An Overdue Barrel Rate Refresher
Before the 2020 season, I wrote a series of articles that looked at how much control batters and pitchers exerted over various outcomes: home runs, strikeouts and walks, fly balls, that kind of thing. I found the conclusions helpful, if mostly as expected: batters have more to say about home runs and line drives, but both sides have input on strikeouts, walks, grounders, and fly balls.
I decided to apply the same methodology – which I’ll detail below – to check on something that I thought I already knew the answer to: do pitchers exert any control over barrel rate, and how much do hitters do the same? Barrels are essentially batted balls hit extremely hard and at dangerous angles; I think they’re a great way of thinking about hard contact.
There’s already been plenty of research on the year-over-year stability of batter barrel rate. There’s been plenty on the fact that pitchers don’t do the same. Here’s a preview of my findings: I didn’t find anything that disputes that. I still think it’s useful confirmation, however, and I’m also pretty proud of the method. Thanks to Tom Tango, there’s even a rough rule of thumb to use if you want to estimate future barrel rates. Without further ado, let’s get to it.
The method in question works like so: first, take every batted ball from one season and calculate the barrel-per-batted-ball-event (Barrel%, let’s call it) for every pitcher and batter in the game. Next, divide them into quartiles with an equal number of batted ball events in each. For example, the quartile of pitchers that allowed the least loud contact gave up barrels on 3.3% of their batted balls. The quartiles that allowed the most checked in at 7.9% of batted balls. One minor note: I manually defined barrels for this exercise, which gave me slightly different barrel rates than Statcast (though within half a percentage point of batted balls overall).
That was just the setup. With every batter and pitcher tagged with a year-one quartile, I used the next year’s data to see how persistent barrel rates were. I tossed out pitchers hitting. Then, for every single batter, I separated their plate appearances into five groups: one for each quartile of pitcher, and one for pitchers who didn’t appear in the majors in the previous year. I tossed out the pitchers who didn’t appear in the majors; I don’t have a prior against them. Then, I looked at how each hitter did in terms of barrel rate against each quartile of pitchers. I did this using two years as Year One, 2016 and ’17, which let me avoid using either ’15 (when balls likely changed mid-season) or ’19 (when the coefficient of restitution in the balls may have changed, producing higher exit velocities) in my data set.
For example, in 2018 Mike Trout put 74 batted balls into play against the stingiest (lowest barrel rate) pitchers from ’17. He barreled up 12.2% of those batted balls. He put 83 balls into play against the next-best group (I’ll call them by their quadrants from here on out, so Quadrant 2) and barreled up 18.1% of them. He had 90 batted balls against Quadrant 3 pitchers – 11.1% barrel rate. Finally, he had 71 batted balls against Quadrant 4 pitchers and checked in at an 18.4% barrel rate.
I repeated this process for every hitter in baseball. I then weighted each by the lowest number of pitchers they faced in a quadrant – I didn’t want the data thrown off by Aaron Judge facing 130 pitchers in one quadrant and only 30 in another, or something of that nature. The results were clear. If the pitchers had repeated their Year One barrel rates, there would have been a huge spread between quadrants. Instead, there was almost none:
| Quadrant | 1 | 2 | 3 | 4 |
|---|---|---|---|---|
| Expected Using Y1 Rates | 3.1% | 4.9% | 6.0% | 8.2% |
| Observed Y2 | 5.7% | 6.2% | 6.2% | 6.8% |
Pitchers displayed a wide range of barrel-suppressing numbers in Year One. In Year Two, the “hard-to-barrel” pitchers from Year One hardly performed better than their easy-to-crush counterparts.
On the other hand, batters who barreled up a lot of baseballs in Year One mostly did so again in Year Two. I repeated the same process for hitters; for each pitcher, I separated out their Year Two batted balls by batter barrel rate quadrant in Year One and threw out the batters who didn’t appear in the majors in Year One. The spread was meaningfully wider:
| Quadrant | 1 | 2 | 3 | 4 |
|---|---|---|---|---|
| Expected Using Y1 Rates | 1.6% | 4.1% | 6.5% | 11.2% |
| Observed Y2 | 3.1% | 4.6% | 6.9% | 9.5% |
What does this mean? In simple terms, using this year-over-year quadrant methodology agrees with intuition and previous research: hitters supply the power, just like they supply the home runs. Whether a pitcher surrenders a ton of hard-hit balls in the air has a lot more to do with who they face than how they pitch. On the other hand, there isn’t zero pitcher impact; even equalizing across batters like we did, the low-barrel hitters from Year One allowed fewer barrels than average in Year Two.
