The Limits of Projections: Nelson Cruz
In the post comparing a variety of projection systems and what they make of the M’s, I noted that the entire AL is clumped around 81 wins (see Jeff Sullivan’s fangraphs post on this today), and as a result, *individual* projections mean a lot more to the M’s playoff chances. A career year from Kyle Seager or Ketel Marte can impact the pennant race in a way it wouldn’t if the M’s were chasing some 95-win behemoth. Of course, individual forecasts are really, really hard to do, and the M’s have a number of players for whom traditional projection methods might be ill-suited. The point of all this isn’t to say that these cases are where the projection systems are wrong, or this is why they’re dumb. Having an objective starting point, with the same methodology applied to each player is incredibly valuable, in addition to making February almost bearable. Besides, it can be instructive to think through what we mean when (if?) we say that the M’s will “beat” their projected win total. As fans, we actually DO have access to all sorts of material that PECOTA, Steamer, ZiPS or whatever don’t. But as fans, we’re often ill-equipped to make sense of that extra information, or to apply it even-handedly. It’s not clear that we *should*, of course. In any event, we’ll start this off with one of the hardest players to make sense of in the American League; a player whose career arc makes less and less sense the more you look at it: DH (hopefully) Nelson Cruz.
In general, projection systems are going to have a lot more disagreements about players with short track records. The younger the player, the higher the fraction of minor league PAs to MLB PAs, etc., the more projections need to focus on things other than looking at past performance and applying an aging factor to it. For players with plenty of MLB time, this is much less of a concern. You don’t need to look at major league equivalencies or spend computing power weighing the strength of the Texas League versus the Southern League in 2014 – you can just take the prior X years and take a weighted average. But what are they – or we – to do with Nelson Cruz? Four years ago, at the age of 31, Nelson Cruz was terrible, a more or less league average bat with a decidedly less-than-league-average glove. He had some power, as he’d shown in 2010, but that power seemed to be on the wane: from 2010 to 2012, his ISO fell from .258 to .200… in Arlington. His fly ball percentage fell as well, and he made less contact than he did earlier in his career. The raw stats were bad, and if you looked at the components, you’d see what looked like characteristic signs of declining bat speed and production.
From 2013 on, though, Cruz has thumbed his nose at aging curves and component stats and enjoyed some of his best years as a pro. He abruptly halted his decline in ISO and started racking up HRs, setting a career high (by a mile) in 2014 before besting it the following year. Even in the midst of this, projection systems and analysts saw warning signs. After his 2014, Tony Blengino penned this piece at Fangraphs saying that Safeco would sap the production out of the fly balls Cruz hit, and that aging would hurt his batted ball authority. Blengino also noted that Cruz was pulling more of his balls in play, leaving him vulnerable to infield shifts that would hurt his production on *non* fly balls. It all makes a certain kind of sense, and when you remember that Cruz turned 35 last year – he’s older than the cratering Josh Hamilton – Blengino’s prediction that Cruz would never hit 30 HRs again doesn’t seem so harsh. But trends and component stats aren’t destiny, as Cruz’s 2015 shows us. His pull tendency stopped and went into reverse, and Cruz’s fly ball percentage *dropped* markedly. In its place was a huge uptick in Safeco-defying batted ball authority, especially fly ball authority. This is precisely the skill that Safeco was supposed to depress, but Cruz got more out of his fly balls in Safeco than he ever did in Baltimore of Texas. His HR/FB rate had touched 21% a few times, but it was over 30% last year. He hit fewer flies, and was spectacularly unlucky on those that stayed in the park, but it didn’t matter: 2015 was Cruz’s best season at the plate by far. Cruz has not just made adjustments, he seems like an entirely different hitter. At the very least, he’s made a series of adjustments, from pulling more pitches to last year’s radically different batted ball profile. Whatever he’s doing, it’s worked.
What do we do with this? Nelson Cruz’s authority is increasing, and his move to a pitcher’s park and his mid-30s hasn’t slowed it down, as Jake Mailhot at Lookout Landing noted a few days ago. That’s odd enough, but the more you look, the harder it is to find any kind of good precedent for Cruz’s late-career surge. To be clear: plenty of sluggers have enjoyed long careers and had very productive seasons in their mid to late 30s, but these players generally had very productive 20s and early-30s as well. Jim Thome was great in his last 30s, but he was great in his 20s, too. In the main, it’s hard to find players who weren’t so hot at 31 and then turned into all-star candidates years later for the obvious reason that most players who aren’t great at 31 are busy applying for a real estate license or dreaming of a charter fishing business.
