Watchtower Review: Lions at Buccaneers

>> 9.13.2011

From the season’s first Watchtower:

While highlighting the extreme paucity of data, I’m going to swallow hard, wipe the sweat from my forehead, tug at my collar and project the most likely outcome to be another narrow Lions victory: 24-21, albeit in regulation this time.

Now, the final score was 27-20, but the Watchtower only accounts for offense-defense interaction. For the purposes of Watchtower evaluation the “real” score was 27-13. Before you folks cry foul, a quick reminder: the goal of The Watchtower is to spot systemic advantages that the Lions’ offense and defense have over their opponents’ schemes. Counting special teams and return scores throw off our evaluation of those effects.

Detroit Lions defense vs. Tampa Bay Buccaneers offense:

I can’t identify a systemic advantage from one data point. Disproportionately disrupting scoring by stopping drives with sacks is the design goal of the defense. I’m kind of stunned it worked with Blount being so devastating; I expect the Bucs to give Blount more than 15 carries this Sunday. We have no current performance data to go on, so I’ll have to recycle last year’s. I project the Bucs’ offense will meet expectations, scoring 20-23 points. I have extremely low confidence in this projection.

Right off the bat, we see the enormous difference: instead of LeGarrette Blount destroying the Lions at a 7.33 YpC clip, he was held to just 15 yards on 5 carries; a 3.0 YpC. As a team, the Bucs didn’t fare much better: 16 carries for 56 yards, 3.50 YpC.

The Lions’ pass rush didn’t hit home quite as often as last year’s contest, or for as many negative yards (two for -7 vs. three for -25). But, there was just enough pressure to rattle Freeman a bit.

Freeman’s completion percentage was nearly identical, 65.1% to last year’s 65.6%, but his yards-per-attempt was down dramatically. Freeman netted only 5.78 YpA this year—well off last year’s average of 6.80, and well well off the 7.64 he managed against the Lions last season. Between the there-enough-to-be-felt pressure and the excellent downfield coverage, the Bucs’ passing offense had to settle for underneath stuff—again, depressing scoring.

The Lions also intercepted Freeman once, and forced four fumbles, recovering one. Last season, the Bucs didn’t give away the ball at all, so two turnovers this time doubtlessly depressed scoring. We don’t have season averages to work with, but I ascribe this better-than-expectations performance to the Lions’ defense skill level being vastly improved. Over the course of 2011, I expect the Buccaneers to ouutperform their 2010 scoring average—meaning this Lions defense performed like an above-average unit on Sunday.

Detroit Lions offense vs. Tampa Bay Buccaneers defense:

The story is basically the same in reverse: the Lions more-or-less met expectations last time. Without any established trends for this year, I can only project the Lions’ offense to meet expectations, scoring 21-24 points. I have extremely low confidence in this projection.

Of course, the Lions offense did outperform my “wildazzed guess,” as one MLive wag (downbeat) called it. As with the Bucs’, the Lions rushing attack wasn’t nearly as effective this time around. 35 carries for 126 yards netted only 3.6 YpC—nearly identical potency. But twice as many carries is a strong indicator that the Lions were A) moving the chains and B) winning.

There’s an old saying amongst young statisticians: “You’re not winning because you’re running, you’re running because you’re winning.” The Lions weren’t pounding the rock any better than the Bucs were. They were just effective enough to control the ball and deny the Bucs possession. Tampa Bay couldn’t do the same thing because they were playing catch-up. Once they abandoned the run, the Lions could focus on pass rush and coverage—which, I’ll say it again, was excellent.

Through the air, Matthew Stafford’s execution was visibly worse than it was in the preseason—which, since that was quite nearly perfect, was to be expected. He completed 24-of-33 passes—that’s an astounding 72.7%. He wasn’t just dinking and dunking either: he netted 9.24 YpA; equally astounding. He threw three touchdown passes, not including a perfectly-placed TD pass Brandon Pettigrew dropped, and a wide-open Calvin Johnson He threw one interception, counting one tipped catch that came down in Aqib Talib’s hands, and not counting another one Jahvid Best basically shovel-passed to a stone-handed defender.

