SSNS: C. Anderson, Stroman, L. Castillo
Last week, I reintroduced my Small Sample Normalization Services (SSNS), analyzing strong starts by Dylan Bundy, Jose Berrios, and Patrick Corbin in the context of other small samples within their respective careers or recent histories. This time, I discuss three more odd starts among starting pitchers and their implications.
Chase Anderson, MIL SP
Anderson’s 6-game rolling swinging strike rate (Swstr%) is at its lowest point (7.1%) since before he was good (aka 2016). You wouldn’t know it by his results thus far — a 2.86 ERA and 1.01 WHIP through 34.2 innings — but the negative wins above replacement (-0.1 WAR) and 4.70 ERA speak volumes about his good fortune (.196 batting average on balls in play (BABIP), 96% strand rate (LOB%)).
His change-up, his best pitch last year, has seemingly lost its bite, managing an anemic 10% whiff rate compared to last year’s 16%. The pitch, however, also induced an equally weak whiff rate last April before picking up steam and carrying its momentum through September. Maybe, for whatever reason, this specific pitch takes a few starts to heat up. I’ll cut Anderson some slack and assume this is somehow a reliable trend.
Regardless of the change-piece, it’s Anderson’s four-seamer that should bring us concern. Something magical happened in September 2016, during which Anderson’s fastball notched a 13.6% whiff rate (and actually set the table for 2017’s breakout): he changed his release point. As I mentioned last week, I’m still not sure if a change in horizontal release point means Anderson simply moved over on the rubber or he actually changed his arm slot, especially when there’s no discernible change to the vertical release point. Regardless, it’s an adjustment that produced strong results — an adjustment he carried through 2017 and then seemingly abandoned in 2018. Concurrently, the pitch’s whiff rate slipped into the single digits.
As it stands, Anderson is a ticking time bomb. But, also as it stands, Anderson may be a small adjustment away from recapturing 2017’s glory. He’s an obvious sell-high candidate, but keep an eye on his release points moving forward.
Marcus Stroman, TOR SP
It has a rather unflattering start for the Blue Jays ace, whose 5-game rolling average chase rate (O-Swing%) reached its lowest point since mid-2014. But that’s not necessarily a function of declining “stuff” — his overall swing rate (Swing%) has bottomed out in dramatic fashion.
For this, I don’t have an answer. Stroman’s release points and pitch locations look normal enough, as do his zone profiles. So why are hitters leaving their bats on their shoulders? Fortunately, Stroman’s whiff rate is normal, as is his strikeout rate. But the lack of swings have made him especially vulnerable to free passes. Until it gets sorted out, or at least until some kind of explanation surfaces, he’s kind of a risky starter in shallow formats.
All that said, his 46% strand rate is improbably bad. Stroman has never excelled in this regard, but also, no one in the last decade has stranded so few runners in the first month of the season. It’s unsustainably bad, and it’ll regress. On this basis, Stroman is good bounce-back candidate, and his 4.18 SIERA, fueled by a 65% ground ball rate (GB%), suggests he’s plenty valuable even with a concerning walk rate. Buy low, obviously, but don’t hold your breath for a 2017 repeat (it was a little lucky anyway).
Luis Castillo, CIN SP
It’s taking every ounce of my strength not to write about Masahiro Tanaka or Luke Weaver again… so, I’ll turn my focus to Castillo. Let’s first deal with the obvious: the 58.1% strand rate, like Stroman’s is remarkably bad and, as history shows us, will not continue to stay this low. Moreover, it’s too early to expect that, after exhibiting above-average contact manage through his first 15 starts, he’s suddenly below-average in this regard. Through 118 innings, his career BABIP stands at .271; until further notice, he shouldn’t be considered worse than league-average (.300 BABIP). That 7.85 ERA ain’t sticking around.
What’s weird with Castillo is the disparity between his strikeout and whiff rates. His in-season rolling 6-game swinging strike rate has never been better, yet his rolling strikeout rate has tanked. Last time the aforementioned rolling strikeout rate was this low — 19% from July 15 to August 10, 2017 — Castillo was inducing whiffs 10.8% of the time, which was also nearly a career-low at the time. The two rates have tracked each other pretty well to this point. My mental rule of thumb for converting SwStr% to K%: double it. Castillo’s 13.5% swinging strike rate warrants a strikeout rate closer to 27%, give or take a couple of percentage points but clearly much higher than his current 18% K-rate. Castillo’s 4.62 SIERA paints the picture of a wildly mediocre pitcher, but it, of equal wildness, underestimates his strikeout potential.
