clock menu more-arrow no yes mobile

Filed under:

2008 Stats Relevance Watch: A Brief Introduction

The last two years, I've preceded all of my "Stats Relevance Watch" posts (see sidebar left) with the cliché, "There are, as they say, lies, damn lies and statistics," a nod to the inherent malleability of any set of numbers to prove whatever it is you want them to prove. They can, however, serve as a useful shorthand - in lieu of watching every team play every game, which is impossible, the box score is the window to the "truth," the economical synthesis of the sum result of 22 bodies gliding and colliding in coordination some 150 times over a few hours, give or take, to one specific end. For any understanding of the game beyond the scoreboard, to assess the story of the game without the benefit of watching it unfold - and especially to form some predictive idea of each team's strengths and weaknesses and overall performance in nuts-and-bolts terms, the every-down blocking and tackling - that synthesis is essential to making sensible, informed conclusions that tell the truth about the elemental push-and-pull that the final score or overall record occasionally (only occasionally) does not.

So to provide the level of analysis I try to provide on SMQ, numbers are everything, but only so far as it remains perfectly clear that that specific end - the final score, winning or losing - is the ultimate trump, the only number that really matters. The box score is worthless if used in some futile attempt to "overrule" the final score. A win, in the end, is a win (is a win is a win), and a loss is a loss. Statistics only inform the result; they never define it. In discussing numbers, the first, most essential aspect of the enterprise is that they don't "lie" - that is, that the numbers in question actually have some bearing on the final result, and aren't tossed around for the sake of tossing.

The first caveat of any statistical analysis, then, is that the numbers are only good suggestions, not sacrosanct truths, and the first test of its relevance is whether the numbers being used have some direct relationship to the ultimate number, the final score. This is the goal of the Stats Relevance Watch, which sits at the foundation of so much of what I write for the rest of the year: not only to demonstrate that statistics are typically an accurate reflection of the "truth" about any particular team or performance by demonstrating an overwhelming, quantifiable correlation between stats and victory, but also to identify specifically which stat categories are most prominent in that correlation and which categories are less useful. Numbers that don't match up with wins and losses should be ignored.

If past experience is any indication, what we'll see over the next couple weeks is that the smart money is always on the box score.