Imagine that a right-handed hitter steps to the plate. Now draw a line from home plate through to center field. It's a bit simplistic, but consider the space to the left (from the hitter's perspective i.e. toward left field) to be pulled territory and the other half, to the right, to be the opposite territory. If the hitter were left-handed, draw the same line and just flip the labels.
Most wouldn't consider a line drive hit just to the side of second base to be "pulled" or "hit the other way", but for the sake of a quick comparison, of the home runs hit last season, 82% of them came on the pulled side of that dividing line. Looking at it from a rate view, 23% of fly balls that were on that pulled side went over the fence and only 4% hit to the opposite ("oppo") side became home runs.
It shouldn't be surprising that most home runs are pulled. Perhaps you, like me, didn't know the magnitude to which it was so skewed until I, unlike you, looked it up. Perhaps you also knew that more fly balls (home runs and all) are hit the other way rather than pulled (foreshadowing). Perhaps you didn't and are skeptical. Would a visual suffice?
Instead of using only two halves though, I'll bisect the field many more times. By using the MLB.com Gameday provided location of the batted ball* and math (hooray, math!), I can grab the angle of the batted ball, with 0 set to the line growing out of the hitter's back, 90 set to center field, 360 set to the line growing out of the hitters' back, 450 set to center field, 720 set to the line growing out of the hitter's back and so on... What, sometimes the balls travel in spirals.
*No, they are not perfect. They are close enough. This isn't homeopathy after all, being off by a little won't kill anyone.
With the angles of all the batted balls (hereafter: "bat angle") available, I directed a program written by people (or possibly robots) not me to separate the angles into 5-degree-wide buckets and then produce me a histogram. Before you look at it and then open your gape hole to ask an obvious question, note that while the x-axis of the histogram portrays the right-handed hitter's perspective (with the "pulled" areas to the left of center), this is an accurate reflection of fly balls hit by all hitters. For left-handed hitters, I used the supplementary angle.
It's a fairly even graph, especially compared to what will come later (additional foreshadowing!), but there's definitely more fly balls to the opposite side of center field, which, again, represents toward right field for right-handed hitters and toward left field for left-handed hitters.
Contrast the above with the angles for home runs hit last year.
As mentioned earlier, a solid majority of home runs are pulled. That is fairly common knowledge in baseball circles, though I still think it's worthwhile to put numbers and pictures to it.
Those are just frequency histograms however and I think what really brings out the stark difference among fly ball behavior is by dividing the two into a rate, the familiar home run per fly ball (HR/FB) rate. You might know that it averages around 11% and that pitchers greatly beating that or getting beat up by it (after park adjustments) are typically pointed out as targets for regression. What I found interesting though is just how skewed the HR/FB rate gets when you break it down.
Look at the difference in rates between the definitively pulled fly balls and everything else! In a large swath of territory stretching from the pulled foul pole to about halfway toward center field, about half of fly balls are home runs. And then it drops to almost nothing remarkably quick.
I'm intrigued by this. Home runs incur massive costs to pitchers and yet neither raw home run totals nor any home run rate stabilizes enough over a season to be worth putting much stock in as far as a performance/talent indicator. On their own, those two points aren't too terrible, but combined, it means that the weightiest factor in many pitching metrics is essentially unreliable. It's the reason for my preference of xFIP over FIP; xFIP replaces the HR term in FIP with an expected home run total based on a pitcher's ground ball rate.
Ultimately, I think we can do better than that. We obviously could with hit F/X data including ball speed off the bat and launch angles instead of just directional angles, but we don't have hit F/X data and I'm doubting we will. I don't think that means all is lost however. There's a chance that we can still use already existing data, examined in a more granular way, to better our approximations.