Carl Edwards: Riding Spin to Success

Maybe you know who Carl Edwards Jr. is. Maybe you’re a Cubs fan, or remember him from prospect lists. Maybe you wondered who that 120-pound dude with the 95 mph fastball was one day and did some light googling. Maybe you, like me, have stared at this list of top four-seam fastball spin rates in the league so long you know by heart that he’s got the second-most spin in baseball.

Or maybe you don’t know who he is, and you’re just now getting acquainted. One of the first things you should maybe know about are his hands.

Maybe his fingers more than his hands. Check out how long they are on his fastball grip.


The relationship seems to scream out from that image: Edwards features one of baseball’s highest spin rate because he features some of baseball’s longest fingers. Ask him, though — as I did earlier in the year — and he just shrugs: “I don’t know. It’s just a four-seamer to me.”

Though it’s not definitive, the early returns suggest that the link between finger length and spin rate isn’t particularly strong. Kyle Boddy took a look at 16 pitchers that trained at his facility and the data was not helpful. You usually want a p-value of under .01 or less, and his showed p-values in the .5-.6 range:

Fun-stealer is right. Imagine scouts trying to steal peeks at finger length to gain an advantage for their evaluations. Or, even crazier, imagine baseball’s version of the combine, in which height, weight, and finger-length measurements are published for every draft-eligible prospect.

Dystopias aside, it’s possible Edwards gave me some clue to how he gets the spin when he talked about the keys to his delivery. “Stand tall and don’t collapse too quick,” he said of his work in bullpens. “Stay on top of the ball,” he emphasized.

“[To] keep those breaking balls down” is his reasoning for those mechanical cues — and that’s important to Edwards’ success, of course. His breaking ball features above-average spin (2466, 2308 is average), velocity (81 mph, 78 is average) and drop for curves over 80 mph (-6.5, -3.9 is average). So he’s done a good job of staying on top of his breaking ball.

But maybe staying on top of the ball also leads to a high spin fastball? We already know that going over the top is a good way to create ride on the fastball, and that Rich Hill went over the top to make the most of his high-spin curve and fastball combo, so could arm slot also be the way to a higher spin rate?

Unfortunately, no. Despite a larger sample (480), “true” (height-adjusted) vertical release point has no relationship to four-seam spin rate (p value = .455). The graph doesn’t even show a relationship if you squint really hard.

Just as weird for Edwards is the movement on his fastball. There’s only one four-seam fastball that’s “straighter” than his, or closer to that mythical “zero” horizontal movement. Sonny Gray gets one-tenth of an inch of movement towards his arm side, and Edwards gets five-tenths of an inch of movement towards his glove side. Some weird, weird four-seam fastballs right there.

Of course, the further a pitcher deviates from the norm, the more interesting (and probably difficult) his pitches becomes. A glance at an example of Edwards’ fastball reveals that reacts to the pitch as though it will have some fade towards his arm side — most fastballs do, if not cutters. Instead, he whiffs on the inside of this straight pitch.

But Edwards didn’t call it a cut fastball, and while he admitted his grip was a little off kilter, he still thought of it as a four-seamer. He’s heard it’s weird — “I’ve heard it, and I look at video, I look at my mechanics, I don’t look at my pitches. Once the pitch leaves my had, there’s nothing I can do.” — but he doesn’t do anything to make the pitch weird.

When it comes to taking advantage of the weird movement, he’s just as intuitive. “Coach says to change eye level,” he grinned, “but my whole approach is to go right at them.” That seems to mean throwing high and away no matter which hand the batter is using:


By movement, especially when compared to his straight four-seamer, Edwards has a good changeup. He may yet start for this team in the future. But, given his injuries, and the team’s timetable, they decided he would relieve this year. The decision was “more about getting me here,” Edwards said. “As of now, this is my job, to come out the pen and get people out.”

So there may yet be time to be more nuanced than high fastballs and low curveballs. There’s a lot to that combination, anyway. Enough for now. For now, Edwards will keep his mind clear. “I was taught by my dad how to pitch,” he said quietly, “so I don’t really think too much about the different parts about what my arm is doing. I just basically hit the target. Just get ’em out, whatever gets ’em out.”

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.

newest oldest most voted

Kyle’s regression analysis is potentially faulty. The hand size, middle finger size, and index finger size are likely highly correlated. I would not recommend conducting regression with correlated inputs. See the wiki for collineararity, or multicollinearity, for more info. Doing so creates regressions that are potentially meaningless on out-of-sample populations, which is essentially what you and Kyle are trying to do, extrapolate the learning to evaluate the broader prospect universe.

The fact that the regression has a positive coefficient on index finger length and negative coefficient on middle finger length further suggests something erroneous may be going on. I would recommend starting with univariates (does a single input predict X) first, before dumping 3 inputs in.

PS. I would not require a p-value of .01 or less on a regression like this with 16 observations. A p-value of .1 would certainly be exciting evidence suggesting further study is warranted. The 16 pitchers in the sample may not necessarily be representative of pitchers in general.


good post–what were the correlations between the predictors and the outcome here out of curiosity.