Two things might be true: (1) Castillo probably overperformed his strikeout rate by a couple of ticks last year; (2) Castillo may not sustain the lofty level on his current swinging strike rate. Neither detracts from the high likelihood of his strikeout rate climbing several ticks over the course of the season. (The whiff rates on his individual pitches are all over the place, so I’m not yet relying too heavily on them. It’d be nice to see his slider rebound, though.) His current metrics with a 25% strikeout rate produce a mid-3.00s SIERA. If you drafted Castillo, you have to hold him. All pitches face adversity — even sophomores who seemed immune to it as rookies.
What do we make of guys like C. Anderson, who have ‘lucky’ results by traditional metrics, but are acquitted well by statcast? https://baseballsavant.mlb.com/expected_statistics?type=pitcher&year=2018&position=&team=&min=100
He shows up with a .310 xwOBA (or a .293 xOBA in Perpetua’s version).
I wouldn’t deny that, of the balls in play Anderson has allowed, a significant portion have been of weak/poor contact. But not all pitchers constantly allow what might be their player-specific average quality of contact. Like, maybe the expected outcomes properly reflect the quality of contact Anderson has incurred to date. But that contact quality allowed is not necessarily predictive (although, in large samples, metrics like xwOBA have better predictive power than other metrics) because player performance ebbs and flows. And it ebbs and flows not only by outcomes (getting lucky on particular EVs and LAs, like a cheap double on poor contact) but also by inputs (that is, the specific EV-LA combination produced by a certain pitch, such that the pitch produced a lucky EV-LA combo to begin with but appeared to produce a cheap double). Player performance is rarely a straight line… looking at any player’s graph, almost any metric, whether input or outcome, has some semblance of a wave.
So if I obscured my point here, it’s that I think a pitcher can acquit himself well even in the prism of Statcast because contact quality, as with almost every other baseball statistic ever, is not entirely skills-based. And while a particular set of outcomes occurred, and the inputs that produced them suggest they were warranted, it does not necessarily mean those inputs (and, thus, those outcomes) will continue to occur.
(Hope that made sense.)
Totally makes sense. I wish it were easier to get a sense for how we might expect those batted ball distributions to normalize. I think the EV/LA data is useful for HR-rate, at the very least, which stabilizes very very slowly by traditional metrics. I bring it up for Anderson because it seems a big part of his breakout last year was suddenly reducing his homers allowed, and that’s something that statcast could help us diagnose in the early-going here. He’s already allowed 7! But by xStats, should have allowed 4.5. So he has some luck in the LOB%/BABIP right now, but some bad luck in the HR’s.
At the very least, I think this provides some solace that he’s not in complete free fall. But as you’ve said above, definitely hard to ignore the loss of K’s so far.
Word! It’s kind of hard to articulate the point, too, that it’s not only outcomes, but also the inputs, that regress The latter is sometimes difficult to conceptualize when you’re caught up in “expected this, expected that,” thinking it’s enough to explain away some of the variance, when there’s actually variance in the things that create more variance… There’s so much variance!
All that said, I’m willing to buy an above-average contact-manager version of Anderson post-2017. It’s just that so much of that success was underpinned by his improved fastball (which I think helped play up his other pitches, too). So, fingers crossed!
Well we might start by asking how predictive/stable those sorts of metrics are for pitchers. Just because there’s an “x” in front front of it doesn’t mean that it represents anything close to a true talent level. From what I’ve seen, they’re pretty volatile, although they’re definitely not without value. I believe Anderson has some legitimate contact management skills, although not nearly to the extent that it would make up for a 16.9% K-rate.
This has more down-votes than up-votes but I agree with it. I think the disconnect is the second sentence, which — I’m assuming on behalf of the author — carries the caveat of “at this point in the season.” But there still is some truth to it in the long-term for some pitchers. Not all breakout performances are necessarily repeatable just because they’re validated by xMetrics, especially for those nearer the tails of a distribution of outcomes (that we unfortunately cannot ever see).