Sports Illustrated's Tom Verducci came out with his annual "Year After Effect" column yesterday, based on his hypothesis that that young pitchers tend to break down the season after an increased workload. Specifically, a pitcher 25 and under is supposed to be at risk if he pitched at least 30 more innings than his previous career high.
Dubbed the "Verducci Effect" by Will Carroll at Baseball Prospectus, it's one of the most prominent early examples of a happy marriage between analytics and old journalism. Sabermetricians were some of the loudest critics of the overuse that may have contributed to the early decline of Mark Prior, Ben Sheets, and other pitchers who debuted in the early aughts.
But the Verducci Effect probably doesn't exist. Its continued popularity has little to do with the power of numbers to support rational observation and everything to do with their power to baselessly reinforce existing beliefs. The article is an example of three pervasive mistakes that the general public makes about statistics:
• Regression to the mean: When an outcome is far above or below expectation, the subsequent results tend to be closer to the average. How does a young pitcher make Verducci's list? By having been healthy and successful enough to earn a greater workload. So by chance alone, you'd expect some members of that group to pitch worse, and you'd certainly expect to see some of them get hurt. Derek Holland's good health allowed him to pitch 71 more innings than he ever had before. If he gets sidelined in 2012, it will have more to do with the random nature of injuries than with Rangers mismanagement.
• Confirmation bias: People tend to rely on anecdotal examples that confirm what they already think. There's little attention paid to the pitchers who repeat their healthy seasons the year after; instead, fans fixate on the ones that get hurt. Mets fans might look at Jon Niese's 2009 injury and point to Verducci's warning label. They'd be ignoring the fact that no other pitcher that Verducci identified in '09 spent a single day on the disabled list.
• Correlation does not equal causation: Verducci is correct that some pitchers who pitch 30 innings more than their career highs tend to get hurt. However, that doesn't mean that his rationale for why that happens is valid. After I read his article, I went to the bathroom and peed. That doesn't mean I peed as a result of reading.
Every study I could find on the Verducci Effect suggests that it at best doesn't exist and at worst is backwards. David Gassko's 2006 study focused on the possibility of a decline in performance, and found an increase:
OK, so what happens if we limit ourselves to pitchers who threw at least 100 innings in year two? Actually, a funny thing. The pitchers who best their career high by at least 30 innings go on to throw 90% more innings in year three than they do in year one, and those who didn't only throw 78% as many innings. What's more, while the [year-after effect] candidates have an ERA 9% lower in year three than it was in year one, the guys who were accustomed to the big workload do not improve their performance at all.
Jeremy Greenhouse's 2010 follow-up focused on injuries and also found nothing. JC Bradbury came up empty. Brian Burke used a card game to show how randomness, not overuse, is the likely culprit. Tom Tango expressed his concerns (there's elaboration in the comments.) Scoresheetwiz found nothing too.
Verducci's find is the Sports Illustrated Cover Jinx in a new dress. Athletes grace the SI cover because they've done something extraordinary; it's reasonable to expect that they would return to being average by the time the issue is in your recycling bin. Yet the idea of the jinx endures, because people notice when an athlete sucks after being on the cover and ignore the players who continue to do well.
It's time for Verducci to retire the Verducci Effect until someone comes up with evidence that an increase in workload has anything to do with a decrease in productivity in the following year. It's not even a rule of thumb or a kludge at this point, as the pitchers he picked out are probably more likely to succeed in 2012, not less.
I don't blame Verducci much for continuing to write the articles. He deserves a lot of credit for coming up with an interesting hypothesis—it's not his job to use advanced analytics to prove or disprove it. A lot of the fault lies with Baseball Prospectus's and Carroll's stubbornness on this particular point. He now insists that it's just a general warning about overuse and that "people who quantify it strictly are missing the point." But given the rule's origins as a hard-and-fast limit, he's just asking for writers without statistical backgrounds to misinterpret it. Here's the BP glossary entry for "Verducci Effect:"
Named for Tom Verducci of Sports Illustrated, this is a negative forward indicator for pitcher workload. Verducci, who called this the 'Year After Effect,' found that pitchers under the age of 25 who have 30-inning increases year over year tend to underperform. Will Carroll independently found that pitchers who break the "Rule of 30" tend to get injured. Carroll renamed this 'rule' the Verducci Effect in honor of the man who initially found the evidence.
It's time for a new definition.