The Knuckleball Is More Conventional Than It Seems
We often treat knuckleball pitchers as if they were members of some long-forgotten sect, practicing their secret ninjutsu on the rest of the league with a pitch that defies gravity and cannot be classified. That’s fine, the knuckler is the rarest pitch in baseball, and it has its iconic moments. Let’s not begrudge anyone a little fun.
But once you peal back the layers on the pitch, you start to see that each truism about the knuckler isn’t necessarily true. In fact, there are probably more ways in which the art of throwing a knuckleball is similar to the art of throwing other pitches than it is different. At least, that is, in terms of strategy and outcomes. Mechanics are obviously a different story.
Let’s unpack some of the things we might hear about knuckleballs, and then us the data and the words of R.A. Dickey and Steven Wright to guide our analysis.
Velocity doesn’t matter.
Maybe this isn’t a thing that’s said a ton, but nobody breathlessly reports knuckleball velocity readings the way they do fastball readings, so at least implicitly we’ve decided that speed doesn’t matter as much with the floating butterfly.
But we might be wrong. Our first clue was that R.A. Dickey got better with better velocity. Our second might be this look at different buckets of knuckler velocity and their outcomes.
| Velocity Bucket | Swinging-Strike Rate |
| >77 | 13.1% |
| 72-77 | 9.1% |
| 68-72 | 7.4% |
| <68 | 8.8% |
Looks like more velocity means more whiffs on the knuckler, just like with other pitches.
You can’t command a knuckleball.
Dickey felt like throwing a knuckler was like letting the butterfly loose, so it seems that even knuckleball pitchers don’t think they can command the pitch. Sorta.
The Blue Jays starter felt that his hard knuckler had different movement, lesser movement. The Red Sox knuckleballer du jour, Steven Wright, agreed recently. “It’s a little bit sharper, yes, moves a little bit less,” he said before a game with the Athletics. “The biggest thing is to throw it for strikes.”
Steven Wright’s knuckleball grip. Imagine trying to command that.
So Wright has a method by which he attempts to throw the knuckler for strikes. That’s one way to show command. Another is to continue to vary the pitch. “I change the speeds on the knuckleball. Hard, soft, medium,” Wright said. “I try to work in a mid-range and then I can go a little bit more on top and then a little below that speed, too. Gives me three different pitches instead of one. I think I’ve changed speeds more than I have in the past.”
That matches up with the data, shown below. Wright’s knuckler is harder at the median point, but he’s also extended the upper range while keeping many of the slow knucklers in his arsenal.
Wright may not be able to control location as much as other pitchers, but he can command the velocity. And the velocity determines the movement, which determines the location. So Wright can command the knuckler. Sorta.
If it’s high, let it fly.
This is a truism preached by batters who are facing the knuckler — and it’s true. But! It’s also true of all pitches. High pitches of all sorts are hit for home runs. High knucklers are, too, but not in any pattern that discerns them from other knuckleballs.
Check out all pitches slugging rates from righties versus righties by location (left) versus all knuckleballs (middle) and Wright’s knuckleballs this year (right).
In order to argue that high knuckleballs are more prone to home runs, you’d have to zero in on that one little quadrant in the top of the knuckleball circle. You’d have to ignore that Wright’s heat map looks nothing like that.
And you’d have to ignore Wright’s words. The high knuckleball? “If it’s thrown well, it’s one of the hardest ones to hit,” he said. And it’s unavoidable, anyway. “You can’t aim too low, because what happens if it drops. You have to start it high. You can get away with it.”
For the most part, Wright has gotten away with those high knuckleballs. Just not when they drift to the outside corner to a righty.
Knuckleballers are homer prone.
Since 1974, knuckleballers have given up .923 home runs per nine innings. Since 1974, all pitchers have given up .923 home runs per nine innings.
Everybody has a knuckleball.
This is a tough one, because if you ask around any clubhouse, they’re all working on their knuckleball. They’ve all “got one.” But they don’t.
“A lot of guys can do it in the bullpen, but then to go out in the game and actually throw it, it’s not an easy task to do,” laughed Wright. And that’s the ticket to all the new pitches we hear about in Spring Training that don’t make it to the big-league games. They look good in the bullpen but the pitchers don’t believe in them enough to throw them.
So everyone has a knuckleball. And nobody has a knuckleball.
Knuckleball pitchers are erratic.
I asked Wright about the fact that it took him so long to get here. And it took Dickey until he’d lost his chance as a conventional pitcher before he made it. Why does it take so long for knuckleballers to make it in the bigs?
“You can go out there one day and look like the best pitcher in the game, and then go out the next day and you’re throwing batting practice,” Wright suggested. “So there’s a trust issue with it.”
