Do hitters get more jumpy during a slump?

One of the criticisms thrown at Sabermetricians is that we are analysts who do not appreciate the richness of the psychology that goes into the game and its players.  And for what it’s worth, the charge isn’t completely without merit.  Many of our models look at baseball as simple agglomerations of probabilities without any sense of what’s going on inside the players’ heads.  The place where this particular argument has gotten the most play is in the clutch hitting debate.  After all, say the doubters, some people have a psychological ability to perform in the clutch while others freak out.  And they’re actually right, at least generally.  The problem is that over the course of a baseball season, the actual effect of this “clutch” ability is fairly minimal.  Whether Bill James’s fog is to blame or whether it’s Mike Stadler’s, author of The Psychology of Baseball, theory that while this ability exists in the general public, baseball players make it to the Majors in part because they all have this particular “clutch” ability, clutch hitting ability has consistently shown itself to be a (very) minor player in explaining the variance in actual outcomes.
But, clutch hitting isn’t the only place where a player’s mental state might affect an individual at-bat and make him something of a different man from one at-bat to another.  Consider the slumping batter.  He’s had a bad couple of days (weeks?) and he just can’t seem to get a hit.  He might be feeling a little desperate.  Will he ever get on base again?  Perhaps he should swing a little bit more or a little bit harder to try to break out?  Sportscasters like to call this “pressing.”  But does it really happen?
In The Book — Playing the Percentages in Baseball (which if you haven’t read, you are a horrible human being), StatSpeak friends Tangotiger, MGL, and Andrew Dolphin laid out the case pretty convincingly that as far as the actual outcome of the at-bat, a hot streak or in a slump has very little predictive power over what will happen next.  You’re better off betting that a player will do what he normally does over the course of the season.  But, that doesn’t mean that our esrtwhile batter gets to that outcome in the same way as usual.  A slumping player’s outcome may be the same as might be expected were he not slumping, but he may go up to the plate with the mindset that he needs to do something different in this plate appearance.  Perhaps he might take a few more chances on some pitches and swing a little bit more.  It makes sense that he might try this strategy.  Let’s look at the data.
First, let’s define a slump.  I took the 2006 season and eliminated all the pitchers batting.  I then set up to look at each plate appearance and the ten that came immediately before it.  A player was in a slump if during the last ten plate appearances he had made an out in at least nine of them.  That’s a really rough definition of “slump”, but it keeps things manageable. 
Now, how to tell if a player is pressing or not.  At first, I looked at pitches per at bat.  Do players who are in slumps have shorter at bats (they poke at the first ball near the strike zone) than when they aren’t slumping?  The answer is no.  I took everyone in baseball with at least 50 PA in 2006 and calculated their average pitches per PA when they were slumping and when they were not.  Then, I ran a paired samples t-test to see whether there was a significant difference between the two groups.  A paired-samples t-test has the advantage of comparing people to themselves, so that there’s not the confound that batters who have longer at bats might be better hitters overall and thus less likely to go into slumps.  Players saw an average of 3.70 pitches when not slumping and 3.69 when in a slump.  There’s a problem though: number of pitches doesn’t tell you what the player did on those pitches.  For example, a player could take two balls and a strike, then put the ball in play (1 swing in 4 pitches), or swing at four pitches (and foul one off) for a strike out.
I needed a better idea for how to measure a player’s willingness to swing at the plate.  His jumpiness factor, if you will.  Thankfully, a brilliant researcher named Russell A. Carleton published a paper (pdf warning) last year in SABR’s Statistical Analysis newsletter, By The Numbers.  In it, he uses signal detection theory to measure whether a player’s willingness to swing, correcting for the fact that some players don’t see a lot of swing worthy pitches, comparatively.  A strike may be a strike, but whether it was a called strike or a swinging strike tells us something about a player’s attitude toward swinging.  He called the stat “response bias.” 
So, I calculated the response bias for all players, both when they were in a slump and when they weren’t and compared the two, again with a paired-samples t-test.  Most players in baseball have a response bias around 1.0, which is ideal.  Greater than 1.0 means that they swing too much, less than 1.0 means they don’t swing enough, but a higher number means a greater likelihood of swinging.  Players when they were not in a slump had an average response bias of .965.  When slumping, it jumped to .990.  That difference was significant.  There’s no units to put on those numbers, so you can’t interpret them as .990 somethings, only that it indicates a little bit more of a willingness to swing.  The effect isn’t huge.  Players don’t turn into Vlad Guerrero-like free swingers when in a slump (Vlad was overall a 1.671 in 2006), but they do seem to go up to the plate with a little more urgency.  A little.
I wanted to rule out one possible alternate hypothesis.  Perhaps players who like to swing a lot compensate for a slump by swinging more, but those who are more reserved about their swings actually go to the plate even more reluctant to swing.  I split the group into halves and looked only at those who were above 1.0 in response bias (the free-swingers) when not slumping and then those below 1.0 (the takers) when not slumping.  Both groups increased their overall response bias when in a slump.  Looks like everyone gets a little jumpy from time to time.
So sure, psychology is in play in baseball.  How could it not be?  Players are human beings.  Now, are the effects on actual behavior and outcomes that big?  No.
This is what happens when people practice psychology without a license.  People assume that most (other) people crack under pressure and thus, are unable to come through in the clutch.  In fact, when there’s an actual emergency situation (and here I’m talking about something actually important, where people might get hurt), most people report that while they felt a little afraid, they were able to put it aside and do what had to be done.  Are there people who freeze?  Sure, but they are actually fairly rare breeds.  And I agree with Mike Stadler’s explanation that none of them make it to the Majors. 
Why then do we believe that there’s this amazing performance-to-mental toughness link in baseball?  Because we get most of our information from media members who get paid to lay on the drama and romance.  Dramatic sells, and “mental toughness” is a wonderfully romantic concept, because anyone can be “mentally tough.”  Not everyone can hit a 95 mph fastball 400 feet, no matter what his mental state at the time.  But where’s the fun in that explanation?


6 Responses to Do hitters get more jumpy during a slump?

  1. dan says:

    This seems to have a lot of potential. Although I’m not sure how much people like to categorize the human mind with numbers like that. Way to mix the two sides of your double life in one post.

  2. Pizza Cutter says:

    Psychology, at least scientific psychology, is all about quantifying the human mind and human behavior. You’d be amazed at how many people out there make their living by doing this exact sort of thing. I just happen to use it on baseball.

  3. KCDC says:

    Call me a Royals fan, but 9 outs in 10 appearances isn’t exactly the type of slump that you’d expect to see players getting desperate. That’s a 1-3 night (with the hit at the beginning) followed by an 1-4 and an 0-4. Maybe that’s a slump for Pujols, but if we’re looking for players to get desperate enough for a hit to measurably change their approach, I’d like to see something that resembles a bad week rather than a bad day. The risk here is that you start isolating bad hitters for your “slump” subset, but that’s not too difficult to control. If you’ve still got the model, would it be too difficult to run it again changing the “slump” definition?

  4. Pizza Cutter says:

    I could change the slump definition, but the more exclusive I get (say it needs to be 2-for-20), the lower the sample size of slump at-bats. I wouldn’t worry too much about isolating bad hitters, because everything is a comparison of hitters against themselves. If I get a chance, I will fool around with it.

  5. KCDC says:

    I don’t know how accessible the data is, but if you’re worried about sample size, there’s no need to limit to ’07.

  6. Pizza Cutter says:

    Actually, the reason I only used one year was that to manipulate the data in the way that I did requires a lot of memory. My little laptop eventually starts creaking…

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: