Quantifying the Trade-Off Between Power and Contact

Intro – Setting the Table

For most hitters, batting is an optimization strategy between power and contact, finding that point where the incremental benefit of harder contact is offset by the incremental cost of less contact. Giancarlo Stanton could potentially be a more productive hitter if he tweaked that balance, sacrificing some of his power for gains in batting average. This might not work so much for Ben Revere on the reverse side, where there is little power upside. Is there a way we can quantify how hard a batter is swinging, from a purely numerical standpoint and give us some insight into which batters are sacrificing contact on the altar of power?

A simple approach would be to simply look at the correlation between contact percentage and isolated slugging, which would reveal a negative relationship (higher contact percentage links to lower ISO and vice versa) and a not insignificant .28 R squared correlation. I would argue that this is likely simply measuring the effect that pitchers will pitch around powerful hitters, thus reducing contact percentage. In other words, it’s only half the story; the real question we should be asking is: given that a power hitter will see pitches farther from the center of the zone than the average hitter, are power hitters making more or less contact than we would expect?

To do this, we’re going to take a four-step approach:

1. Demonstrate the link between pitcher respect and batter slugging. Use “Distance from the Center of the Zone” as the key respect measure, quantified as feet from center-center.
2. Demonstrate the extreme link between distance from the center of the zone and whiff percentage (swings and misses/swing), comparing a linear model and an exponential one
3. Develop a metric, dubbed “True Contact,” which measures a hitters whiff percentage less the expected whiff percentage based on where each pitch was thrown in the zone. For example, if Ben Revere had all his pitches thrown right down the middle, he’d have a huge advantage when it comes to making contact, even if he were swinnging harder.
4. Demonstrate the link between True Contact and Slugging on Contact, and show that there is little multi co-linearity between this relationship and the distance/power relationship.

Demonstrate the link between pitcher respect and batter slugging

Last year, Jesse Wolfersberger built upon some prior research that demonstrated we can tell a lot about a hitter simply by the way the pitcher approaches the hitter, using the following metrics: zone percentage, first strike percentage and fastball percentage. In this article we are going to take the concept a step further and replace binary metrics with fluid metrics, specifically Distance from Center of the strike zone (as per PitchFX, measured in feet).

Let us begin by looking at the relationship between pitcher respect and SLG on Contact. On the Y-axis we have Distance from the Center (measured in feet) and on the X-axis we have SLG on Contact. All batters who have faced at least 400 pitches (in their career) are included in this analysis, which shows a 0.47 R Squared correlation, when looking at a batter’s entire career.

Joey Gallo stands out for both having prodigious power and clearly being pitched around more than any other hitter. Miguel Sano, while actually doing more with his contact, is perhaps forcing pitchers to pitch closer to the center of the zone due to his other elite skill. Who in the world is Mikie Mahtook? The latest, greatest example of SSS noise, where a roughly league average minor league player can put up a 168 WRC+ in 115 plate appearances (based on some other research I’m doing he did have the fly ball + HR distance to back it up, so maybe he’s a fantasy sleeper).

When we browse through the names that are displayed, we see at the top of the distance axis guys who have a lot of power and at the bottom a lot of pitchers. Free swingers such as Pablo Sandoval and Josh Hamilton are well above the trend line, suggesting pitchers are taking advantage and throwing them a lot of bad pitches. Chris Carter is below the trend line, which might suggest that pitchers feel confident that there are enough holes in his swing, or his patient approach is forcing pitchers to throw closer to the middle. I can’t explain why David Ortiz is in the same spot as Josh Hamilton; that does seem odd to me.

It is important to note that this relationship does exist in season as well, but is not as strongly, with a 0.28 R squared.

Interestingly, this relationship maintains its predictive power when predicting SLGContact in the following year, with a similar distribution and an R Squared of 0.27:

So, essentially, we’ve re-performed prior research that confirms that there is a strong relationship between how a pitcher approaches a hitter and the quality of contact that hitter produces, which is hardly ground-breaking news. This brings us to the second piece of the puzzle and another relationship that is well known, specifically the probability of a whiff, based solely on how far from the center of the zone a pitch is.

Demonstrate the extreme link between distance from the center of the zone and whiff percentage

I’m going to share two pictures, both showing whiff percentage graphed against distance from the center of the zone, as measured in feet. On top, you’ll see the linear trend line and on the bottom you’ll see the exponential trend line. Using a basic linear model we get a very strong relationship of 0.9, with the exponential model we get an almost perfect model.