That aforementioned Tango data twist? It gives us a rough estimation tool. Based on how far each group regressed to the mean in Year Two relative to Year One, we can estimate a “ballast amount” – how many league-average batted balls you should add to a player’s Year One barrel rate to predict Year Two barrel rate. For pitchers, it’s large – 131 plate appearances, to be exact. For batters, it’s tiny: 15 plate appearances. That’s because pitchers “regressed” 78% of the way back to average (in an average sample of 37 batted balls per quadrant) as opposed to only 33% regression on 30 batted balls per quadrant for hitters.
Why do this kind of study if I already knew the answer? Knowing the answer doesn’t mean it isn’t interesting to test it out another way, and I find this method to be intuitive. It helps me to think about what happens to the group of pitchers who were best at suppressing hard contact in one year and see how they did the next year. The same goes for hitters. Understanding what happens to the group as a whole makes it easier for me to think about how to treat individual hitters.
In any case, it’s great to see the same results no matter how you bake the pie. Does a hitter make plenty of loud contact? Believe it! Does a pitcher surrender barrels left and right? He might be doing something wrong – but you’d have to see a lot of hard contact to believe much of it. I suspect that a huge amount of the pitcher variation in barrel rate, in fact, comes down to the fact that you can’t barrel up a grounder, so groundball pitchers have an edge at barrel suppression. If you keep those two rules of thumb in your head, you’ll be a long way to understanding why pitchers focus so much on what they can control (walks, strikeouts) and hitters focus on what they can do best (pummel the ball into the stratosphere, at least until this year).
Ben is a writer at FanGraphs. He can be found on Bluesky @benclemens.
This is interesting–so groundball pitchers do have an advantage in barrel suppression in a different way than other pitchers do.
I think it’s partly a chicken->egg phenonmenon and partly an egg->chicken phenomenon.
On the one hand, if a pitcher is a groundball pitcher due to their ability to suppress launch angle on balls where the hitter makes contact around the actual barrel of the bat, you convert fly balls that would have been barrels into hard-hit non-barrels (for your high launch-angle hitters) and convert line drives into harmless grounders (for low launch-angle hitters). (Side note: LA-suppressing pitchers do still have a relative platoon disadvantage vs. high-LA hitters, because they’re also less likely to induce balls in play with too-high launch angles – the ones that results in high hang-time air balls with low xwOBA.)
On the other hand, if a pitcher has above-average ability to induce weak contact due to inducing contact too far away from the barrel of the bat (Mariano Rivera being the GOAT of this skill set), his main skill is obviously barrel suppression. However, as a side-effect of this skill, balls in play that were either (a) hit near the bat handle and rolled over to the pull side and/or (b) from contact that actually broke the bat…will generally lead to weakly hit ground balls, more frequently than average.
TLDR – pitchers with skill in suppressing launch angle will produce fewer barrels, and pitchers with skill in suppressing barrel-level quality of contact will tend to produce more ground balls than otherwise (at least, that’s my conjecture.)
I think there’s a question here about whether groundball rate is the cause of inducing weak contact, or whether groundball rate is a reliable indicator of weak contact. Either way, GB% seems relevant!
Great article Ben.
I have a few questions as a follow up:
– do you plan to perform the same exercise on any other years? Although there are pitfalls regarding some of these years, as you mentioned, a single year-to-year sample seems quite small given the days available
– could there be a future article regarding the correlation of ground ball and fly ball rates to barrel% along with year-to-year correlations of the aforementioned rates?
– is there any days available regarding the timing off contact or the accuracy of the contract, for hitters? In other words: how often is the hitter out in front of behind? How often does the hitter contract the ball squarely rather than be a little high or low (most of this should be caught in pop ups and ground balls, but it may be more predictive to go to the source)?
Sorry for the long response – got carried away.
Lol Matt, didn’t see your questions til just now, but those literally go to my hypothesis that I just posted in response to sadtrombone. I’d bet anything that this is the kind of stuff that the analytics folks working for most teams have researched up and down, because the implications are pretty darn huge.
Really quickly on the first one: there just really aren’t any other years to use yet. When 2022 is over, that’ll help, and I could do 2020-2021 I suppose, but 2020 is deeply strange and there haven’t been any other consecutive years. I don’t really think the sample I used (two years worth of y1/y2) is too small, though I’d of course love to have more.
Definitely yes on the second one.
As to the third one, not in Statcast. The data exists — companies like Blast Motion track it, but I’m sure lots of others do as well — but not publicly for all ML batters, so it’s hard to do too much digging into it from my seat.
