matchup, Scott Linehan versus Lovie Smith, I had much better data. It led to a pretty specific conclusion:

Given greater, equal, or lesser talent, Lovie Smith's relatively aggressive Tampa 2 will surrender a disproportionate amount of yards to Linehan's balanced offense, but also generate high numbers of sacks and turnovers, disproportionately disrupting scoring.

• Stafford was sacked five times, for a loss of 42 yards.
• Stafford lost a fumble on one of those sacks, and threw an interception.
• The Lions generated a season-high 398 yards of total offense, and scored 24 points--for reference, they scored 20 points off of 231 offensive yards in Week 1.

That is the kind of accuracy that makes me want to keep honing this. With enough data, we get a great glimpse into how the upcoming game will go.  In the fifth installment of The Watchtower, I expanded the data I was looking at, by adding in average defensive yards per attempt/carry. This gives us a better handle on exactly how effective an offense is, aside from total points output.  I studied Pittsburgh OC Bruce Arians' historical matchups against Guntherball. My conclusion was:

The evidence is clear, and the verdict is terrifying: With quality quarterback play, and/or suspect secondary play, Bruce Arians's multi-WR downfield passing offense is disproportionately successful against Jim Schwartz's balanced 4-3, regardless of talent.

I then turned my eye to Scott Linehan’s historical success against Steelers DC Dick LeBeau--and found I had only one data point to

Given the only data point on LeBeau, and fitting it into the broader picture painted by the Capers and Williams info, I think I'm safe to say that Scott Linehan's balanced offense significantly outperforms expectations against aggressive, blitzing 3-4 defenses like LeBeau's.

Well no, I was not safe to say that; I forgot to include talent into the equation. With Stafford out, and Kevin Smith dinged up, the offense actually underperformed on a per-play basis. I was tantalizingly close, though; I could feel it. At this point my obsession with figuring this out was reaching fever pitch. In the sixth, most recent Watchtower, I spent the first several paragraphs breaking down the results of the fifth one--no need to regurgitate it all here.

I broke down the stats of Packers' head coach Mike McCarthy two prior meetings with Gunther Cunningham, and drew this conclusion:

given equal or greater talent, Gunther Cunningham’s aggressive 4-3 disproportionately disrupts Mike McCarthy’s downfield flavor of the Bill Walsh offense. Given lesser talent, Gunther’s 4-3 will cap offensive production with sacks and turnovers, even while allowing better-than-average offensive effectiveness between the 20s.

And what happened?

• The Packers scored 26 points, which exactly matches their average on the season so far.
• Aaron Rodgers completed 29 of 37, for a whopping 358 yards (and 9.68 yards per attempt!).
• Rodgers, however, was sacked five times, and intercepted once. The Lions also forced three fumbles, recovering one.
• Rodgers passed for only two touchdowns, and those were on the first two drives (one of which started on the Lions' 17).
• In fact, the Packers as a whole did not score a touchdown after cashing in on the opening-drive Culpepper turnover.

That is spot-on, to the letter, exactly what I'd predicted. The Lions' secondary was victimized by Rodgers and the Packers' WRs, but thanks to the sacks and turnovers and red zone defense, the Packers repeatedly failed to cash in. If the Lions offense could just outperform their averages a little bit . . .

As we've seen with Gregg Williams and Dick LeBeau, Scott Linehan's balanced, conventional offense is disproportionately successful against an aggressive, blitzing 3-4. This will be the third such defense that the Lions face, and they've outperformed averages against the two prior units. If Kevin Smith is his usual, steady self, and Matt Stafford is able to play, I expect the Lions to score between 24-28 points.

Well, there we have it.  All else considered equal, the Lions' defense should hold the Packers' defense at or just below their season average, and the Lions' offense should slightly outperform their season average. Essentially, it's a push.  The data says that if Stafford, Smith, and Megatron can go, it's a 50/50 shot.  However, I was feelin' my oats. I liked this matchup, and all the Packers info I could find had Pack fans assuming a blowout was a forgone conclusion.

I decided to stick my neck out; have a little faith. I not only called for a Lions victory (albeit a narrow one), I sent it out to Packers bloggers everywhere.  Immediately afterwards, it became obvious that Stafford and Megatron were not going to play, and I had a real problem on my hands.

Obviously, pure scheme-on-scheme interaction cannot account for everything. I have to account for outside influences--like, the Lions banged up and desperately needing a bye, the Packers coming off a bye, the wind at Lambeau Field, Daunte Culpepper.  However, I've been most inaccurate when I've let those sorts of things color my predictions--or when I've tried to fill gaps in historical data with subjective guesses.

So, going forward, The Watchtower posts will consist of two components.  One will be pure, objective analysis based on historical data--and I'll be noting my confidence level, based on the amount and quality of data I have to work with.  The other will establish relative talent and execution levels, and note mitigating or influencing factors.  This way, we'll have as good of a grasp as is possible on the results of scheme/scheme interaction--yet also understand what might prevent those Xs and Os from playing just the way they do on the chalkboard.

Critically, I will also begin incorporating 2009 data into The Watchtower.  Up until this point, the Lions were really an unknown quantity--but now, we have sufficient data to make the words "outperform expectations" have meaning; we now know enough about these Lions to have accurately expectations going forward.  Perhaps most intriugingly, this will factor heavily into the remainder of the divisional games.

So there we have it.  I'm not sure what the ulimate endpoint of this series is. Will I end up with incredibly accurate projected final scores, key player statistics, and mathematical levels of confidence in each, and us all winning millions of dollars in Vegas?  Probably not. Will the greater football analysis world lift me up on their shoulders carry me around and crown me King Of The Football Geeks and I rule wisely for ever and ever?  Unfortunately unlikely.  But even if I routinely make a complete and total jackass of myself, as I did last week, I think this is a very cool way of breaking down Lions games, and focusing on what makes this team succeed . . . and fail.