Are Sinkerballers Affected by the Team's Spring Training Location?

On Tuesday, the Baltimore Sun had an article about Andrew Cashner and how he feels coming into the season.  It read:

Cashner said having spring training in Florida for the first time in his career — he trained in Arizona when with the San Diego Padres and Rangers – allowed him the opportunity to get a better feel for his sinker.

“You go to Arizona and you really don’t have a chance to work on your sinker because it really doesn’t sink there, so I think I really had the chance to work on it all spring,” Cashner said. “I feel like I’m in the best place I’ve been in a long time with my breaking ball.”

I have not seen this suggested before and it is a fairly interesting idea.  If you are unable to see the outcome of your pitches, it might well be difficult to get ready.  Last year, Cashner threw a sinker 54% of his pitches, so he certainly would be a prime candidate to be impacted if this is a real effect.

I looked at seasons from 2015 through 2017 for pitchers who threw sinkers 40% or more of the time.  I considered these pitchers to be heavily dependent on sinkers.  For Florida, the data set contained 23 player seasons.  For Arizona, the data set contained 28 player seasons.  A few players are counted twice for their separate seasons.  The metrics I used for comparison were FIP and wOBA.

40% SI/2S	FIP		wOBA
	April	Rest	April	Rest
Florida	3.70	4.00	.296	.319
Arizona	4.35	4.32	.317	.320
t-test		0.34		0.26

The data is messy.  That is the first thing to notice.  In general, it appears that Florida sinkerballers see their performance erode past the first month while Arizona sinkerballers remain the same.  However, none of the data here is significant (t-test).  If this was a significant finding, it would perplex me a little bit.  The narrative should be that Florida pitchers would remain constant while Arizona pitchers improved.  We do not see that.  Really, the only aspect that came up with a significant t-test was when we compared Florida pitchers wOBA for April to the rest of the season (0.02).  They performed better in that month and then significantly reduced their performance.

I decided to take another stab with, of course, worse data.  I looked at pitchers with a sinkerball rate greater than 50%, which is more representative for the kind of pitcher Andrew Cashner is.  At first, I took a subset of the above dataset, which was a far reduced amount of data.  A couple weird things emerged, so I thought it was best to increase the data set.  For this, I looked at seasons 2011 to 2017.  This yielded 17 Florida player seasons and 23 Arizona player seasons.

50% SI/2S	FIP		wOBA
	April	Rest	April	Rest
Florida	3.61	3.83	0.296	0.317
Arizona	4.04	4.02	0.301	0.312
t-test		0.42		0.59

And, so we got more of the same, which is not interesting and also a little bit interesting.  It is interesting that we again observe an insignificant decrease in FIP for the Florida pitchers while the Arizona ones are level.  We then see a significant increase in wOBA for the Florida pitchers (0.05).  Meanwhile, nothing else seems to emerge from the data.

So, what we might be seeing is just a weird artifact that will disappear with a larger dataset.  Strangely, what we do not see is what Cashner appeared to claim we would see, which would be a poor handle in the beginning of the year and then improvement.  The only thing that might be happening is that for some reason sinkerball pitchers who train in Florida seem to come out of the gates very strong in the first month and that advantage appears to escape them as the year progresses.

When To Worry About Hyun Soo Kim...Not Any Time Soon

Kim's First Spring Training Hit Orioles Hangout

This spring, Hyun Soo Kim is 1-24 with only three strikeouts (the one hit being an infield single).  That is rather impressive to get out so often on a batted ball.  I have yet to see him play, so I do not know about the quality of the batted balls.  However, it seems that they might mostly be rather weak contact.  This might cause some concern for people who were expecting him to play a major role in left field this season, but how valid is that worry?

Last year, Matt Wieters was coming off his arm injury and was 0-23 with a walk.  That performance did not appear to reflect his performance during the season.  Wieters was certainly still getting back to playing form and was unable to catch consistently at that point, so maybe the comparison is not exactly apt.  That said, Kim is adjusting to a new league, a new culture, and even a new Spring Training setup that is condensed from the Korean three-month norm to that American standard of about 33 days.  Some might remember last year when Jung Ho Kang also had a long acclimation period that was as bleak as Kim's current experience.

We can try to discern something by taking a quantitative approach.  Looking only at MLB season production, Russell Carlton found that the only indicator that might be useful with less than 100 PA (which no one ever achieves in Spring Training) is strikeout rate.  Nothing else reflects season numbers beyond that metric.  I would argue though that when we see Kim not performing, we are not trying to discern what his seasonal average will be.  What we are trying to figure out is more of an up-down conclusion.  Will Kim be good or not?

As a quick first look, I looked at all of the players with rookie status in 2014/2015 and compared their performance in Spring Training against their performance during the regular season.  Very simply, I bucketed the 60 players into three groups with ascending Spring Training slugging percentage. 

Bucket	sOBP	StDev	sSLG	St Dev
1	.300	.034	.382	.069
2	.312	.032	.410	.062
3	.312	.033	.406	.059

Traditional standard of significance (0.05) was not met, but Buckets 2 and 3 were far more similar than Bucket 1 was with either 2 or 3.  In other words, Bucket 1 gives the incredibly noisy appearance of being different than Buckets 2 or 3.  However, if you look at even five or six of the players in Bucket 1, that vague difference will in no way give any suggestion that there are differences.  I would say that this initial test informs us that there is not a great reason to think there are differences, but further investigation should be considered.

I next tried to see how important the Spring Training data was for predicting actual results in conjunction with a projection model (i.e., ZIPS).  The result there was that Spring Training data account for less than 2% of the final performance.  For players with slugging below .400 during Spring Training, the contribution to the model rose up to 4%.  In other words, trust ZIPS. That model has Kim at 272/338/424.  If Kim goes 0-60, then this model would adjust him down to 270/335/420.

Keep in mind that this is a pretty rough cut to answer the question.  Players that were awful and simply were assigned to the minors are not included here.  The players within this population were seen as meaningful enough to stay in the majors for a considerable amount of time.  That likely results in a survivor bias.  That said, we see similar effects for players who greatly excelled in Spring Training.  If we were to think that Spring Training provides enough meaningful events to discern how poorly one might do, then we should probably be able to discern how well someone might do.  I could not establish that with this data.

Sunday Comics: Happy Spring Training!

I apologize for disappearing last week - I was swallowed by graduate school work, but now that I'm able to breathe again here's this week's cartoon!

Really, Spring Training is just an excuse for baseball players to play golf and all of us to get really excited.

Jake Fox will hit 30 home runs?

Just a short post, but I saw some folks quite excited as Jake Fox hit his 9th home run of Spring Training.  I do not mean to be pessimistic, but just to temper expectations. Since 2007, there have been 26 instances of a player hitting at least six home runs in at least 50 at bats.  Actually, there have been 28.  I did not include two players who did not appear in any regular season games.

I compared HR per At Bat from Spring Training to Regular Season.  Players had a 56 +/- 18 % decrease in home run rate.  Jake Fox is averaging 7.22 at bats per home run.  A 56% decrease would mean a home run every 16.4 at bats.

That translates to 30 home runs every 500 at bats.

The 95th percentile would put him between 18 and 42 home runs per 500 at bats.  Anything above or below would be significantly different.