Don't Blame Tony Romo: How Our Love of Football Statistics Ruins Our Perception

The rise of statistical analytics as a means of appreciating sports in modern America is a puzzling conundrum. We are currently a nation that is regularly ranked among the least proficient in mathematical ability and yet we are among the most dependent on mathematics, statistics and data management to navigate the labyrinth that is contemporary life.

And the sporting life—that of drills and laps, conditioning and weight training, grunting and physical punishment—has historically been viewed as distinguishable, if often entirely antithetical to, the life of the mind.

To be sure, there are exceptions. The ancient Greeks and Romans, early Israeli settlers in the twentieth century, the R.A. Dickey’s of professional athletics: they represent cultures, civilizations or individuals who believe(d) that the body was intrinsically tied to the mind, that being cultured was not mutually exclusive from being physically gifted.

But the conception of modern sports as mathematically describable—and predictable—is new. The popularity of sports-math is similarly new and, perhaps more importantly, the result of the unforeseeable rise of “viewer-ownership” in fantasy sports, SABRmetrics and the popularity of the NFL in particular.  As our love of sports has become democratized, so the ways in which we view sports have changed.

We are a nation of avid sports fans. We watch with an emotional and financial dedication that nearly baffles the mind. Though we are in the midst of a "Great Recession"—and what for many has been a 48-month-long financial misery—in the last four years, we have found ways to spend billions of dollars on tickets, jerseys, memorabilia, stadiums and other sports-related products.

Our love of sports seems to outweigh our hope for financial stability. Cities like Philadelphia, Chicago and Boston—among a host of others—live and die by the success of their Eagles, their Bears and their Patriots. Economics be damned; if our team wins the Super Bowl, it is a salvo for all other ills.

But we are also a nation of people who are fairly impatient, who like events boiled down into easily digestible chunks, who would rather get the gist of an unfurling narrative than engage with the narrative itself; and statistics help us facilitate these impulses: numbers provide shortcuts for comprehension.

This is not unique to professional sports. We of the public who know little of economics rely on simple Gaussian curves and employment rates to describe the state of a nation’s fiscal health. We’re perfectly comfortable assessing the merits of a financial regulations, even without knowing much of accounting or finance or international trade or defaults or swaps or otherwise, through the use of simple percentages and correlations with past trends.  

But the problem remains. We are a nation that is bad at math. And the more we rely on statistics to determine the value of a plan of action—or in sports, of a player—the more we betray a fundamental flaw in our reasoning.

In the NFL, the best examples of this are reference to amalgamated data without reference to context—interceptions, sacks, incompletions—and the use of derived values that almost no one knows the content or background of—namely, quarterback and passer ratings. We tend to use statistics that paint a broad swath in order to describe specific outcomes, and though our statistics might be useful to represent some general trend, they’re not particularly nimble nor do they account for inconsistency or statistical deviation. 

And no player or event has exemplified the problem better than Tony Romo of the Dallas Cowboys.

Depending on who you talk to, in what city, at what time of year, assessments of Tony Romo’s quarterback play vacillate from his being an abysmal underachiever to a solid quarterback with all the tools for success. Given that he’s also labeled as inconsistent, perhaps those two divergent labels make sense.

Quarterbacks bear the load of responsibility for a football team’s success because they are team leaders. They are the drivers of the offense, the movers of the chains. But this is a false assumption. A quarterback can’t throw a pass to himself and in only very limited places can a quarterback run a play alone. A quarterback must rely on and have faith in his receivers, his backs, his offensive line.

In the Dallas-Chicago game on Monday night, Romo could do none of that. His best receivers looked like a squad of ham-fisted sprinters. And his running backs were a realization of total ineffectiveness. Want to blame Romo for taking risks in the second half? What else is a guy whose team is down 17 points to do when none of his primary weapons are doing the most basic things they’re supposed to be doing?

Tony Romo’s statistics lines for that game were fairly unforgiving. On the upside, he wasn’t sacked and he did throw a touchdown. On the downside, he threw five interceptions and had a passer rating of 60.1, well below the statistical average of around 67. Then again, he did complete nearly 75 percent of his passes for over 300 yards. And yet, his Total QB Rating, according to ESPN, was a 14.2. On a 100-point scale, that is about as poor as a person’s ratings can be.

Something is evidently strange here. How does a guy who threw for the length of three football fields and found success on three out of four passing plays still come up so short? And why did the Cowboys lose that game so badly?

Well, let’s start with the starters. On two occasions, Cowboys receivers left points on the field by not catching passes that were placed perfectly for them. The Cowboys run-game didn’t get moving under almost any circumstances. And in the latter stages of the game, the Cowboys offense was forced into a position to pass almost exclusively—they were down by 17 and couldn’t find their way into the red zone.

