“Numbers Never* Lie,” but ESPN Does

I’m what you could call a stat head. I’ve spent far too many hours staring at data distributions, constructing confidence intervals, and inspecting survey questions for potential bias. Some sickness makes me enjoy these things so much that I chose to endure unintelligible statistics classes with Algerian- and Egyptian-accented professors throughout college. You might say I’m a young Nate Silver. You’d be wrong, but you might say it.

My obsession extends into my life as a sports fan. I have just as much fun looking at WHIPs and GSCs from the comfort of my bedroom as I do on summer days at the ballpark. I have a whole set of theories about why the metrics used to evaluate tennis players are fundamentally flawed, and I could spend hours boring you with the details of these theories. The point is, when I first heard about ESPN’s “Numbers Never* Lie” in advance of its 2011 debut, I cried out with unbridled nerd joy. ESPN’s official description of the program says that it “[uses] advanced statistics and analytics to settle the biggest debates in the world of sports.” Sweet Jehovah! Advanced statistics! Analytics! Biggest debates! The world of sports! My heart goes aflutter at the mere mention of those things. I just had to behold this wondrous television show with my own eyes. The next day, I did.

It turns out the description of the show should have read like this: two ESPN employees who don’t know what a standard deviation is debate a sports topic for a minute or two. When they’re through with the topic, they pull one (1) arbitrary statistic out of an envelope, and use this number to decide which debater was right. Never mind explaining how the statistic was calculated, or considering even one of the slew of other figures that may influence the issue at hand. Nope. The one number taken off Google by a broadcast journalism major is the key to unlocking The Truth of sports. Next topic!

Okay, maybe I was dumb to expect rigorous statistical analysis from a show on ESPN’S weekday afternoon lineup. After all, this is a block of programming in which Skip Bayless is given the opportunity to express his opinions to other people, so my optimism was surely a bit misplaced. But such a show isn’t completely unheard of. Over on the more niche-friendly MLB Network, Brian Kenny and company have a good ol’ sabermetricious time on “Clubhouse Confidential,” a show that I, as a subscriber to the 100-channel cable package instead of the 900-channel one, unfortunately miss out on most of the time. But ESPN shouldn’t throw out words like “advanced statistics” and “analytics” in reference to their mess of a show.
After it flunked the ten-minute tryout I gave it when it debuted, “Numbers Never* Lie” had been off my radar for a long time. On Friday, I decided to give it another chance. Perhaps Bill James or J.C. Bradbury had visited current hosts Jemele Hill and Michael Smith in their dreams, and imbued them with the gift of statistics.

Alas, Hill and Smith’s analysis fell straight into traps of conventional wisdom that statisticians have easily debunked. When discussing the Miami Heat’s 24-point blowout loss to the Spurs, the hosts contended that the “numbers” were on Miami’s side. You see, after their last two losses of twenty-plus points, the Heat went on to win eight out of their next nine, and 11 out of their next 13, respectively. And it was definitely getting blown out that spurred these runs! Strength of schedule or increased offensive efficiency couldn’t possibly have been factors, so why bother looking into them?

Things got even crazier when Hill and Smith talked about the Houston Texans and their first-overall draft pick. Hill argued that the team shouldn’t select a signal caller in a draft class full of weak or risky quarterbacks, while Smith disagreed, preposterously claiming that the team made a mistake drafting Mario Williams first overall in 2006. Then came the number: Houston should draft a quarterback, because last season, their quarterbacks increased their win probability by 46 percent.

What I interpreted this to mean (with no help from the hosts) is that, over the course of 16 games, the quarterbacks increased win probability by an aggregate of 46 percent, or a little under three percent in each game. But is the increase relative to the average NFL quarterback’s performance, or a tackling dummy’s performance? I’d have to assume the latter, but it doesn’t matter, because the stat itself says nothing about what the Texans’ draft strategy should be. What effect does the typical first-overall rookie quarterback have on win probability? How did the top free agent quarterbacks fare in this metric during recent seasons? Answering these questions may or may not uncover any pearls of statistically significant wisdom, but asking them, and other questions like them, would at least begin the process of substantive analytical thinking. But “Numbers Never* Lie” doesn’t have time for such silliness.

The show failed its first ten-minute audition back in 2011, and the 2014 version lasted just three minutes longer before it was dishonorably discharged in favor of a much more enlightening program: that “Friends” rerun where Ross dresses up as an armadillo. I’ll keep my eyes open for another show that can satisfy my statistical cravings, but I’m 99 percent confident that ESPN will never have anything for me on its menu.