Further Adventures in Pull Rate
I don’t think I’m alone in my fascination with pulled fly balls. In fact, I know I’m not, because Alex Chamberlain wrote about them today too. These days, we’re practically drowning in data: exit velocities, launch angles, chase rates, aggression rates — the list goes on and on. There are so many different ways of thinking about exit velocity that you can read an entire great article about what they all mean. If you want to translate how hard someone hits the ball into how they’re likely to perform, there’s no shortage of instructive articles. But in that deluge of data, horizontal angle has been left out, for reasons both purposeful and accidental, and the unavailable is always interesting.
Earlier this month, I did some idle digging into what pull rate means for production on contact. The takeaway was, to be generous, middling. It seems like pulling your aerial contact results in better overall production on that contact, but the effect isn’t huge. Perhaps the more interesting takeaway was that xwOBA on these batted balls had a bias: the more pull-happy the hitter, the lower their xwOBA was on the balls they hit in the air. That was the case despite greater overall production on those balls.
That’s a weird little artifact, though I didn’t think too much of it because I kind of knew what it would say in advance. Every time I look at a dead pull fly ball hitter, they’re getting home runs out of batted balls that xwOBA hates. But that doesn’t mean the statistic is working incorrectly; it’s doing exactly what it says on the label by bucketing batted balls based on exit velocity and launch angle.
After I wrote that article, I did a little chatting with xwOBA creator Tom Tango, who wrote a followup piece that took a more holistic look at that gap between wOBA and xwOBA. He found that after taking all batted balls into account, pull-happy hitters produced roughly equivalent overall lines to their xwOBA, and vice versa for the guys who went oppo most often. Here’s the chart from that article, reproduced in lovely FanGraphs green:
Quartile | Pull Rate | Best Speed | wOBA | xwOBA |
---|---|---|---|---|
1 | 39.1 | 99.8 | .326 | .324 |
2 | 32.5 | 99.8 | .321 | .321 |
3 | 27.7 | 99.7 | .319 | .320 |
4 | 21.0 | 99.6 | .315 | .318 |
A slight methodological issue: To come up with those four quartiles, I used my original data, which was first bucketed out into four quartiles by exit velocity, then split in four again by air pull rate. The pull-happiest group in the above chart is actually the top quarter of each exit velocity group. That’s very marginally different from grouping it exclusively by air pull rate, so I’ve done that as well just to show that nothing too weird is going on:
Quartile | Pull Rate | Best Speed | wOBA | xwOBA |
---|---|---|---|---|
1 | 39.1 | 99.8 | .326 | .323 |
2 | 32.5 | 99.9 | .320 | .321 |
3 | 27.7 | 100.0 | .323 | .325 |
4 | 20.9 | 99.3 | .311 | .314 |
On the whole, these are extremely similar numbers, and they tell the same story: You can’t look at air pull rate (the pull rate in question in this study) and use that to imply either outperformance or underperformance relative to Statcast’s expected statistics. I suspect that the reason for this is that an approach that prioritizes lifting and pulling also leads to a lot of rolled-over grounders; our lift/pull heroes are harmlessly tapping the ball to third (or first if they’re lefty) more often than the average hitter.
