clock menu more-arrow no yes

Filed under:

What You See, Statistics, and the Middle Ground Somewhere In Between

New, comments

Statistics are tricky. I remember sitting in a Michigan computer lab many years ago attempting to learn statistics from a graduate student whose grasp of English was, at best, limited. It wasn't an easy proposition. Statistics are, at their heart, the study of trying to discern how and why certain sets of numbers provide you with a reasonable answer to a question you haven't asked. At least that's what I told myself when I withdrew from the class. In hindsight, the fault for my lack of understanding may equally have been my own. College being what it is.

However, as I've gotten older I've found statistics to be a useful tool when trying to uncover meaning in sea of random numbers. Whether it's my spending habits, understanding economic theory, or summarizing a complex concept, statistics are a useful tool. And, I've found, they're the most useful when I'm using them for illustrative purposes.

For instance, if I wanted to make the assertion that offensive efficiency is not demonstrated by the total yards an offense racks up, but rather is demonstrated by the number of points per yard a team scores I can easily find numbers to support that. As a matter of fact, I can find "concrete" numbers to support this contention. It's a simple set of numbers and a simple formula. The concrete numbers are readily available, Total Yards and Total Points. Those are pretty concrete, right? So what's next? You take total points and divide them by total yards to derive your points per yard offensive efficiency statistic. Let's take a look at the Big Ten for the 2010 season.

Rank School Pts/Yard Yards Points
1 Wisc .0931397 5787 539
2 OSU .0864197 5832 504
3 Ill .0819449 5162 423
4 MSU .0758265 5051 383
5 Iowa .0755323 4978 376
6 Ind .0701075 4650 326
7 NU .0674665 5084 343
8 Mich .0670549 6353 426
9 PSU .0657160 4839 319
10 Minn .0641143 4336 278
11 Pur .0631184 3739 236

Wow. 6353 yards on the season, and Michigan ranked eighth in our new make believe statistic for offensive efficiency. As a matter of fact, they ranked behind Northwestern and Indiana. But, this isn't a real statistic. They're just arbitrary numbers mashed together. But for giggles, lets see what happened during the 2010 conference season.

Rank School Pts/Yard Yards Points
1 Wisc .1044431 3466 362
2 Ill .0907072 3153 286
3 OSU .0821673 3359 276
4 MSU .0758027 3021 229
5 Iowa .0740072 2770 205
6 PSU .0670205 3014 202
7 Pur .0664585 2242 149
8 Mich .0656914 3760 247
9 Minn .0645161 2666 172
10 NU .0610561 2030 185
11 Ind .0571822 2903 166

Holy crap! Are you serious!!?? Michigan's in conference offensive efficiency was worse than Purdue's! The funny thing is, if you look at the rankings, well... with the notable exception of Illinois, they make sense. Wisconsin, OSU and MSU were all at the top of the conference. Iowa, Penn State and Illinois were all bowl teams. And if you watched these team play for any extended period of time, the numbers makes sense.

Here's the rub. These numbers aren't "concrete" at all. They're a select set of figures derived from a complicated game that tell a limited story. Frankly, they're cherry picked numbers that support a contention. You look at the numbers and they obviously include overtime points, defensive scores, short fields, field goals, two point conversions, etc. There are too many variables to count, but if I want to stick by these numbers, I can. I can dismiss whatever I want as chance or irrelevant, because the numbers are the numbers. These numbers are right there in front of you to see, and they are irrefutable. You know. Because they're numbers.

Fortunately, I'm not going to make that argument. You can take the above percentages for what they're worth: flawed, cherry-picked numbers. The only thing you can really derive from these numbers is that Michigan had to gain more yards to score a point than seven other teams in the conference. As a reference measure for offensive efficiency, it's not good.

It doesn't tell the whole story. No statistic ever tells the whole story. It's just a fact. Some are just better at explaining a part of the story than others. We just have to be careful not to substitute the part for the whole.