Love the thought process and article. It takes me back to the “defense of FIP” from a few days ago.
I think WAR is usually used as a measure of what players did that year, and so pWAR shouldn’t include FIP even though it’s more predictive of their future than RA9 or ERA. It’s good to see how barrel rate influences that variation
You seem determined to blame and reward pitchers for things they have no control over.
Hazelrah is determined to not understand FIP no matter how many times it’s explained to them
Neat stuff, again. In regards to barrels and how a hitter achieves such. Is opposite arm pitcher as a % of barrels much higher than same side pitcher? Talking % of balls in play. What are the splits? Also are pull hitters generally higher in barrels overall?
Ben, one more question then I have to go back to my day job (:|)….
Has Fangraphs considered using Barrel FIP in lieu of “IFFB-adjusted FIP” as its basis for pitching WAR?
Who cares about barrels? I would settle for any contact whatsoever. It is getting late early. With almost 1/5th of the season played the major league batting average is .232 even after the NL adopted the DH. For those who say the game has changed and BA isn’t a factor any more I offer a .306 OBP and a .376 SLG. To those who say it is now a TTO game, 21 of 30 teams have hit less than one HR per game. In 2019 only 2 teams hit fewer than one per game. The Tiger are averaging .40 HR’s per game. Yesterday there were more games in which a team got less than 5 hits, 8, than got more than 10, 5 and only one of those got more than 11. In 17 games the batting average was .217 and only 24 HR’s were hit. There is no more dedicated fan than me and I see the game becoming less intriguing when it becomes little more than a game of the pitch and catch.
The reduction in offense is mostly due to MLB messing around with the baseball. I don’t know if there’s been any large change in league-wide plate approach since the launching pad years, but the decreased BA, OBP, and SLG is more likely a product of decreased exit velocity across the board. The new balls come off the bat like they’re full of pudding.
The ball is certainly part of the problem and only makes things worse but there is still a serious lack of balance between pitching and hitting. Before 2019 there had never been a year in which there were more K’s than hits. Each year since the K’s have gone up and I feel contact is the heart of
baseball. Not only more baserunners, but more spectacular fielding by today’s brilliant athletes. Back in the 50’s and 60’s the K/BB ratio was closer to 1/2 and even during the tough stretch in the middle 60’s, culminating in 1968, there were about 5 hits for each K and that was with pitchers hitting. I see nothing that can help restore balance until hitters can catch up to the overpowering velocity of the modern hitter. There was a time when only a few could dominate with sheer velocity but that is clearly no longer the case. A dead ball and limited contact is a program for disaster and I don’t like the only solution but it is extending the distance between the mound and the plate.
I watched another boreathon today and wasn’t the least bit excited even though Nestor Cortes took a no-hitter into the 8th. That is a sad situation for the game.
A pitcher from the Yankees almost throws a no-hitter and a Red Sox fan calls it boring.
Shocking.
I meant 5 hits for each 3 K’s in the mid 60’s.
If MLB sticks with this new ball the definition of a barrel might have to be adjusted.
I vote for “mini-foeder” (0.5 BBL)
I wonder if the deadball problem shows up in the barrels metric?
The two tables are presented very confusingly. It’s hard to tell from first or second glance whether the hitter or pitcher rates are more stable.
I love your stuff, Ben, partly because your writing is so clear and accessible on even the murkiest of subjects, but I have to confess I did not follow the paragraph below.
“That aforementioned Tango data twist? It gives us a rough estimation tool. Based on how far each group regressed to the mean in Year Two relative to Year One, we can estimate a “ballast amount” – how many league-average batted balls you should add to a player’s Year One barrel rate to predict Year Two barrel rate. For pitchers, it’s large – 131 plate appearances, to be exact. For batters, it’s tiny: 15 plate appearances. That’s because pitchers “regressed” 78% of the way back to average (in an average sample of 37 batted balls per quadrant) as opposed to only 33% regression on 30 batted balls per quadrant for hitters.”
The lynchpin of my confusion is that I am not sure at all what you mean when you say to add (1) “league-average batted balls” to (2) “a player’s Year One barrel rate”, and then in your next sentence you switch instead to adding “plate appearances”. Those things seem like apples and oranges and mangoes to me (in part because you seem to be adding a number to a rate).
The concept is that since pitchers ‘regressed’ 78%, then we need to add PAs such that (regressionPAs)/(regressionPAs+battedBallsPerQuadrant)=.78.[131/(131+27)]=.78.
However, I’m not sure what ‘regressed 78% of the way back’ means? R2=.22?