Swinging Strikes Are Good For Pitchers, But Not Bad For Hitters

by Lucas Apostoleris on Feb 16, 2011 9:00 AM EST in Research

I'm here today in an attempt to illustrate something you probably already know - it's really good if your pitcher can get hitters to swing and miss, but it's really not bad, necessarily, if your hitter swings and misses. The chart below shows runs above average per 700 plate appearances plotted against contact rate for MLB pitchers, minimum 500 plate appearances, from 2002-2010.

As you can see, there's a solid correlation; r = .473 for runs above average and contact rate among the 1152 pitchers that qualified for this study. If we look at hitters, it's a different story:

For hitters, there is a slight negative correlation. R = -.204 for the 1412 hitters here, which says that the hitters who swing and miss the most are more productive than those who make a lot of contact. And makes intuitive sense. Think about the players who are going to lead the league in whiffing - guys like Adam Dunn, Ryan Howard, and Mark Reynolds. They'll miss, but they'll also hit a ton of home runs. The table below is a different way at looking something similar. It shows the mean RAA for pitchers and hitters with a particular contact rate. Keep in mind the relatively sample sizes at the extremes.

PITCHERS

Contact Rate	% of Qualified Players	PA	RAA/700
65%-70%	.006	5640	-24.68
70%-75%	.051	46552	-19.59
75%-80%	.268	242006	-9.83
80%-85%	.496	432663	-0.03
85%-90%	.172	147088	7.85
90%+	.007	5929	10.80

HITTERS

Contact Rate	% of Qualified Players	PA	RAA/700
60%-65%	.004	3799	20.08
65%-70%	.021	18358	27.72
70%-75%	.101	87552	17.64
75%-80%	.260	223048	13.54
80%-85%	.314	273977	7.32
85%-90%	.232	200458	5.59
90%+	.067	59774	6.86

I think the next graph is important. It shows the same data in the table, except with smaller class intervals. The lines have been smoothed to better accentuate the trends.

The large divide between the two lines at the left portion of the graph shows how low contact is often better for both pitchers and hitters. Something that's extremely important to note is that by setting the minimum number of plate appearances at an approximate season's worth, we run into a bias - the guys that stick around for a season must have something going for them, and if they didn't, they'd be off the team. So, definitely don't take this to mean that swinging and missing brings success for hitters, not that you would anyway. The guys that can't make contact, walk, hit for average, or hit for power haven't factored into this sample. If anything, take this to mean that among successful major league hitters, the ones who provide the most value typically also make the least contact. For pitchers, it appears to be more simple. If you miss bats with frequency, you're probably going to do better than a pitcher who allows a lot of contact.

The data here are courtesy of Fangraphs.

**ADDENDUM: I would have loved to do this without having the full-season sample bias, but I would also need enough plate appearances to make contact rate significant. Any thoughts on how to improve upon this are greatly appreciated.

Tweet Comment 15 comments | 2 recs |

Do you like this story?

Comments

Not an expert, but

I think the real bias you’re dealing with here is selection bias, which is inherent to studying major-league hitters — by definition, you’re studying a population that has succeeded, at least to some degree. I think the only way to avoid that would be to expand to include the minors, and then you’re doing a different study (though one that would be quite interesting).

by The Ancient Mariner on Feb 16, 2011 9:18 AM EST reply actions

Yes

Good point. I hadn’t thought about expanding to minor league data, which, admittedly, would be a much more difficult task.

by Lucas Apostoleris on Feb 16, 2011 9:22 AM EST up reply actions

Interesting stuff

I’ve taken very brief looks at hitters along similar lines for contact rates and stuff, although this is more in-depth and promising. Nice job at the work and acknowledging potential problems/limitations.

Making watching baseball as fun as doing your taxes.
My Twitter feed.

by Matt Klaassen on Feb 16, 2011 10:13 AM EST reply actions

Thank you, Matt.

It’s definitely something that I’d like to look at more and improve upon in the future.

by Lucas Apostoleris on Feb 16, 2011 10:24 AM EST up reply actions

Just to reiterate, in case it wasn't clear earlier

The title makes it seem like swinging and missing somehow causes hitters to be successful, which is definitely not true. The point is simply that the hitters who play at the major league level can get away with not making a lot of contact, while pitchers who allow a lot of contact typically aren’t going to be successful. Given that, it makes sense that contact rate is more widely distributed for qualified hitters than for pitchers. Here is a graph that illustrates that (using the “qualified” samples):

As I mentioned before, I’d be interesting in looking for ways to expand on this topic without dealing with selection bias.

by Lucas Apostoleris on Feb 16, 2011 12:47 PM EST reply actions

So my question is why

Why do sluggers get away with whiffing more? Is it that high contact correlates to adjusting swings whichan while increasing contact decreases the impact of contact?