FEI is often trumpeted as one of the best statistical tools for determining offensive efficiency. Yet, Michigan ranked second in FEI in 2010. Ahead of Oregon, Stanford, Wisconsin, Ohio State, Alabama, Oklahoma State, et al. That ranking is strange because Michigan turned the ball over 29 times and ranked 106 in turnovers nationally. Michigan also trailed the aforementioned schools in actual scoring. Can there be a legitimate argument that Michigan's offense was actually better than Oregon's, Stanford's, Ohio State's, Wisconsin's, etc.? Not if you saw the teams play.

And that is what this comes down to. Common sense. Your eyes. If your eyes are telling you that you're watching a turd of a football game, well... you are. If your reaction to the Wisconsin Michigan game was that Michigan just got completely curb stomped by Wisconsin in the first half, mounted a minor comeback when Wisconsin took a third quarter nap, and then still got blown out by 20 points at home, well... that's what you saw. Perhaps the stats tell a different story. Maybe. But while the stats say that Michigan ran up an astounding 442 yards against Wisconsin they don't relate what actually happened at the game.

Just to summarize Michigan's first half against Wisconsin, Michigan's first half drives against Wisconsin were as follows (plays/yards): 7/34 (punt); 11/59 (missed FG); 3/4 (punt); 3/3 (punt); and 2/24 (end of half [:30 sec]). That is an astounding 124 yards of offense and three punts. And Michigan was down 24 points at the half. The final score was 48-28.

Statistics at the end of the season and the end of a game can be just as misleading as they are informative. Where someone might see a close game based on statistics after the fact, anyone actually at the game saw a butt whupping that they haven't forgotten. I've made this case before, Michigan wasn't as good offensively as the end of the year totals, or end of game totals, seem to indicate. However, this is not to say the offense was awful. It was fun to watch and it provided a lot of nice memories and highlights. It put up some mind-bending numbers that rightfully deserve praise.

This brings me to my main point. The answer to the disparity between the numbers and your eye sight lies somewhere in between the two. But no one seems to be looking there. If you read the recent piece on the spread offense at MGoBlog, you probably understand why I'm writing this. There seems to be a battle raging in a segment of the Michigan fanbase between the reality of what has happened in the changeover and where that segment thought Michigan would be headed. Nothing that has been written over the summer of discontent sums that up better than this two sentence paragraph.

Michigan's turnover issues aren't fate, should improve naturally, and are not related to the spread. Most of Michigan's other issues at turning yards into points are not really the offense's.

Both of these statements are supported in MGo's post with numbers and statistics. They seem to be the dividing line between spread supporters and people who are all too happy to leave the spread offense behind. If you've watched football, and watched this team, these two statements seem to fly in the face of reality. Turnovers were a constant problem and appear directly related to the coaching staff because they never improved. And the second portion seems to be written in crayon. Of course the offense is responsible for scoring points.

But that's not the whole story, and it's not that easy to dismiss the above statements. In particular, the turnover statement is true as a general rule of football. The type of offense you run doesn't have a discernible impact on your turnovers. There are numbers to back this up if you study the averages of turnovers year to year of all the teams in D1.

On the subject of turnovers and the statistics that support them, the premise that they are "random" is a major pillar of the thesis that I, and a lot of reasonable people, do not share. Turnovers, in my view, are the by product of poor ball protection, poor scheming, poor decision making, and/or poor coaching. Maybe a batted ball or a helmet on the ball are an occassional random event, but sustained high turnover numbers are not random. Whether or not the high turnovers Michigan incurred in the spread were the result of the spread is up to the individual observer. I contend that the offense and the turnovers aren't related. I do contend, however, that turnovers are not a random event that can be simply brushed aside for purposes of arguing that the spread would've worked except for all the turnovers.

However, the contention that turning yards into points not being a responsibility of the offense is something I can not find support for. But I think this is something that is best explored in a minute.

The real issue at bay before you can assess the merits of the MGopost is whether the post is just a long apology for Rich Rodriguez, or whether it is a discussion of the merits of the spread offense in general. Given the tenor of the article it is easy to come to the conclusion that it is the former rather than the latter, and many people have suggested so in the comments. I'm not sure that's the intention.