Even though the story of the game was correctly told as, “Stafford had the butterflies early and then settled down,” Stafford’s “jittery” performance was still outstanding by any measure. The ongoing performance of the Buccaneers’ defense will tell us a lot about how good the offense was this day. If the Bucs’ D turns out to be as good as many expect, this was an excellent statistical showing by an offense likely to average 30+ points per game the rest of the way out. If the Bucs’ defense is more hype than substance, then this was a pretty okay showing—but not near what we think the offense will be capable of.

Conclusion

I projected a close Lions win in regulation, 24-21. I can’t call it “right” or “wrong” because there was neither a long track record of coordinator matchups, nor any in-season data with which to establish expectations. This 27-13 performance showed us a lot of good things, though: the defense stopped the run and covered downfield better than we expected, and the running game was just good enough to control the game after Stafford built a lead.

There were a lot of mistakes, a lot of missed opportunities, and a lot of room to improve. But overall, this Lions team won on the road against a notional playoff contender—and looked better doing it than we could have hoped for.

7 comments:

Matt,  September 13, 2011 at 11:05 PM  

I think you need to find a way to consider defensive and special teams returns in your analysis. I understand what you're shooting for, but you leave some stuff out because of it. For instance, you're discounting the 7 points from the Talib Pick-6. What if a Lion had managed to catch him at the 1 and then Blount hammered it in from there? You'd count it against the Lions' defense, but is that more fair than NOT counting the Pick-6? You also count the Bucs first field goal against the Lions, but that was set-up by a 78-yard kick return. The Bucs started off in field goal range, went nowhere, and kicked the field goal. Again, is this really fair to count against the defense?

I understand your objective with the Watchtower posts, but, if upon post-game analysis, you're going to pick-and-choose which stuff "counts" and which doesn't, the whole endeavor becomes a lot more subjective.

Rob Hockey8player,  September 14, 2011 at 8:13 AM  

It is true. But I still love the watchtower

Ty Schalter September 14, 2011 at 8:32 AM  

Matt--

Well, it's worth pointing out that I've been doing that all along, both in the Watchtowers themselves and in the reviews. When I go back and analyze historical matchups, I check for D/ST TDs and discount them.

I realize this is imperfect, since the special teams can dramatically affect the field position each unit starts from. But projecting that would be an exercise in futility. Special teams units vary wildly in personnel from year to year, and even the same players vary wildly in production from year to year (remember Eddie Drummond?). How could I project that the team with the fewest kickoff return TDs to their name would send out a guy with 14 career kickoff returns in two seasons as a part-time returner, and he'd burn a theoretically-excellent coverage team for a 78-yarder?

Besides that, a lot more gifted analysts than me have shown that whether or not turnovers get returned for scores is completely random; fumbles and fumble recoveries are random, too. Interceptions are not--they're a function of coaching, decision-making, and defensive back skill. As a consequence, I do track turnovers, figure them in the analysis, and note when I think they'll affect my projections.

Again, simply removing special teams play, and defensive scores, isn't perfect. But there's no consistent way to project them, and including them in my analysis of offense vs. defense will just ruin whatever value that analysis has.

Peace
Ty

Anonymous,  September 14, 2011 at 4:17 PM  

I see what you are doing there Ty. If you can factor out ST and defensive scores, why cann't you give points to an offense say if they turn the ball over in the Red Zone. If the team scores touchdowns at a fifty percent rate and thirty percent for field goals in the Red Zone, do the math of (7*5)+(3*3)/10=4.4 for an adjust to the score. I agree with Matt that without the return, the watchtower score would have been 27-10.

Next subject. Why is Cherilus getting the biggest bone head play late in the game. Looking at thier field position, did he really cost them yardage, considering the punt was a touch back? Why isn't Aaron Berry's mental mistake not being talked about? After a great Buc's catch, he allowed his player to roll out of bounds to stop the clock. To me that could be considered just as big as the first one.

Rip

Unknown September 14, 2011 at 9:32 PM  

The Gosder mistake wasn't about the yardage, it was about stopping the clock on 3rd down, when the Bucs had no more timeouts. Without that penalty the Bucs have 45 less seconds to work with and get the ball with roughly 30 seconds left, and no timeouts.

Berry's mistake was not smart football, but Gosder was just downright stupid. There is a difference in my opinion.

Good work Ty.