Writer at Beyond the Box Score and tortured Mets fan (is there any other kind?)

by Bill Petti on Feb 16, 2011 1:56 PM EST via mobile up reply actions

That was the first thing I thought of

It seems very reasonable.

by Lucas Apostoleris on Feb 16, 2011 2:20 PM EST up reply actions

but why isn't it symmetric?

That is, if slap hitting raises your contact rate, but doesn’t do much for your production, then why can’t pitchers be successful by allowing everyone to be a slap hitter (i.e., pitch to contact)?

Empirically, the correlation analysis you’ve done shows quite clearly that what’s good for pitchers is not necessarily bad for hitters. Theoretically, I still can’t get past the fact that any individual good outcome for a pitcher is necessarily an individual bad outcome for the hitter, in exactly the same amount.

Any insight on that front?

"The USA despite its flaws and corruption and overall messiness is still a great and powerful instrument of freedom and hope for the entire world." - Madville

by bbjones on Feb 17, 2011 2:46 PM EST up reply actions

If I'm understanding your question, it's all about selection bias

The pitchers and hitters in this study needed a certain number of plate appearances to qualify, so naturally you’re going to get the better players in the sample.

by Lucas Apostoleris on Feb 17, 2011 3:12 PM EST up reply actions

so, they are different samples, of course

If I follow you, suppose the same study were done where we selected out the at-bats that included BOTH a qualifying pitcher AND a qualifying batter.

Then in that sample, there would have to be the same correlation for pitchers and batters, since every at-bat would go into the total score for each.

That implies that the batters that do not qualify have roughly the same distribution of RAA vs. contact rate as those that do, so including them in the sample against which the pitchers threw doesn’t alter the distribution. But the pitchers that did not qualify have a different distribution of RAA vs. contact rate, such that the remaining qualifying pitchers have a substantially different correlation.

Did that make any sense? I’m not sure I’m stating it clearly.

"The USA despite its flaws and corruption and overall messiness is still a great and powerful instrument of freedom and hope for the entire world." - Madville

by bbjones on Feb 17, 2011 4:17 PM EST up reply actions

I think I get most of what you said

So, at least to start with – you’re proposing a study in which only plate appearances with qualified hitters and pitchers are used, correct? Theoretically, the distribution there would be the same, as you said.

by Lucas Apostoleris on Feb 17, 2011 5:26 PM EST up reply actions

I think so

It stems from the first question that pops into my head when I read the title of this article:

How can ANYTHING be good for a pitcher without being bad for a batter, in equal and opposite measure?

You correctly point out that the set of at-bats used in the pitcher plot is different than the set of at-bats used in the batter plot. That could certainly be the source of the difference.

The “common at-bat” study, if it did show the same distribution for batters and pitchers, would prove that the sample sets were the cause of the original difference. And, if we thought hard about it, would could probably say something interesting about performance of bad hitters vs. good pitchers, or vice versa, or something similar. That might turn out to be interesting.

"The USA despite its flaws and corruption and overall messiness is still a great and powerful instrument of freedom and hope for the entire world." - Madville

by bbjones on Feb 17, 2011 6:00 PM EST up reply actions

I agree!

It’s a very interesting idea for a study; thank you for your thoughts.

by Lucas Apostoleris on Feb 17, 2011 6:49 PM EST up reply actions

selection bias

to investigate the contribution of selection bias, what you could do is bin the pool of hitters based on plate appearances and see if it truly exists. this would allow you to also open up the group to, say, 300 PA players, and see if that group behaves similarly, or if they are so scattershot as to muddy the waters.

it’s a different study, but i’d be curious to see how contact rate/raa changes relative to the age of the hitter, and also by season.

by metsmarathon on Feb 16, 2011 2:49 PM EST reply actions

What happens if you combine the two tables above?

Have a row across the top for the pitchers going from low to high contact rate, then a column going down for hitters from low to high. More involved, certainly, but it would be interesting to know how the “Mark Reynolds” group fares against the different grouping of pitchers, and contrasting that with how the “Ichiro” group looks against each pitcher group.

by nathaniel dawson on Feb 17, 2011 9:49 PM EST reply actions