As you’ve probably gathered, I have a little more to add on the subject. (Otherwise, this would be a pretty short article that only reiterates stuff from elsewhere online.) I wasn’t entirely satisfied with that answer; it still seems like the José Ramírezes and Isaac Paredeses of the world are getting a little extra out of their batted balls. So I went back and grouped by contact quality again, in the form of Tango’s “best speed.” This time, I was looking for overall wOBA differentials, rather than merely what happens on balls hit in the air. Here are the hardest hitters in the game:
Quartile | Pull Rate | Best Speed | wOBA | xwOBA |
---|---|---|---|---|
1 | 39.2 | 102.8 | .346 | .352 |
2 | 32.1 | 102.9 | .351 | .356 |
3 | 27.8 | 102.8 | .345 | .349 |
4 | 21.7 | 102.5 | .338 | .346 |
As before, the hitters with the most power posted the best overall numbers, but pull tendency didn’t provide any additional information. Across the board, everyone slightly underperformed their xwOBA. Figuring out exactly why that’s the case is outside the scope of my analysis today, but it’s certainly of interest. Anyway, time for the next tier, and it’s hardly conclusive:
Quartile | Pull Rate | Best Speed | wOBA | xwOBA |
---|---|---|---|---|
1 | 39.4 | 100.4 | .331 | .329 |
2 | 32.8 | 100.4 | .322 | .324 |
3 | 28.2 | 100.3 | .325 | .329 |
4 | 22.2 | 100.3 | .321 | .323 |
There’s not much to see here. The pull-heavy hitters do the best, and xwOBA thinks they should do the best. The magnitude of the misses in either direction simply isn’t very big. But take a look at the last two groups:
Quartile | Pull Rate | Best Speed | wOBA | xwOBA |
---|---|---|---|---|
1 | 39.3 | 98.8 | .322 | .316 |
2 | 33.1 | 98.7 | .305 | .305 |
3 | 27.7 | 98.7 | .307 | .306 |
4 | 20.8 | 98.8 | .303 | .306 |
Quartile | Pull Rate | Best Speed | wOBA | xwOBA |
---|---|---|---|---|
1 | 38.5 | 97.0 | .305 | .295 |
2 | 32.1 | 96.2 | .295 | .287 |
3 | 27.2 | 96.4 | .290 | .287 |
4 | 19.4 | 96.0 | .293 | .292 |
Now we’re getting somewhere. Here, the gaps look more meaningful to me. Pull hitters in my third group – below-average but not the true bottom of best speed – outperform oppo hitters by 19 points of wOBA overall, against only 10 of xwOBA. In the lowest-speed tier, the gap between the pull-iest and oppo-iest groups is 12 points of wOBA, against only three of xwOBA. These data sets are meaningful, too; three years of data across all of baseball makes for a robust sample.
To be clear, this still isn’t a huge effect. I’m not saying that hitters who look awful might be sneakily great if they just pull fly balls 5% more often. But for hitters with little or middling raw power, visiting the pull side more frequently is associated with better results, both overall and relative to a model based on exit velocity and launch angle.
I think this is an intuitive conclusion. The more fly balls you cluster to the pull side, the more balls that have a chance of sneaking just over the wall where the fence is closest to home plate. That matters considerably more for hitters who are frequently hitting the ball between 95 and 100 mph. Take a look at this chart, which I’m reproducing from an earlier article:
Speed | Pull | Straightaway | Opposite |
---|---|---|---|
<90 | .091 | .107 | .084 |
90-95 | .214 | .015 | .050 |
95-100 | .812 | .079 | .289 |
100-105 | 1.043 | .598 | 1.082 |
105+ | 1.853 | 1.505 | 1.728 |
There are undoubtedly tradeoffs in the groundball department. There might be tradeoffs in the strikeout and walk department, too; if you’re altering your normal approach to prioritize getting in front of the ball, it stands to reason that you might be committing to each swing earlier than you’d like. But on the whole, the numbers suggest that if you can’t mash the ball over the fence the regular way, with overwhelming force, you should probably try to tuck it into the pull side corner. After all, home runs are the currency of today’s game — hitting more of them seems like a good plan.
There’s still more to investigate on this front, but that’s for another day. It’s still February, and this one has gone on long enough as it is. My next line of inquiry in tracking down this small edge available to hitters is to look at two things: how groundball/fly ball tendencies affect this phenomenon and how pull rates vary from year to year. The batted ball tendency part seems important in figuring out who might benefit most from the all-pull life. The year-to-year correlation part seems important because this effect is only interesting if it’s repeatable. But that’s for the future. For now, I’m gonna take a nap. All this data acquisition is exhausting.
Ben is a writer at FanGraphs. He can be found on Twitter @_Ben_Clemens.
Bottom line – it’s always best to nap in preparation for Mardi Gras…