In 2012, Ellsbury was sidelined for more than half the season due to injury, posting pretty paltry numbers at the plate (84 wRC+, .099 ISO). He had a much better all-around year in 2013, but the power Ellsbury displayed was pedestrian compared with his 2011 campaign (.128 ISO, 6.6 percent HR/FB).
We can all think of similar scenarios in which a player, coming into his peak, puts up fantastic power numbers only to see those numbers come back down to earth as he ages. It got me thinking about the extent to which power — and the aging of power — could be studied by looking at batted ball distances, not just outcome statistics.
Absent HITf/x data, I turned to my colleague Jeff Zimmerman and his PITCHf/x database. While not perfect, PITCHf/x is what we have available publicly. Also, it is possible to calculate batted ball distance and angle (i.e. field) based on coordinates provided through the GameDay data. The biggest caveat is that the coordinates mark where the ball was fielded, not necessarily where it landed. So, one can imagine a few data points in which the ball rolled to a deeper part of the park after landing. This, of course, would appear to be a deeper driven ball in the data.
I am using data from 2010 through 2013, so there are some limitations in terms of sample size (hence, the curves beginning with age-24 seasons). Pitchers were excluded from the sample, for obvious reasons. And I am looking only at batted balls that are coded as home runs or fly balls, and those that traveled at least 100 feet, and we are pushing the limits of the data. But, as I said before, until we have access to HITf/x we must make do with what we have.
In terms of breaking the batted balls into pull, center, and opposite field I used the following batted ball angles (numbers represent the degree of the angle relative to dead center field):
|Batted Ball Angles|
|Pull||<= -15||>= 15|
|Center||> -15 to < 15||>-15 to <15|
|Oppo||>= 15||<= -15|
For those familiar with our work on pitcher aging curves, the same method for calculating change by age cohort was used. In this instance, the harmonic mean of the first and second seasons’ number of batted balls was used for the weighting. (You can read more about the method here and here.)
Let’s dive in.
Aging Curves by Batted Ball Direction
There are some slight differences in terms of how distance ages depending on what field the ball is hit to. Distance on balls hit to center appears to decline straight through a player’s career. Opposite field power sees a slight increase at age 27, and then begins a decline similar to center field power. Pull power, however, experiences an increase that is akin to what we’ve come to expect when looking at a player’s outcomes–peaking between ages 25-27, and then beginning a decline. Pull power is also more “choppy” than to the other two fields.
I would caution against reading too much into this for a few reasons. First, the data. We have to remember that we’ve got limitations with volume, quality, and of course the survivor bias that can’t be completely controlled for in using this method. Second, it could be that as players start to feel the effects of aging (slower pitch recognition, slower bats, etc.) they adjust their approach such that they are trying to hit the ball up the middle and the other way with more power than before.
Aging Curves by Handedness
If we split the curves based on batter handedness, we observe some slight, but interesting, differences.
First, if we look at pulled balls we see that right-handed hitters generally gain distance between ages 24-27 before experiencing a sharp decline. Left-handed hitters, in contrast, experience a slight increase between ages 25-26, but then start their decline as well. The key difference appears to be the pace of decline.
Between ages 31 and 35, right-handed hitters’ distance dramatically declines, while lefties seemingly stabilize into a more gradual decline at age 32. I’m not sure what’s behind the difference. Could it be that right-handed hitters are able to get away with less power for longer? My initial guess would have been to say no (due to generally shorter porches in right field than left field across ball parks, speedy lefties may have an advantage productivity-wise without power that righties don’t have), but maybe I am missing something.
Let’s return to Ellsbury. In 2010, his average pulled home runs and fly balls traveled roughly 276 feet. During his breakout 2011, however, his average distance skyrocketed to 304 feet. Ellsbury added 28 feet on average to the balls that he pulled and elevated. In 2012, however, his pull distance declined to 280 and in 2013 he averaged 287.
Now, compared to 2011 he clearly suffered a decline. But relative to left-handed hitters over the same age span, that increase goes against the general aging curve. Lefties, from ages 26-29, generally lose about seven feet of distance when pulling the ball. So while Ellsbury isn’t hitting home runs at the same pace he was for that magical season, his physical power trend should be encouraging.
The patterns for opposite field distance are remarkably similar by handedness. Outside of the slight increase at age 26 for right-handed hitters, the two curves are almost identical. There is a slight bit of accelerated loss by left-handed hitters around age 31, but overall there isn’t a lot to tease out when it comes to going the other way.
