Obligatory Monthly Update on Carlos Gonzalez by Alex Chamberlain July 28, 2015 It’s time for another monthly analysis of Carlos Gonzalez’s tumultuous 2015 season. When we first tuned in, Carlos Gonzalez was bad. Like, really bad. Despite peripherals that suggested some bad luck, the rest detailed a hitter struggling mightily. When we last tuned in a month later, Eno Sarris determined CarGo had been unlucky up to that point, but only slightly. Things had started to turn around, but it was hard to be optimistic. We tune in now, another month later, to find Gonzalez falling short of his previous levels of production but still performing admirably considering the circumstances. To paint a fuller picture, observe CarGo’s statistics at the time we published each of the aforementioned posts: Date PA HR SB AVG OBP SLG ISO BABIP K% BB% wOBA wRC+ as of May 13 110 2 0 .188 .245 .297 .109 .233 24.5% 7.3% .242 32 as of June 16 240 8 2 .242 .310 .400 .158 .272 19.6% 9.2% .310 80 as of July 26 354 18 2 .270 .326 .497 .227 .294 20.3% 7.9% .352 109 You can see the improving batting average on balls in play (BABIP), but you can also see the improving everything-else, too, with the bulk of his value provided in the form of a home run surge. The improvements look much more impressive when evaluated at the margin: Date PA HR SB AVG OBP SLG ISO BABIP K% BB% wOBA wRC+ April 6 – May 13 110 2 0 .188 .245 .297 .109 .233 24.5% 7.3% .242 32 May 14 – June 16 130 6 2 .289 .364 .491 .202 .303 15.4% 10.8% .369 120 June 17 – July 26 114 10 0 .327 .360 .692 .364 .342 21.9% 5.3% .441 170 From May 13 to June 16, Carlos Gonzalez actually performed like a mostly typical Carlos Gonzalez. His home runs paced out to about 28 over 600 PAs, and his linearly weighted metrics such as wOBA and wRC+ would have ranked as his fifth- and fourth-best seasons, respectively, if sustained for an entire season (“entire,” as defined loosely by Gonzalez’s injury history). Since June 16, CarGo has done preposterous, perhaps illegal things to baseballs thrown at or near him from short distances. In some sort of karmic apology for his month and a half of misery, the baseball gods smiled down upon him and cried, “Regress past the mean, old friend.” Speaking of words being said to a person(s), I also said this once, to you: it might be in your best interest to get as large a return as you can for him right now before it’s too late. In that instance, him refers to Carlos Gonzalez, by no coincidence the same Carlos Gonzalez about whom I write today. So let’s address the elephant in the room: if you’re mad at me for suggesting you trade Gonzalez before he lost more value, that’s fine. I’ll own up to that, because I look like a real dingus now. There’s a reason I used ambiguous language to suggest your abandonment of him, but even that wasn’t ambiguous enough to save face. Still, I think Gonzalez’s incredibly volatile 2015 campaign has taught us — or, at least, me — an important lesson related to statistical reliability (aka what one refers to when asking “when [insert metric here] stabilizes”). And that lesson goes as follows: I looked beyond outcomes and into peripherals, which largely validated everything we had seen to date. Eno also looked beyond outcomes and into peripherals, which, again, largely validated everything we had seen to date. Yet at neither point did Gonzalez’s peripherals offer much prophetic wisdom regarding a forthcoming surge. About a month ago, Jonah Pemstein and Sean Dolinar posted a ridiculously awesome dynamic tool to visualize for statistical reliability tested at various sample sizes. It looks only at outcome statistics — home run percentage (HR%), isolated power (ISO), BABIP, etc. — and provides lower and upper bounds for what a hitter’s true level of performance can be at any point in time. For your convenience, I have consolidated the results and presented them below. PA Actual ISO Lower bound Regressed ISO Upper bound as of May 13 110 .109 .039 .129 .195 as of June 16 240 .158 .082 .156 .202 as of July 26 354 .227 .140 .208 .244 The regressed ISO represents the actual ISO regressed toward the mean — that is, based on Jonah and Sean’s sample of hitters, Gonzalez’s actual ISO ought to look more like the regressed ISO. I have included it for sake of demonstration, but it is not as important here. What is important, however, is the upper bound column. I specified 99-percent confidence which, in other words, offers lower and upper limits three standard deviations from the mean in either direction. In layman’s terms, 99 percent of all probable outcomes fall between the lower and upper bounds. Thus, any outcome falling outside the limits is what one could deem “improbable.” Gonzalez’s current level of power production exceeds his upper bound on June 16. It also exceeds his upper bound on May 13. Jonah and Sean’s reliability estimates are descriptive, not predictive, but at 99-percent confidence, it certainly seems statistically improbable that CarGo, as deep into the season as we are, would end up achieving an ISO outside the range of expected outcomes once calculated for him. One could conclude, then, that the extent to which he improved would have been very hard for me (or you or anyone) to predict based on 2015 data alone. Unsurprisingly, several years’ worth of excellence proved more predictive than a mere six weeks’ of toiling. With CarGo, we witnessed a little bit of everything: a recent six weeks of improbably good baseball over-corrected for an early six weeks of improbably bad baseball. Many a fantasy owner overreacted — this author, not an owner, included. In hindsight, we’d all be less alarmed if CarGo’s season thus far happened in reverse. The slump would have been masked by a good start rather than subject to clean slate, for better or for worse. I hesitate to make any kind of claims whatsoever about Gonzalez at this point, but I think it’s safe to say he’s not broken, at least not now. I wondered aloud in May if he was hurt. Maybe his swing was off-kilter. These kinds of things don’t necessarily reveal themselves in the statistics, although you didn’t need me to tell you that. An argument could be made here for the important cooperation of quantitative and qualitative analyses, i.e. statistics plus scouting (or film). But I digress. If Gonzalez was broken, he’s not anymore. He currently portrays a declining but still very CarGo-esque middle-of-the-lineup bat who hits for power and average. He no longer steals bases, but he also no longer hits infield flies, too. His batted ball profile isn’t optimal relative to his past dominance, but I can point to multiple seasons during which he recorded worse ground ball rates (GB%) or hard-hit rates (Hard%), so it could always be worse — as we all know very well by now. He’s currently on pace for 31 homers across 600 PAs, but I’d peg him for closer to 25 if his ratio of home runs to fly balls (HR/FB) regresses toward his 2014, rather than 2013, rate. That’s prior to considering his perpetual inability to stay healthy. Ironically, after all the hubbub I helped stir about cutting him loose in fantasy, he’s probably a legitimate sell-high candidate given the high likelihood that he 1) slows down and 2) hits the disabled list prior to the end of the season. But maybe all he needs to stay healthy is me writing about it.