Tiger
TigerReport.Blogspot.com

Matt,  September 15, 2011 at 12:36 AM  

First, I totally disagree that fumbles and fumble recoveries are "random." Unpredictable, sure, but not random. Runningbacks get reputations as "dirty fumblers" or the opposite (remember the old Marshall-Faulk-in-a-video-game-motion-capture-suit commercial where he insisted "I don't fumble"?) and defenses get reputations for "flying to the ball." That can't be random. The Tampa 2 in it's heyday with Derrick Brooks, John Lynch, etc. excelled at creating turnovers because it was part of the system. Players were coached up on the fundamentals of the Tampa 2 (keep everything in front of you, force the offense to execute long drives, all 11 defenders swarm to the ball, etc.) and "sold" on these concepts because forcing long drives meant more playmaking opportunities (i.e. sacks, fumbles, INTs) for these defenders. Michael Vick & Brett Favre had twice as many fumbles in 2010 as Eli Manning & Matt Schaub. I don't think Eli & Schaub simply fell on the "lucky" side of the probability curve.

I'd be interested to see what these other analysts looked at in determining that turnovers are "random" as I would also disagree that turnover-returns-for-TDs are purely random. For example, the thing that made Stafford's throw on the Pick-6 so bad was that a pick there is almost surely going for 6 (as opposed to a pick further down the field). His lack of touch on that particular throw raised the percentage chance that one of the two "bad things" that can happen when you throw would happen. Also raising this percentage are the defensive system (the Tampa 2.1 still frequently has the corners prowling the flat after releasing the WR) and the particular CB, Aqib Talib, a reputed ballhawk, that the Lions were facing. I'm not saying someone could have predicted that play, but that play wasn't exactly "random."

I guess the problem I have with the analysis is that you are specifically focusing in on the coordinator vs. coordinator match-ups using historical data (ignoring, somewhat, the players that make up those coordinators' units), but then predict a specific game's SCORE from this discounted data. You then go back and analyze why that prediction was near or far from the actual result. This is the exercise in futility, not trying to actually incorporate more data. The answer to the question "Why wasn't this Watchtower prediction more accurate?" is the same every week. . .'cause you're leaving a bunch of stuff out and massaging the numbers to try to account for it.

I'm not trying to bash you for not having a perfect system (obviously, no one does) and you always include the necessary caveats and disclaimers. The Watchtowers themselves are very interesting and the score predictions based simply on Watchtower analysis is practically obligatory anyway. Also obligatory is going back to each Watchtower after the game and seeing how you did. In this respect, if the aim is coordinator vs. coordinator analysis, that's where the re-analysis should focus and kinda' leave the score out of it. Again, it's basically the same story each week: "Well, I didn't account for this, that, or the other thing, so I predicted this but the score was that. The End." If you want to hang your hat on score predictions, you're gonna' need a deeper system. Leaving out certain phases of the game assuming they all just even out stastically isn't going to cut it.

nwashuta September 15, 2011 at 2:58 PM  

Matt,

There are a ton of problems with trying to make the report as detailed as you said. First, you'd have to determine specific offensive formations and the resulting targets that are likely to happen(which need a huge sample size and are very context specific). From that, you'd need to determine the likelihood of a turnover from that target (again, an even bigger sample size), and you'd need account for the probability of a turnover from that spot on the field resulting in a defensive TD. Once that's done, you would need to adjust for the defensive scheme (do corners cover the flats or do linebackers? cornerbacks are faster and have better hands, so they're more likely to intercept and return the ball). With all of this completely unavailable data, you could get some decent model of what will happen. However, with the standard deviation of those probabilities probably being fairly high, your error bars would be huge and the lions would probably just end up running the ball and wasting countless hours of your time.

The watchtower does a nice job in that it takes a big picture view. How does the coordinator typically gameplan and how effective are his teams within his scheme? You know the patriots are going to throw the ball and most likely very well. You know the lions typically have a pretty poor pass defense. Therefore, the pats will probably put up a ton of passing yards and TDs on the lions. (ahem, thanksgiving day, 2010).

The watchtower is using statistics in a very unique way to try and smooth out the random variances to scientifically arrive at a score prediction rather than a beat writer or FM jockey pulling a number out of nowhere. Sure, it's not perfect, but I enjoy it.

Post a Comment


  © Blogger template Simple n' Sweet by Ourblogtemplates.com 2009

Find us on Google+

Back to TOP