The Role of Statistics and "Advanced Metrics"

In case you missed it, on Monday a Dallas sportswriter went off on baseball math gone bad. He chose to go off on pitch counts and claim that math's turning baseball fans into worrywarts as soon as a pitcher hits a high-enough number. And he threw in an odd shot about the shock that something not 100% certain to happen didn't happen (Nelson Cruz making a catch. I know, really odd).

He's already been taken to task by 670 the Score's Tim Baffoe (which is really a must-read that just excoriates the daylights out of him), and the sportswriter and Baffoe had an email exchange to kind of clear the air (also a great read, though much lower on the vitriol meter than the other two entries).

Now, Mac Engel's original point was that what all these counting numbers and stats are trying to accomplish is the ability to play games merely on a spreadsheet or computer and that the reason we love our sports is the "unpredictability" (he says "Math is never wrong. Baseball very much is, which is why I love it." Along with using the actual word "unpredictable.").

Here's the thing - statistics aren't trying to do that. Not in baseball, not in hockey with Fenwick numbers and Corsi, not with basketball with PER, and certainly not in football, not even with Pro Football Focus ratings and metrics such as DVOA. Liking the unpredictability of sports and embracing numbers in sports are not mutually exclusive things.

Let's just briefly go over what statistics are meant to accomplish:

1) Statistics are a means to quantify or attempt to clarify a player's impact or show what is happening.

2) Statistics are quantified or calculated to attempt to find correlations between on-field capabilities and victories.

3) Statistics count things that have already happened.

Maybe 3 goes before 2, but you get the point - by their nature, statistics don't get into the business of taking the play out of the game since the play is required to get the statistics. That doesn't even get into the part that some advanced metrics can be opinionated by what is or isn't included in the calculations or how they're calculated, or some borderline base statistics changing with even a referee missing a spot of the ball.

Look at advanced metrics this way. In baseball, batting average used to be the biggest go-to statistic for a long time. But batting average hardly tells you anything - merely how often the batter gets a hit, regardless of the type of hit. If someone had a high batting average but just wasn't having the same impact you thought he'd have with a .300 batting average, you'd ask someone why that was the case, right? Then they'd tell you he was hitting a lot of singles or not having anybody on base in front of him. It's information - it'd be the same if you looked and saw the guys ahead of him having sub-.300 on base percentages or the batter having a slugging percentage of .330.

Football is the same way with yardage. Total yardage is nice, but with any gross number, it has its problems - quantity of plays, time, attempts, all that sort of thing. So, yards-per-play, net-yards-per-pass-attempt - it's the same concept. Quarterback rating's been around for seemingly forever. Offensive lineman play has been unidentifiable numerically for years, so Pro Football Focus' ratings are pretty much the go-to source for judging offensive line play, judged ratings (by people watching the games, mind you) though they may be. And don't forget ESPN's QB(S)R!

Granted, baseball is much farther along in advanced metrics than football is, but the point remains the same - advanced metrics are developed to attempt to provide extra information. And it's all done using past data, to further attempt to clarify past data.

Could it be used for future performance? Well, trend analysis already does this work, so that's already been out of Pandora's Box forever, and any analyst is more than willing to toss out a two-cent prediction based on what their expert opinion is. But just to make a point, ZIPS projections and Bill James projections have been done in baseball for years, and while they're occasionally accurate in a range, that's nowhere near enough to turn the sport into a computer simulation with concession stands. And then there's the matter of how they (the players) get those projected stats and records.

And even taking that all into account, even the best prediction of what will happen in a future season is just a (sometimes educated) guess. You can't quantify what hasn't happened yet.

But that's not to say advanced metrics don't have a place in building the future of a team. The key for front offices is identifying which statistics and metrics help contribute to winning baseball games. Any edge gained here isn't enough to trump a superstar talent or anything like that (i.e. you don't want to have a team full of nothing but guys with high OBP or will collectively not have an error, or in football never drop a pass or anything like that), but where a team will get that edge is in the bottom of the roster, where the minute differences really tend to matter more as you get into a lesser talent differential. Wins-Above-Replacement is probably the most comprehensive metric in sports (attempting to measure a player's wins added above a Quad-A-level/fringe major leaguer), but as nice as that is for a quick look of how a player's performed, ultimately it's not particularly useful in that regard.

(The point I'm trying to make here is you don't want to replace Brandon Marshall with Earl Bennett because Bennett drops fewer passes. Marshall starts because he is a damn excellent receiver despite the increase in drops.)

So, to bring this to a point. If you're looking at numbers and statistics through the lens of "They're gonna the be-all-end-all and they're tryin' ta' take mah sports away," please stop. Those stats and metrics might just be working to get your favorite team a title.