Cruz’s power allowed him to stick around and make some adjustments, but what he’s done with those opportunities are just about unprecedented. Mark McGwire’s age 34-35 peak seems like a good comp, but that obviously comes with a pretty big asterisk, as does Bret Boone’s 32-35 peak. Barry Bonds’ peak is the most famous case of a player’s peak occurring in his late 30s, but Bonds was a Hall of Famer before it, and for a number of reasons, it’s just not fair to use it as a comp.* The obvious M’s parallel is Edgar Martinez, who broke in late (like Cruz), and then took off at age 32, but Martinez never really struggled: Martinez’s peak came late, but Edgar was never a 1-win player (excepting his injury years) the way Cruz was. David Ortiz enjoyed a late-career renaissance after two poor years in his early 30s, but his best years were from age 29-31. The best examples I’ve come up with are Ellis Burks and Hank Sauer, two corner OFs from very different eras who struggled with inconsistency before putting up brilliant late career numbers.
Burks broke out in Boston in 1987-88 along with RF Mike Greenwell. Burks was a standout CF for a few years, but scuffled in his age 26-28 seasons, resulting in him playing for three teams in three seasons. A move to Colorado in 1994 certainly helped revive his slash line, but a trade to San Francisco seemed likely to put an end to his years of gaudy statistics. Instead, Burks thrived, peaking at age 35 with a .344/.419/.606 line in a tough park. Yes, yes, steroid era, and yes, his WAR numbers were better either at 23 or 31, but from a park-adjusted batting standpoint, that 2000 season was his peak. Sauer was a middling OF before losing a few years to military service in WWII. He came back and had another mediocre year for Cincinnati and then found himself stuck in AAA for his age 29 and 30 seasons. He topped 30 HRs in 1948 and 49, but a low average and the high scoring environment meant he was still a solid-average regular as opposed to any kind of a star. Things turned around at age 32 when he was sent to the Cubs in mid way through 1949. He made the all-star team in 1950, his first full year in Chicago (at age 33), and won an MVP award at age 35 in 1952.* Both Burks and Sauer aged fairly gracefully from there – Burks hung around as a useful DH in Cleveland for a while, and Sauer’s 1954 may have been better than his MVP year.
But that’s essentially it. Any projection system is going to apply generic aging curves, and those curves are going to *destroy* a guy like Cruz. This the reason why Steamer and PECOTA have surprisingly weak batting lines for Cruz. Steamer’s line of .255/.321/.476 and PECOTA’s essentially identical line of .256/.315/.472 are marginally better than his 2011-2012 trough, but nothing remotely like his 2014-2015 peak. CAIRO’s projection looks a bit better, at .272/.338/.505, but it may not be park adjusted. At least CAIRO’s got him down for over 3 WAR, while Steamer has him below 2, or a below-average regular. PECOTA’s 2.6 WAR forecast looks better, but it sees him as a RF (with a curiously conservative estimate of his defensive cost), whereas Steamer, CAIRO and I think he’ll DH this year.
While Cruz’s black swan aging curve means the projection systems can’t really figure him out, it’s not like the rest of us are in a better position to judge his production over the remainder of his contract. If he really is (nearly) unprecedented, humans will throw up their hands just like the CPUs that produced PECOTA or Steamer. Jermain Dye is Cruz’s top PECOTA comp, but he was done at age 35. Sauer and Thome (and Burks, to a lesser extent) were productive throughout their late-30s, but Michael Cuddyer (and Duke Snider) weren’t. As we’ve mentioned, it seems like a fools errand to find a meaningful trend in any of his peripherals. Ultimately, if you’re an optimistic sort, and want a reason to believe that the M’s are better than their 83-84 win projections, Cruz is a great place to start. Fangraphs/PECOTA missed badly last year, and you can accurately say that these systems just don’t know what to make of him. If you’re not, you can say that it’s all downhill from here, and that many/most of Cruz’s comps fell off a cliff after 35, and that no player is guaranteed a slow, dignified descent from star to beloved regular to bench-bat/heart of the clubhouse. What you make of the fact that no one knows anything and math can’t save us is up to you. I have no idea what’s coming, and have simply decided that that’s kind of awesome.
* Was it deserved? Eh, probably not. Stan Musial and Jackie Robinson were both far better, but I’ve seen less-deserving years rewarded.