Batted balls and cheese

Stringers are in every major league park, and most levels of minor league ball, too. They manually record various aspects of the game as it progresses. If you’re watching an MLB.com Gameday feed, you’re seeing a combination of PITCHf/x data (speed, location, pitch type, etc.) and stringer observation (where the ball went, how it got there, who fielded it, etc.). There’s a level of detail that’s not often discussed—or present—in the Gameday information, that could provide assistance in evaluating pitches, and the pitchers who throw them.

Two weeks ago, I looked at the variation, or bias, shown by stringers when classifying batted balls. I take interest in such things since I’d like to calculate pitch-by-pitch run values based on the type of batted ball, not the outcome. In other words, I’m interested in line drives and fly balls, not outs and hits directly. While Gameday consistently provides classifications for line drives, fly balls, pop-ups and grounders, it often provides an nice little descriptor—soft or sharp. If you look closely, you’ll find that it is possible, according to some stringers, to bunt the ball sharply, or even hit a sharp pop-up.

Refresher on batted balls

Since clicking the link to an old article (two full weeks!) is taxing, here’s a breakdown of batted ball types and value and likelihood of various outcomes. (Outcomes being hits and outs, values being Linear Weights.)

  Home Run Single Double Triple Out
Line Drive .022 .524 .174 .015 .224
Ground Ball .000 .219 .018 .001 .693
Fly Ball .119 .057 .082 .013 .597
Pop Up .000 .013 .014 .000 .975
LW 1.468 .489 .768 1.052 -.289

Contact types

Now let’s layer on the cheddar cheese. Soft or sharp, or none of the above. Home runs are never sharp nor soft, and any play that has an error results from a normal batted ball. Allegedly. I give home runs their own contact tag, errors (and bunts) I’m usually ignoring.

event contact #
Bunt soft 6
Bunt normal 2,550
Bunt sharp 1
Error normal 1,203
Fly ball soft 1,109
Fly ball normal 27,867
Fly ball sharp 165
Fly ball Home run homer 3,924
Ground ball soft 752
Ground ball normal 46,718
Ground ball sharp 1,209
Line drive soft 1,644
Line drive normal 18,787
Line drive sharp 812
Line drive Home run homer 471
Pop-up soft 147
Pop-up normal 8,475
Pop-up sharp 1

Not every park (i.e., stringer) tags batted balls at the same frequency. The deeper we go, the more we need HITf/x.

Park Tag Freq.
sln .0957
bal .0815
flo .0766
nya .0741
was .0732
nyn .0684
bos .0671
phi .0670
kca .0635
cin .0629
atl .0625
mil .0576
chn .0574
tor .0542
sfn .0520
tba .0514
cha .0448
cle .0442
col .0395
oak .0391
sea .0388
tex .0371
min .0364
ana .0324
det .0293
pit .0274
ari .0206
hou .0198
sdn .0196
lan .0159

AT&T is smack on the average (.0524). The difference between Busch III and Chavez Ravine is six-fold. That’s a problem, but I’ll forge ahead.

What’s a sharp liner worth to ya?

Breaking down the batted balls by contact type (and ignoring home runs), here are the event probabilities by batted ball type.

Line Drive # Single Double Triple Out(s)
all 21,243 .537 .178 .015 .211
normal 18,787 .527 .188 .016 .213
sharp 812 .413 .209 .018 .249
soft 1,644 .707 .052 .001 .167

Line drives are the most likely to be tagged—nearly 12 percent. The sharp line drive yields more outs than the other types. It also gets fewer singles and more extra base hits. The soft line drive is turned into fewer outs, more singles and far fewer extra base hits. Intuitively, beyond the sharp liners being turned into more outs. I can speculate about the human factors involved, but I’ll leave that for the comments.