Given how neatly the model fit the exponential curve (and without any advanced stats knowledge telling me that I’m over-fitting) I decided to use the exponential model to create an xSwStr% (expected swinging strike percentage) for each pitch based on its distance from the center of the zone. The formula, in case you were wondering, is xSwStr% = -.114*D⁴+0.478*D³-.415*D²+.16*D+ 0.093.

Develop a metric, dubbed “True Contact,” which measures a hitter’s whiff percentage less the expected whiff percentage

The table below gives metrics for all batters who have swung at at least 1,000 pitches in their career. Remember that the distance and whiff percentage metrics here are only on swings, not takes. A higher distance from center indicates the batter is swinging at more pitches farther away from the center of the zone (less selective).

It’s always comforting when the list generated produces names that you would expect to be at the top of the list and the bottom of the list. Ichiro performed almost a full percentage point better than the next closest. When you combine his elite ability to make contact, with his ability to hit pitches anywhere in the strike zone, you get a lot of hits and a great batting average. Salvador Perez and AJ Pierzynski seem to swing at everything, yet don’t swing and miss very often. Marco Scutaro and Chris Iannetta both swing at pitches on average 0.73 feet from the center, but Scutero whiffs seven percent of the time and Ianetta 28 percent.

Robinson Cano and Pedro Alvarez are both at 0.92 feet, but are at 15 percentand 32 percent and respectively. Michael Brantley is clearly exceptional at pitch selection and combines that with exceptional contact abilities as well, all of which jibe with what we know of him as a ball player. George Springer is pretty selective as well, but swings and misses a ton.

Demonstrate the link between True Contact and SLG on Contact

 

The graph above charts Slugging on Contact to “True Contact” as described above. This yields an impressive 0.28 to 0.32 R-squared relationship (depending on the threshold for swings), implying that we can predict 30 percent of batters’ SLGContact (a good measure of batted ball quality) knowing only how much they over or under-perform their expected whiff percentage based on the location of the pitches they swing at. The underlying assumption is that this can serve as a proxy for how hard the batter is swinging and adjusts for the fact that powerful batters will be pitched around more (and thus have a higher probability of whiffing simply due to pitches being farther from the center of the zone).

One thing that pops out at me from this chart is the clustering that occurs where you have hitters clustered in almost identical locations, suggesting these are very stable skills year to year. Look at Albert Pujols 2008-2010, Ryan Howard 2008/2009, Carlos Pena 2012/2013 and Giancarlo Stanton 2012/2015. It also highlights something very interesting about Mike Trout:

Trout has seen his SwStr% (on swings) move up steadily since 2012, but this can be mostly attributed to pitchers pitching away from him more and more (20.4 SwStr% in 2012 to 22 percent in 2015 is mirrored by average distance increase of 0.84 feet to 0.89 feet). This would suggest that his true talent level for making contact has remained steady, while he has significantly increased his damage on contact since ’12/’13. It looks like there is still some upside. Interestingly, 2012 Ryan Braun was very similar to 2012 Trout, but never got close to ’14/’15 Trout. I would have published this to Tableau public but the data sets are way too large, unfortunately. Bryce Harper 2015 has a very similar profile to Giancarlo Stanton, but swings at pitches about 0.1 feet closer to the center of the zone.

The relationship holds up when we look at things from a career standpoint. Look at Kris Bryant clustered with Chris Carter, Mark Reynolds, George Springer and Carlos Pena. The latter fou definitely profile the same way, so a little bit troublesome that Bryant is in that zone.

Why not just use SwStr%?

There is a stronger (slightly) relationship between SwStr% and SLGContact, which begs the question, why not just use SwStr%? Well, essentially, I was looking for two distinct variables with no measurable correlation (True Contact and Distance from the Center of the Zone). SwStr% has about a 0.12 R Squared correlation with Distance, so we’d have to deal with multi co-linearity. 

True contact has almost no relationship with Distance from Center (0.01 R squared), which indicates that the formula above did a good job stripping out the location variable. This brings us to our final chart of the day:

This relationship has a 0.50 R Squared correlation, implying we can predict almost half of a batter’s Slugging on Contact on two variables: How far away the average pitch is from the center of the zone and how often he swings and misses (adjusted for how far away the pitch is from the center of the zone).

Conclusion

So, what have we learned from all of this? True Contact may be a useful measure to classify hitters who either have long swings or are swinging harder than average (or vice versa). This may lead to some clues to the batters who would benefit from tweaking their swings to be shorter. Giancarlo Stanton leads the list above with an .805 SLGContact, but he may be better off shooting for a Goldschmidt-esque -4.5% and end up as a more productive hitter.