The concern here is that Michigan is moving away from what should be an efficient, deadly offense in 2011. MGo is not making a direct argument that Rodriguez should've been retained, it is instead making an argument that his offense should've been retained because the statistics support it. To an extent, there is some support for this contention. You need not look further than the spread offense of Oregon or the spread and smoosh offense of Auburn and Tebow's Florida to see it's benefits. And, according to FEI, Michigan was better than Oregon offensively! Just so we're clear, numbers or not, my eyes won't letme believe that's true. But there are statistics to support that contention.

This brings us back to the offense's lack of fault for not scoring points. There are aspects of scoring points outside the 3-down offense's control. Field goal kicking is not a function of the spread offense, and if Michigan had a field goal kicker it's likely the season's offensive numbers would have been substantially better. However, the scoring issue here is not the spread offense. The issue that limited Michigan's scoring in 2010 was coaching.

It's easy to point to spread offenses like Oregon and Aurburn and say "we should be just like them."  But that's not the case. Not all spread offenses are created equal. Not all personnel are the same. There is a high degree of variance between the individual coaching staffs and their play calling within their different spread systems. I think it's over-simplification to point toward two successful spread offenses and suggest that we'll be just like them if we stay in the spread. Football just doesn't work that way, there are just too many variables.

This doesn't take away from the main point that, perhaps, it is a bad idea to abandon the spread in 2011. The answers is a little yes, and a little no. The majority of the Michigan fans' disgust with the spread is how it was implemented at Michigan. In conference, it was a disaster. Second half comebacks and big second half numbers be damned, Michigan's spread was not competitive against any defense with a pulse.

I think there is something to be said for the proposition that these failures were not a systemic problem, but a coaching problem. The spread did not make the offense more predictable or less efficient at scoring points. The coaching staff did with their limited play book, inability to implement a legitimate zone read, inability to take advantage of byes or additional preparation time, inability to find a legitimate running back, and insistence on using Denard Robinson until he was hauled off the field in obvious pain. This simply wasn't true at Michigan:

Rich Rodriguez kept ahead of the curve by constantly adding new wrinkles to the ground game. He was able to do this because of his vast experience with the spread 'n' shred.

So, this brings us to the middle ground I spoke of earlier.

It is impossible to deny the benefits and explosiveness of the spread offense. It works. It's documented-national-championship fact that it works. On the other hand it was only marginally successful at Michigan. This is not because the spread itself "won't work in the Big Ten." That type of argument is nonsense. No, it didn't work at Michigan because, frankly, the coaching staff screwed the pooch in implementing it. But now there is a new coaching staff that wants to run a more traditional offense and foresake the spread and all it's potential glory. So what happens next?

The answer is in the middle. If you watched the Spring Game, you probably noticeda number of Denard designed runs and even a little zone read. If you watched any of San Diego State's rout of Navy, you noticed a very diverse playbook with plenty of spread aspects. You've probably also noticed that Borges offense actually uses zone concepts in its blocking schemes, just not exclusively.

So we're somewhere in between jettisoning the spread into space and reverting to Mike DeBord's boring fullback-waggle offense. And that's not a bad thing. In fact, it's actually fairly progressive. Call it a Pro-Spread if you like. It's not going to be a full pro-offense, and it won't be a spread offense. It'll be somewhere in between.

There are not statistics to back this up. There are no numbers. I can't tell you that the bad things that happened last year will "regress to the mean" because some random study of a bunch of teams who all run different variants of a certain system indicates it will. Statistics do not, and can not tell us the full story. When we rely on them exclusively to make our point, we lose the credibility we believe we are building with them.

Football isn't a statistics game. While I admire people's attempts to quantify it, I have yet to find their efforts helpful or determinative of anyone's particular season. That's why we also have to use our senses when evaluating a football team. We can't just look at numbers on the field, we have to look at how they're playing, what they're doing right or wrong, their opposition down in down out, and observe what actually happens. Someone can break off a 40 yard run on a top ten defense, but a stat sheet won't tell you that the reason for the run was that the weakside linebacker tripped over his own two feet, thus leaving the gap open. Stats don't tell you if All-American player X was lallygagging for a quarter because he thought the game was over.

Stats help us to see things we miss, but they do not create an alternate reality. What our eyes see and what the stats tell us should not be that different. The reality is somewhere in between.