Astros Sabermetric Q&A

Today we start a column answering questions with a sabermetric tint.

If  you have a sabermetric question, leave it in the comment section and we

will try to answer it in the next column.

Question: Brownie mentioned on the broadcast that Kazmir pitches better against right handed batters.  If that's true why did the Astros field an all RH lineup (except for Valbuena) against Kazmir?

Answer: Thanks to Ernie for pointing out this question in the TCB writers' email thread.  I won't get into this from a batter standpoint.  There may be a multitude of reasons for deciding whether a batter should be rested or whether a bench player needs to get a start.  But if a batter needs  a rest day, it may be a good idea to do it when an opposing pitcher from the same arm side is starting.  The question is whether Kazmir, the lefty, is truly a better pitcher against RH batters than LH batters.

Players who can sustain a true reverse platoon split are relatively rare.  I am skeptical of a reverse platoon split which is based on a single season.  Platoon split stats over a single season can be problematic because the splits generally present sample size issues.

Let's look at Fangraphs' splits page. Kazmir had a reverse platoon split for wOBA in 2014: .304 vs. LHB and .285 vs. RHB.  But he only pitched 44 innings against LHB. That's not surprising, since most teams used RHB dominated lineups against him. The reverse platoon split is driven by a high BABIP for LHBs (.345), about 80 points difference between RHBs and LHBs.  BABIP fluctuates more than almost any batting stat, and is dependent to a significant extent on random luck.  RHBs actually slugged better than LHBs against Kazmir in 2014.

How about Kazmir career splits?  He he has the equivalent of 2 or 3 seasons of innings pitched against LHBs over his career. In a much larger sample size, Kazmir pitches much better against LHBs than RHBs (.332 wOBA for RHBs and .287 for LHBs.)  His career BABIP against RHBs is .306 (compared to .287 against LHBs).  Is it possible that the career stats conceal recent improvement against LHBs?  2013 provides no evidence for that idea.  In 2013 he pitched well against LHPs and struggled against RHBs.

The final point is that the reverse platoon split in 2014 did not extend to FIP and x-FIP, which is consistent with his normal career split (example--2014 FIP vs. LHBs 2.86 vs. RHBs 3.21).  So, the peripherals in 2014, which are more likely to be repeatable, do not support the statement that Kazmir pitches better against RHBs.

My verdict: the Astros were correct to presume a normal platoon split by Kazmir.

Question: Can we expect the Astros' hitters to be very streaky because they are so prone to strike outs?

Answer:  With the slow starts by strike out prone hitters like Gattis, Carter, and Springer, it's natural to wonder if the Astros will be a streaky team.  The answer depends on how you define and distinguish the concept of streakiness from volatility.  Streaky players' offensive performance is distributed in periods of hot and cold offensive performance over the season. We  joke about the contrast between Grossman's first and second half performance; that is an example of streaky.  Volatility has been measured in terms of game-to-game consistency in offensive performance.  A batter who is 0 for 4 one day and 4 for 4 the next day is an example of volatility.

Seth Samuels presented a study of streakiness at Fangraphs a couple of years ago.  His study indicates that there is no predisposition to streaky hitting, nor can we predict that a player will be streaky or not based on his performance in the past.  The interesting thing is that basic hitter characteristics like K rate, BB rate, batting average, slugging, etc. have no relationship to a hitter's streakiness.  That's not to say that batters aren't streaky; every year brings a batch of streaky hitters. But that is due largely to randomness, as opposed to innate characteristics of streaky hitters.  Although streaky hitters may not be predictable, he found that there is a tendency for more hitters to be streaky than not streaky.  He suggested that this may be due to factors like scheduling and playing through injuries or illnesses, which may contribute to overall streakiness in baseball.  So, this study indicates that there is no reason to expect the Astros' hitters to be streaky because of strike outs or power hitting.

We reach a somewhat different conclusion if we rely upon Bill Petti's study of volatility at Fangraphs and Hardball Times. He measured volatility as the deviation of a player's daily results with his annual offensive performance. Keep in mind that lack of volatility is not inherently good; a hitter can be consistently bad.  He found that there is some year-to-year predictability of volatility.  Although the year-to-year correlation of volatility is not strong, it is predictable to about the same degree as batting average.

Flyball hitting, K rate, and power (ISO) are all correlated to some extent with greater volatility.  BB rate, base stealing, OBP, and groundball hitting are correlated with less volatility.  This is probably what you expect.  Among elite hitters (during the study period, 2011-2013), Joey Votto and Shin-Shoo Choo are the kings of non-volatile offense.  Do you see a common characteristic?  Both Votto and Choo draw high walk rates and are league leaders in OBP, in addition to traditional power.  That makes intuitive sense.  Batters who can draw walks and get on base, even in games where they don't make hard contact, will be more consistent in making offensive contributions.  The top of the least volatile leaderboard (three year average) is dominated by middle infielders like Scutaro, Andrus, Placido Polanco, and Derek Jeter.  We don't have a leaderboard for 2014, but I am guessing that Altuve would have been one of the least volatile batters last season.

Given the team's power hitting, fly ball hitting, and strike outs, the Astros may end up with more volatile hitters.  But we can't say that the volatility will be reflected in streaks.  If the team has enough power hitters and their big games tend to be randomly distributed, the Astros have a pretty good chance of winning series (even if a clunker game occurs in each series).  At least that's the theory--whether it works out that way, we don't know yet. Gattis' history indicates that he is unlikely to walk as much as the other fly ball hitters on the team.  He may be a little more volatile, as a result.  His difficulties in the first few games may reflect that characteristic.

Question: Is Dallas Keuchel a true ace?

Answer: We frequently read comments that the Astros need a true "ace" starting pitcher.  Every team would like to have as many ace pitchers as possible.  But does that mean Keuchel isn't a real ace?  Since there is no universal definition of "ace," I suppose this is a subjective answer.

Tony Blengino had a good article at fangraphs about Keuchel's contact management skill.  Let's cut to chase.  Here is his conclusion:

he is every bit as good as his gaudy 2014 ERA, and there is no reason to believe he will materially backslide anytime soon. He hasn’t even reached his first arbitration payday yet, so he clearly qualifies as one of the single best pitching bargains in the game at present. He might not look or quack like an ace, but Dallas Keuchel most certainly deserves to be classified as one at present.

Blengino apparently has added hit f/x type data to his analyses:

Keuchel also allowed the lowest average grounder velocity among AL ERA qualifiers, with only Johnny Cueto and Matt Garza allowing lower average grounder velocities in the NL. It should be noted that the average ground ball velocity in the NL is quite a bit lower in the NL thanks to all of those pesky pitchers who get to come to bat on a regular basis.

He adds that Keuchel just nosed out Garrett Richard and Chris Sale as best contact manager in the AL. Blengino makes context and relative production adjustments to pitcher's ERA.  According to Blengino, Keuchel's true ERA was 2.83, below his actual ERA of 2.93.

I'll agree with Blengino and say, yes, Keuchel is an ace.

Inspired by Blengino's contact management methodology, I evaluated Keuchel and the other Astros' starting pitchers late last season.  Link  

I don't have access to velocity data, but my indirect assessment put batters' relative production on all balls in play against Keuchel at 85% of league average--very close to Blengino's estimate of 83%.  Blengino made a context adjutment which reduced that value to 74%, which is well into elite territory.

My evaluation showed Colin McHugh with even better contact management skill than Keuchel in 2014 (batters had 77% relative production on all balls in play).  It would be interesting if Blengino also analyzed the contact management skill of McHugh.  Maybe the Astros can have two aces.