The Stat-Head Revolution

Geeks Infiltrate Baseball's Front Offices; Conventional Wisdom Flees

It's an annual rite of spring: at bookstores across the land, seamheads and roto geeks flock to the new edition of Baseball Prospectus, the thick annual that projects player performance for the upcoming season. This year's is notable not just for the usual cutting-edge statistics leavened with snarky wit (Erubiel Durazo, the hit machine stuck behind Mark Grace on the D-Backs' depth chart, becomes "a prisoner of conscience, held by sadistic autocrats in a desert kingdom"), but because it marks the swan song of co-author Keith Law, who departed in January. His new job title: assistant to the general manager of the Toronto Blue Jays.

For a stat-head to breach the insular world of Major League Baseball is a near-unprecedented leap, and it's tempting to write it off as an anomalous blip, like orange baseballs or Bill Veeck. But many view the signing of Law as a harbinger. "I think what we're seeing is the beginning of something much bigger," predicts columnist Rob Neyer. "In five to 10 years at the most, half of the GMs in baseball will have this sort of background."

If so, the effects could go far beyond boosting BP's Amazon sales ranking in Florida and Arizona. The implications of a "sabermetric" revolution could shake the foundations of baseball common wisdom and convert also-rans into contenders overnight. In fact, it's already begun.

Statistical analysis and baseball go way back, of course, at least since George Stallings of the 1914 Miracle Braves introduced the first full-scale righty-lefty platoons. But the modern surge in sabermetrics—a coinage from the acronym of the Society for American Baseball Research—can be traced back directly to Bill James, a Kansas baseball junkie whose self-published Baseball Abstracts in the '70s brought sophisticated mathematical tools to the masses for the first time.

Among James's readers were the then college students who went on to found Baseball Prospectus. Along with denizens of and other Internet forums, they've expanded on James's work, bringing modern computing power to the task. If the result sometimes looks like a thicket of acronyms—just try to remember the difference between ARP (Adjusted Runs Prevented) and WARP (Wins Above Replacement Position)—the point is to make stats more explicable, not less, by reducing performance to a set of easily quantifiable "metrics."

The new metric that's caught on fastest in the larger baseball world is undoubtedly OPS (On-Base Percentage plus Slugging Percentage), which has the virtue of being both calculable by any grade-schooler and an excellent predictor of runs scored. Whereas the traditional trinity of batting average, homers, and RBI counts doubles the same as singles, and walks the same as never leaving the bench, OBP and SLG credit hitters for doing the two things they're meant to do: getting on base, and moving runners along.

"The notion that OBP measures something more important than batting average is the single biggest change since we started Baseball Prospectus," says Joe Sheehan, the group's managing editor. "When I got to college in the early '90s, RBIs were still really important—Joe Carter was still an All-Star. We've really moved away from that." Neyer is more cautious, noting that while lip service to OBP is high ("It's hard to deny the idea that you can't score runs if there's nobody on base"), that hasn't stopped managers from urging "aggressiveness" at the plate, even at the expense of trading walks for weak groundouts.

But if slugging average, OBP, and OPS are on the verge of household statdom (several teams now even display them on their scoreboards), it'll be a while before they displace the "batting average with two outs on turf at night" junk stats that permeate modern sports coverage. "Clutch hitting" stats are especially loathsome to sabermetricians, because decades of research have turned up no evidence that clutch hitting exists. Likewise, bunting, that age-old shibboleth of baseball fundamentals, is considered worthless by most sabermetricians—when Bob Brenly ordered Craig Counsell to bunt four times in one World Series game last year, it prompted a furious BP editor e-mail exchange under the heading "Bob Brenly is an idiot."

"Hit-and-runs and bunts worked great back in the dead-ball era, when you had to generate runs however you could," notes Voros McCracken of "Ninety years later, when people are hitting home runs left and right, these strategies really don't make any sense."

Then there's McCracken's own metric, DIPS (Defense Independent Pitching Statistics), which exists way out on the sabermetric frontier, with even many statheads turning a cold eye. The DIPS theory, borne out by the numbers, is that for balls in play—i.e., everything except walks, strikeouts, and homers—it's mostly chance and defense, not pitching ability, that determine if they land as hits or in fielders' gloves. "They teach pitchers that the idea is not to strike people out, but to try to get ground-ball outs," says McCracken, but, he notes, even ground-ball guru Greg Maddux garners only an extra five or 10 outs a year on balls in play.

When does McCracken expect to see DIPS join the baseball stat pantheon? "It's a pretty radical idea," he admits. "Probably never."

Next Page »