Pop Up # Single Double Triple Out(s)
all 8,623 .013 .014 .000 .975
normal 8,475 .011 .014 .000 .977
sharp 1 .000 .000 .000 1.000
soft 147 .122 .034 .000 .844

I suppose the soft pops are the bloops over the infield. Less than 2 percent of pop-ups are tagged, so not much to see here.

Ground Ball # Single Double Triple Out(s)
all 48,679 .220 .018 .001 .644
normal 46,718 .198 .016 .001 .666
sharp 1,209 .775 .123 .008 .065
soft 752 .697 .003 .000 .198

More than 9 percent of grounders are tagged. Not surprisingly, the sharp grounders have good outcomes—so good, they’re the best of the lot. Home runs not included, of course. A soft grounder is a good thing, too. This is the only type of the four that has more sharps than softs.

Fly Ball # Single Double Triple Out(s)
all 29,141 .064 .092 .015 .622
normal 27,867 .038 .092 .015 .642
sharp 165 .042 .333 .103 .352
soft 1,109 .707 .069 .002 .162

Fly balls are tagged as often as grounders, but lean heavily toward soft over sharp when tagged. Ground balls are tagged more on the sharp side, but the majority isn’t overwhelming. Sharp fly balls are, essentially, extra base hits. Soft fly ball outcomes are very similar to the same contact outcomes for both line drives and ground balls. I wonder if a soft fly ball and a soft line drive are actually the same thing.

Conclusion

We really need HITf/x. Well, we do have some data: 15,000 batted balls from April. Next week I’ll wrap up this series by comparing HITf/x data to various stringer tags—batted ball type and contact.

References & Resources
Batted ball classifications from MLBAM’s Gameday, data from 2009 MLB regular season through Sept. 13


7 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
Brian Cartwright
14 years ago

…and the problem is even more extreme in the minor league. My Gameday data is 2006-2009, but shows the same patterns at the major league level.

Stevenell
14 years ago

Wow, these are even more biased than the normal classifications.  Obviously, A stringer wants to point out that a guy hit a “sharp” line drive when he makes an out, but doesn’t worry about it as much if he gets a hit. 

Same thing with “soft” ground balls.  they want to make sure t is known that it was a swinging bunt, so they make sure to tag it.  If the person was thrown out, they might not think to tag it as such.

At least those are my theories.  Looking forward to the hitf/x.

Colin Wyers
14 years ago

I think that over time the data quality issues with Hit F/X will prove more malleable than the data quality issues with human stringer data, although I could be wrong. And once you actually track the ball along the entire flight path (which is part of the overall DRE they were showing us in SF) the issues with Hit F/X become a moot point.

But yes, distance, vector and hang time will tell you pretty much everything you want to know. I know you at BIS are starting to track that, and I believe MGL is working on a project along those lines, but then you have to go from someone recording the data to actual analysis of the data.

Alan Nathan
14 years ago

The problems with hitf/x that various of you refer to are indeed tractable.  A couple of months ago, I started work on a technique to correct the hitf/x data for the fact that the ball is tracked over a region that does not include the contact point, then extrapolated to the contact point assuming constant velocity.  The velocity is not constant, but the the change in velocity (i.e., the acceleration) can be estimated and the data corrected.  When this happens, then the reported data will more accurately reflect the velocity of the ball (magnitude and direction) at the impact point.

Unfortunately, as with many projects I begin, I have been sidetracked with other things so I have not finished up this project yet, but will do so in the next month or so.  Too many projects…too little time!

Harry Pavlidis
14 years ago

jedlovec3 – no, I did not, but I should. Maybe that will go into the next follow-up.

Alan – you’re retired, allegedly, so get to work on this! You have the autonomy

fjm(anuel)
14 years ago

I echo Stevenell’s sentiments.  Namely, that there may be an inherent bias in the classification of balls that are turned into outs, so that many “sharp” line drives that are hits are unreported as sharp.

Harry Pavlidis
14 years ago

I agree with Colin. The human factor is far less tractable than the problems with HITf/x