But should there be? Are knuckleball pitchers really more volatile? To answer that question, I got the help of FanGraphs’ Jeff Zimmerman. We looked at the distribution of runs allowed per nine innings for conventional pitchers and for knuckleball pitchers. We then bucketed the conventional pitchers in case the bad pitchers were keeping the whole group down.
| Ave RA/9IP | Standard Deviation | |
| Bottom Third | 5.94 | 1.10 |
| Middle Third | 4.89 | 1.03 |
| Top Third | 4.10 | 0.88 |
| Knucklers | 4.72 | 0.87 |
Standard Deviation = Spread in RA/9IP from start to start for starting pitchers
Since 1974
n = 1391 for conventional, 12 for knuckleball pitchers
Yes, knuckleball pitchers allow slightly more runs than the best bucket of conventional pitchers. But would you look at that? They’re more reliable than any other type of pitcher. If you’ve got a good offense, you’ll take a guy that can give up four runs per nine innings, every night out. And that’s a good definition of a knuckleballer.
It doesn’t move like any other pitch.
See this knuckleball by Wright.
Pick your jaw up. That looks crazy, and maybe it is. But there are crazy pitches of all sorts, even ones that were supposed to be normal like this Freddy Garcia crazy mystery pitch.
Turns out, most knuckleballs actually follow normal pitch movements. Knuckleballs have similar trajectories to other pitches. In other words, knuckleballs move with a “smooth tunnel” that isn’t very different from regular pitches, or at least that’s what baseball physicist Alan Nathan has found. Instead, the knuckleball’s brilliance is the randomness from pitch to pitch — the unpredictability of each pitch compared to the last one. You might bet a cutter or a curve or a change or a splitter, but it will be something you’ve seen before, most of the time.
Of course, there are things that are different for the knuckler.
Nobody teaches it, or nobody cares. “There’s not a lot of guys teaching it,” Wright pointed out. Maybe it’s still a trust factor thing. “Teams aren’t willing to let a guy go out there, wether it’s in the minor leagues, or especially in the major leagues, and say ‘go figure it out.'”
The mechanics are slightly different. “Slow everything down, and stay tall,” Wright explained. “Before I was worried about trying to get extended and out in front. It’s a push, but also a throw. I’m trying to get everything more compact.” And Dickey pointed out that you don’t need to throw at 100% effort, too.
But listen to Wright talk about throwing a knuckleball well, and it sounds like every other pitcher. “If you throw it with more conviction, you get more movement,” he said. “More movement means less control in the strike zone, but if you get them into swing mode by getting ahead, then I can stay aggressive.”
The knuckleball is not a pitch that anyone can master. “It’s hard to do, it’s not easy to do,” Wright laughed. But that doesn’t mean it’s unlike any other pitch. It’s a little bit like every pitch, actually. You just don’t know which one it will be, and that’s the genius of it.
With a phone full of pictures of pitchers' fingers, strange beers, and his two toddler sons, Eno Sarris can be found at the ballpark or a brewery most days. Read him here, writing about the A's or Giants at The Athletic, or about beer at October. Follow him on Twitter @enosarris if you can handle the sandwiches and inanity.
This was great, thanks for writing!
Good timing on this article with Wright pitching today. I’m personally not sold on him. His 2016 BB/K and FB/GB ratios are almost identical to his 2014-15 numbers. He’s actually giving up more line drives this year and there is no drastic change in pop ups or hard hit rate. The notable exception is that he has an absurd 10 GIDP and only 3 HR this year. In slightly fewer innings last year, he had 2 GIDP and 12 HR. Translation: Last year he was giving up hard hit fly balls; This year, hard hit ground balls. With no noticeable spike in K rate or drop in FB%, his absurdly low HR rate is unsustainable. He’s overdue for a 3 HR game and what more likely team than the Orioles tonight. Only time will tell, but I suspect his ERA will end up in the low-mid 3s by season’s end.
I think everybody in Boston would be ecstatic if Wright ended a full season with a low-mid 3 ERA. This is a guy who wouldn’t even have been starting this year if not for Spring Training injuries.
That’s true. I’m in no way suggesting he is garbage. This article is great though because it highlights that you shouldn’t abandon advanced metrics just because a pitcher is a knuckleballer. Look at RA Dickey’s Cy Young season — he led the league in K’s and had fantastic peripherals. He didn’t weak ground ball or weak fly ball his way to a Cy Young. People are loading up on Wright big time in DFS even though his peripherals have extreme luck written all over them. Karabell named him his top pick of the day. Right now is the time to sell high…
The available stats for Wright vs the current Orioles lineup: .098//235/171/406 but after seeing your numbers maybe regression is in order (at least some would be likely with those gaudy numbers and small sample size). One thing I have noticed in his starts (and this is anecdotal) is that Wright seems to often start out shaky and get stronger the 2nd and 3rd time through the lineup.
