Interpreting the Pirates’ Spring Training results

Each year, the MIT Sloan Sports Analytics conference produces a mother lode of interesting research, some of which involves math that ranges vastly beyond the grasp of my limited intellect. On the eve of Opening Day, however, I’d like to discuss a particularly timely article from this year’s conference, one that provides grounds for several fun debates heading into the season.

Prepared by The Economist’s Dan Rosenheck, the presentation in question takes issue with the modern sabermetric consensus that dismisses Spring Training statistics as irrelevant. We’re all familiar with the justly-mocked ‘best shape of his life’ narrative that, if accurate, would see roughly 98 percent of the players in the league elevate their respective games to become perennial All-Stars each year. But the ridiculousness of that narrative, Rosenheck (persuasively, in my opinion) argues, does not mean that Spring Training statistics are useless in predicting a player’s performance for the upcoming season.

Rather, Rosenheck contends we’ve simply been looking in the wrong places—assertions that Spring Training stats are useless often focus on measures like batting average and ERA, which tend to be subject to large amounts of random fluctuation and have limited predictive power even with full-year sample sizes. If, instead, we use statistics that stabilize more quickly (walk rate, strikeout rate, isolated power, etc), we can gain insight into players' true talent levels. Put another way, Spring Training peripherals matter. Rosenheck tests this assertion by weighting Spring Training statistics, incorporating them into Dan Szymborski’s ZiPS projections, and finding that the addition slightly improves ZiPS’ accuracy.

You can find a more thorough synopsis of the presentation here and the original presentation slides here. The math isn’t particularly gory, fortunately. If you’re interested in more detail on ZiPS projections and what they mean, check out this article.

It’s impossible not to be a little excited about this—there’s an exuberant fan inside all of us, and with the return of baseball after the long winter that fan really hungers for performance to dissect, even before the games start to matter. Lending any sort of empirical validity to string training statistics lets us indulge a bit and do just that. Regard, by way of illustration, the table below, constructed by applying some of the methods Rosenheck outlined to generate revised season predictions for each Pirates hitter with 30+ Spring Training plate appearances:

Cool, right? Anybody else pumped about Francisco ‘Mini-Martin’ Cervelli?

I should preface any involved discussion of the table above with two disclaimers about the methodology I used to generate the revised numbers. I didn’t have access to accurate park factors for Spring Training stadiums, so inputs were not park-adjusted. That's no small thing, given McKechnie Field's reputation as a hitter's park. I also used more aggressive forecasting weights than those used in Rosenheck's article. For these reasons, I’m a bit skeptical of the magnitude of some of the effects we see.

Still, there’s plenty of fodder for debate here. Specifically, I want to take a look at three key questions, and see what the community thinks:

Will the real Andrew Lambo please stand up?

The #FreeLambo movement, now multiple years old, isn’t without merit. Lambo, despite posting back-to-back AAA lines of .272/.344/.589 and .328/.389/.563, hasn’t received an extended major league opportunity. Blocked at various points in the past by Travis Snider, Jose Tabata, and Gregory Polanco, it appears that Snider’s trade and an injury to Jaff Decker has now opened the door for Lambo to stake claim to a bench job.

The question is whether he’ll earn it. As a corner outfielder/itinerant first baseman with limited base running value, the burden is on Lambo’s lefty bat to drive virtually all of his value. Both Steamer and ZiPS (in addition to ZiPS’ .242/.297/.415 projection, Steamer projects a .248/.298/.431 line) see Lambo as a league-average hitter, which would give him reasonable value as a bench bat and facilitate a nifty Snider impersonation.

Lambo, though, has tanked his second straight Spring Training—the revised model drops his projection to just .235/.286/.397. If you believe in Lambo’s minor league performance and see him as a league-average hitter, he still probably deserves the final bench spot. But when Decker returns from injury, the Pirates will have an interesting decision to make. Decker’s revised projection is .219/.301/.344, and he’s (by reputation, at least) a stronger defensive outfielder than Lambo. I wouldn’t be surprised if we see Decker in Pittsburgh sooner rather than later, especially if Lambo falters in the early going.

So who is the real Andrew Lambo? Who do you pick between Lambo and Decker?

Can Francisco Cervelli replace Russell Martin?

Nobody had a better Spring Training than the Pirates’ catchers. Both Cervelli and Tony Sanchez improved their respective projections by a significantly larger magnitude than anyone else on the club, with Cervelli gaining nearly 100 points of projected OPS. Cervelli’s always been an intriguing player in that he’s now amassed a career .278/.348/.381 line over 785 major league plate appearances despite the lack of a lengthy track record of minor league success. He hit .301/.370/.432 last year, and the revised projections (much more so than raw ZiPS or Steamer) think that’s reasonably close to his true talent level, forecasting a .260/.353/.422 line. Defense has never been the question with Cervelli, and if he hits the way he’s capable of hitting, the Pirates will replace most of Russell Martin’s production at a fraction of his cost.

Of course, we can’t discuss Cervelli without mentioning his health—there’s a reason ZiPS only forecast him for 211 plate appearances. Enter Tony Sanchez, who currently appears to be the beneficiary of Chris Stewart’s unfortunate injury. Sanchez appeared, as recently as last September, to be falling out of favor with the organization, taking grounders at first base while Elias Diaz started at catcher. After a brutal winter ball showing, prospects for his resurgence looked dim indeed. Sanchez, however, has regained his footing with a torrid spring—this combined with his newfound throwing accuracy, may place him in line for extended time should a Cervelli injury occur.

Can Cervelli replace Martin? Are you comfortable with Sanchez as the starter, or would you start Stewart in the event of a Cervelli injury?

How will Jung-Ho Kang play (and how long until someone nicknames him ‘King Kang’)?

Kang is, with good reason, probably the most intriguing player on the Pirates’ roster. He has his fair share of skeptics, but also some prominent admirers. His Spring Training served as a microcosm of what we might reasonably expect from him this year—good power and tons of strikeouts. It didn’t, however, do much to improve our ability to forecast his performance—Kang’s preseason ZiPS projection is virtually unchanged in the revised model. If he can approach a .700-ish OPS while playing adequate defense on the left side of the infield, he’ll be an asset—think mini-Pedro Alvarez if Alvarez were also a competent defensive shortstop.

Unfortunately, Kang may have to play well if Sean Rodriguez’s struggles don’t resolve themselves. Rodriguez’s main offensive skill (raw power) wasn’t on display this spring, and as an extreme fly ball hitter with decaying plate discipline (see graph below), Rodriguez could be in danger of posting another .255 OBP in 2015.

Rodriguez is chasing more often than earlier in his career, and combined with his increased power he’s seeing fewer pitches in the zone. Given how closely Kang duplicates his core abilities, Rodriguez will need to either be more selective or retain all of his 2014 power to continue to provide value.

Will Kang take the league by storm? Can Rodriguez rebound? Over/under on Kang 2015 home runs?

I’ll go with 15 for that last question. Life’s too short to be a pessimist. What do you think?