Center field power is a little more interesting. Righties and lefties start off quite similar, but diverge after age 29. Left-handers see their distance decline at an accelerated rate relative to right-handed hitters.
Taking all three graphs together one gets the impression that, to stay in the majors and be productive, left-handed hitters might be consciously shifting their approach to driving the ball deep to their pull field. This makes sense when we think about the positions lefties (those who throw as well as bat left-handed) tend to play–first base and corner outfield. It’s hard to be a light hitter and stick around if you are playing one of those positions. Now, these curves aren’t looking just at lefty throwers — you have switch hitters blended in to each curve–but the issue of position would explain some of the differences in the curves above.
Physical Changes in Power and Power Outcomes
We can also see parallels to pitcher aging. In our original study, Jeff and I found that the aging of pitching performance was tightly coupled to change in their fastball velocity. The relationship wasn’t perfect, however, and graphically it showed that pitchers can counteract the effects of a dying fastball. Essentially, performance aged more slowly than fastball velocity.
We see a similar relationship between batted ball distance — particularly, balls pulled by a hitter — and performance from a power perspective. Below is an overall aging curve for pulled batted ball distance plotted against an ISO aging curve. The curves both cover 2010-213 and the same age buckets are represented to keep things consistent:
This graphic looks very similar to what we saw with pitchers. The raw physical skills were important (whether that be velocity or distance), but performance decline was not as sharp as physical skill decline.
It shouldn’t be surprising that a relationship exists, but I wanted to dig a little deeper and see the magnitude of the effect.
I matched up the distance of pulled batted balls and a player’s SLG, ISO, and HR/FB in the same year and ran some simple correlations. The results are pretty good and what we would expect:
Correlations between Distance and Outcomes (Same Years)
|Correlation with Pull Distance|
At first blush, this aligns with the idea that distance can’t (and shouldn’t) explain everything about production, but it does explain a decent chunk. The numbers are also consistent with what Chad Young found about a year ago (although our methods are slightly different and he was not discriminating between all fly balls and pulled fly balls). But what about change in distance–how closely does that tie to outcomes?
Here, I matched up the year to year changes in distance with year to year changes in each of the three metrics above. The correlations are pretty good–not amazing, but a decent size and directionally what we would expect:
Change in Metrics and Change in Distance (same years)
|Correlation with Change in Pull Distance|
Changes in pulled batted ball distance have the strongest correlation to changes in HR/FB. ISO came in with a .34 correlation, and SLG was not far behind at .29.
If we are going to gain anything explanatory or predictive out of the distance data it appears that pulled data is the only way to go. I also ran correlations for opposite field and center field distance, but the results were not robust. In fact, the best correlation was for opposite field distance and SLG (.13).
In terms of predicting change, I took a similar approach but this time matched up the change in pulled distance from years 1-2 and correlated that with changes in each outcome metric from years 2-3:
Change in Distance (Time 1 to Time 2) & Change in Metrics (Time 2 to Time 3)
|Correlation with Change in Pull Distance|
The results here are similar. The correlations aren’t incredibly strong, but they do tell a story that aligns with the aging patterns mapped out above. As distance decreases we should expect–in general–to see decreases in outcomes and power productivity.
We’ve learned that batted ball distance declines as players ages and that seemingly impacts their production. Amazing!
All joking aside, I think the takeaway from this research is a better understanding of the rate of physical aging and the degree to which it impacts performance aging. To stick on major league rosters, hitters have to overcome their declining physical skills and find a way to still be productive. Much as with pitchers and velocity, they can’t simply rely on the raw power of their youth.
It also appears that left-handed hitters make more effort to focus on pulling the ball for power as they age, which is likely a function of the positions these hitters play and what is demanded of their offense. Right-handed hitters may have more flexibility in terms of their approach and may find it easier to remain on major league rosters even as their power declines.
In terms of player evaluation, the data suggest that tracking changes in batted ball distances–in particular, pulled batted balls–does add some explanatory power over current and near-future productivity. No doubt, front offices have already worked these data into their projection models. (And we are talking about even more telling data such as launch angles and speed off the bat.) For those of us without access to HITf/x, however, there still might be something to be gained by incorporating pull distance into our models.
As always, many thanks to Jeff Zimmerman for access to his data and his advice throughout this research. If you want to look up batted ball distances and angles by player, Jeff’s leaderboards can be found here.