Eno expounding on the knuckleball, I don’t think I’ve clicked on a link faster!
The MLB sortable player and team stats site needs to add vs knuckball to the pull-down menu.
I have watched Stephen Wright quite a bit this season. And, based on how much trouble his catchers have had actually *catching* his pitches, I certainly believe in “you can’t command the knuckleball” and ” it doesn’t move like other pitches.” If the knuckle ball *does* move like other pitches, why can’t catchers catch it?
Imagine, as a catcher, that you have no idea what kind of pitch is coming. It could be a curveball, or a sinker, or a slider, you don’t know. I think would make it very difficult to catch.
Okay. So, if I understand, sometimes it moves like a cutter, sometimes it moves like a curve, sometimes a slider, sometimes a change. But before it’s thrown, nobody knows which it will be.
So, individually, a single knuckleball moves like something else, right? But collectively, the set of all knuckleballs moves to its own beat.
Jason Varitek on catching Tim Wakefidld’s knuckleballs: “It’s never relaxing, that’s for sure. It’s like snowflakes – no two are ever alike.”
Joe Torre: “You don’t catch the knuckleball, you defend against it.”
The knuckleball is not as rare as the screwball, is it? The screwball is practically extinct.
Screwball not as rare as I thought. http://www.fangraphs.com/blogs/lets-invent-a-new-pitch/
There are two things that drive knuckleball success other than velocity and change of speed.
One, the quality of release. The pitch must have less than one full rotation to be effective. A perfectly thrown knuckleball has no rotation whatever–the one to Davis broke twice, first in on him and then away. Wright has a consistently high percentage of low rotation knuckleballs. Knucleballs that rotate less than once but do so clockwise or counterclockwise (without turning over) are always ineffective. All knuckleball pitchers throw one now and again.
Two, conventional control. Most knuckleballs that are not strikes were simply badly thrown in the first place and were never going to see the strike zone.
Wright is not pitching in luck. When he loses the fine touch of his release for a time, and he will, there won’t be enough luck in the world to help him. But I have seen all his games and the hitters are very aware he is not pitching in luck, which is pretty much what Showalter said tonight.
I don’t think that is quite right. My understanding is that the movement on a knuckleball is caused by differences in air pressure when air flows around the seams of the ball. A ball with absolutely no rotation would have a more consistent movement than a ball with a slight rotation and be easier to hit.
Zero rotation gives zero stability. That is the object of a knuckleball. A curve ball has as much spin as it is possible to put on the ball to create differences in air pressure. A curve ball without enough spin is more like a knucleball with too much–hung.
Dickey always said he wanted a quarter turn. It’s linked here.
Knuckleballs are chaotic. It’s impossible for the ball to maintain exactly no rotation through air.
I think he was referring to an ideal knuckleball, thrown by a perfectly spherical cow.
Are they consistent? The table of standard deviations doesn’t tell the story (although a histogram would).
Say we have a conventional pitcher who throws for a mean RA/9 of 4.8, with a standard deviation of 1.03 in his RA/9 outcomes, by game (and assume that RA/9 is distributed normally).
And then say we have knuckleball pitcher who throws good games and bad games: for half of his games he throws at a mean RA/9 of 3.8, and for half of his games he throws at a mean RA/9 of 5.8. Within each type of game we’ll say that the standard deviation of his outcomes is 0.28 (and RA/9 is distributed normally within each type).
Both pitchers have mean RA/9 of 4.8, and both have standard deviation of 1.03 for RA/9 over all their games. But they’re not the same pitcher.
(This example is probably unrealistic, as who has sd of only 0.28 for RA/9? But it might be interesting to look at the distribution of outcomes by game, or even by inning, for conventional pitchers versus knuckleballers.)
Velocity is important for a knuckle ball because it only Dances when you achieve a certain Velo. of course 78 mph is not challenging Velo for pro hitters but if you throw it too slowly you still get a pretty laminar air flow. if you throw harder the air flow gets more turbulent and the ball starts to move more unpredictable.
that is also one of the reasons why knucklers don’t work in rec ball (apart from many not being able to throw it). every kid has experimented with a knuckler but a 45 mph knuckler just doesn’t Dance.
Well, also, humans can totally hit something moving erratically and slowly. Erratic movement needs to happening quickly enough for it to be difficult to track. Watching Dickey throw 80+ knucklers for the Mets was like watching the old Kerry Wood power-curve. How do you even track it?