This is typical of football games; if not for one or two plays, the game’s narrative would be functionally different. If the Cowboys won on Monday—which is to say, if Dez Bryant and Miles Austin caught passes when they were supposed to—we very well might have been talking about how ineffective Jay Cutler is and how the Chicago Bears are the most up-and-down team in the league.

Even when a game is won or lost by three scores, usually, the differences are often fairly minimal. Which is to say that football scores and statistical analyses maximize the differences to generate what is, in some ways, a clearer distinction between success and failure.

If you watched the Cardinals-Rams game on Thursday night, you likely saw something similar. Rams quarterback Sam Bradford completed literally 33 percent (7-21) of his passes for 119 yards. And yet, his team won by 14, the difference being the two touchdown passes Bradford threw.

Startlingly, of the 119 yards Bradford passed for in nearly 30 minutes of on-field playing time, 95 came on two receptions; which is to say that Bradford literally threw the ball 19 times for a total of 40 yards in all other cases.

Now, my sense is that everyone can ascertain immediately that between Romo, Bradford and Kevin Kolb (Arizona’s fearful offensive leader). Romo had the best night by nearly any measure except the one we tend to value the most: the win at the end of the game.

But where Romo received almost all of the criticism for the loss—and had disappointment heaped on him even by other Dallas veterans—critical parity would be welcome. Of Romo’s five interceptions from the Monday night game, one was a fumble, one was caused by a receiver misrunning a route and one came so late in the game it hardly mattered.  

Romo was criticized for poor leadership and an absence of productive decision making. But the mark of a leader isn’t simply going from success to success. It’s also responding to failure and attempting to facilitate the progress of those around you. So, when Dez Bryant let a surefire touchdown pass go straight through his hands, on the very next play, Romo threw a first-down pass that hit Bryant in the chest.

I remember something similar occurring with Donovan McNabb during his tenure with the Redskins. The only statistic anyone remembers from that season of his? That he threw more interceptions than touchdowns, 15 to 14. And yet, if one reviews the season’s game tape—as painful as that might be—one sees that of the 15 interceptions thrown, more than 20 percent were not McNabb’s fault.

One pass literally bounced off of Santana Moss’s facemask before landing in the hands of the opposing safety. And yet, it was enough to see McNabb benched before the end of the season and traded by the start of the next.

Oddly, where mathematical analysis is almost always thought of as a tool for clarification, what it actually does is provide us with shortcut for avoiding the complication of accounting for disparate, often contradictory variables and outcomes. And because we’re not inclined to necessarily understand the processes behind statistical analysis, we often, in short, don’t really know what we’re talking about.

After all, in the NFC East, the defining quality of almost every major quarterback that has played for one of the division’s four teams has been their undeniable inability to consistently perform under almost any circumstances, whether good or bad.

Donovan McNabb was able to hit 40-yard bombs more readily than an eight-yard slant. Eli Manning has a career 85 passer rating and yet has won two Super Bowl rings. Romo, statistically, is better than both of them, more accurate and perhaps in some ways more desirable, but he’s only won one playoff game in his career.

What seems to separate them—and others like Michael Vick or Rex Grossman—is what we can begin to describe as "contextual analytics." We tend to engage in wholesale amalgamation when using statistical methods to analyze a player’s performance or value. The problem is that long-term and large-scale data plots aren’t particularly nimble or flexible.

Eli Manning throws approximately three touchdowns for every two interceptions across his career. But, interestingly, he rarely throws interceptions in the fourth quarter of games, during which he has a nearly impeccable record as an end-zone-driving savant. And in point of fact, this is the reason that Manning—who began his career in the same time frame as Romo and Grossman—has likely brought two more Super Bowl rings to New York City.

If we were able to construct a system of football analytics that allows us to value contextualized data, then we would be better able to understand why some quarterbacks—like Manning—rise to success and others—like Romo or McNabb—fall. And it is worth noting, say, for the fourth-quarter statistics, Tony Romo’s apparent inconsistent quarterback performance is, in fact, very consistent: for the first three quarters of any game in which he plays, he is a very reliable quarterback. But once the fourth quarter rolls around, his stat lines tumble.

Statistics matter, but our ability to determine how they matter—and as importantly, how they’re derived—matters just as much. We tend to overvalue career or season-long statistics more than statistics that accurately describe specific situations. After all, we use math in sports as a kind of heuristic that allows us to boil down long-term trends.

What the next generation of sports enthusiasts need is a system of analytics that more accurately and more precisely describes what sports fans see in games.