June 13, 2012; St. Petersburg, FL, USA; New York Mets starting pitcher R.A. Dickey (43) throws a pitch in the fifth inning against the Tampa Bay Rays at Tropicana Field. The Mets won 9-1. Mandatory Credit: Kim Klement-US PRESSWIRE

Having Some Fun With Knuckleballer Pitch F/X


R.A. Dickey has taken baseball, including our very own Tampa Bay Rays, by storm this season. Let’s make some fun graphs and get some insight into just how crazy a pitch a knuckleball really is. The data is from Brooks Baseball and the graphs are my own. Let’s see what’s going on here.

(An explanation of how to read these graphs and the general topic of Pitch F/X is here.)

This graph gives you an idea of some of the variability from year to year. Some years, it averaged movement away from right batters and others it moved towards righty batters slightly. Some years it moved down vertically more, others less. One big thing to notice is the release points, where the lines start. Dickey was messing around with the release point from 2008 to 2009 before finally finding a more consistent release point from 2010 to 2012. From 2008 to 2009, Dickey had a 4.99 ERA and a 5.15 FIP compared to a 2.86 ERA and 3.63 FIP since. Coincidence or not?

And then there were the outliers: pitches that were really knuckleballs but moved so crazily that they were identified as other pitches. Here are some of those.

Part of the reason that the seasonal knuckleballs above had such little horizontal movement is because knuckleballs move in both directions and the movement cancels out at least to some extent. Courtesy of Brooks, here is a graph showing the vertical and horizontal movement on each of Dickey’s knuckleballs over the course of his career.

The yellow points are knuckleballs. As you can see, the movement is all over the place, moving in both directions horizontally and varying quite a bit vertically as well. A lot of that has to do with the velocity of Dickey’s knuckleball, which we’ll talk more about below.

For the “changeups,” we see some crazy run away from right-handed batters. And then there’s the “curveball”- a couple of knuckleballs that went straight down, a five-foot drop. Wow. The “curveball” was more of a soft knuckleball, coming in at just under 65 MPH on average. Dickey throws his knuckleball so hard most of the time, so taking that much velocity off his signature pitch had to make batters wonder what they had just seen.

Now let’s zero in on the 2012 data. In order to attempt to get some actual insight, let’s split the knuckleballs by count. So you can actually see a little bit of the separation between the lines, we’ll do this in a bunch of graphs rather than just one.

This graph doesn’t appear to show very much, but look at the key: as the count progressed from 0-0 to 0-1 and 0-1 to 0-2, Dickey added more  velocity onto his knuckleball. Why would he do that? Because the harder a pitch it is, the more Dickey can command it- at least to some extent. Especially when you’re ahead in the count, you really don’t want to hang a pitch and allow a demoralizing hit or more.

Same pattern here as we see Dickey run his knuckler up to just short of 80 MPH on average on 1-2. Notice that these pitches moved a little bit more horizontally towards a right-handed batter than the ones above.

This graph is where things get interesting. Dickey threw his knuckleball just 54% of the time on 2-0, but when he did throw it, it was the softest pitch and with the most movement that we’ve seen so far. We see more of the same in terms of velocity increasing, but there’s a lot more variation in terms of movement with the 2-ball counts.

On 3-0, we see that Dickey rarely went with knuckleballs, and when he did, he went with the hardest ones we have seen yet, hitting the low-80′s with armside run. On 3-1, he completely shifted gears when he did throw the knuckler, throwing a mid-70′s offering with quite a bit of horizontal movement away from right-handed batters. And on 3-2, he went with a little bit of a slower offering than his usual 2-strike pitch with as much as ever variability. He just let it go and watched to see what would happen with it.

To look at Dickey’s 2012 knuckleball a different way, let’s look at it separated by the number of the balls in the at-bat, but we’ll do it in one graph this time.

The graph here is pretty inconclusive, but the interesting thing is velocity. With 1-ball counts, Dickey threw his knuckelball, presumably with the most control, not wanting to leave a pitch up in the zone. He also lowered his arm slot slightly on average with 1-ball counts as he tried to guide the ball a little bit more down in the zone. With 1-ball and 2-ball counts, Dickey was willing to take a little velocity off to maybe get some nice movement and force weak contact to get out of at-bats or at least to change the tone of the at-bat. And with 3-ball counts, Dickey hopes to get hitters to swing and put the ball in play and basically end at-bats, whether by contact or even pitches out of the zone, or to change the tone of the at-bat with a swing-and-miss or a weak foul ball. The bigger movement helped insure that if contact was made, the contact would almost definitely be on the ground- Dickey has not allowed a single outfield flyball on a 3-ball count knuckleball this season.

All in all, Dickey doesn’t get nearly as much movement as the conventional knuckleballers got. Look at this comparison between Dickey’s average knuckleball movement and Tim Wakefield‘s (I used Dickey’s 2012 knuckleball movement because as we saw in the first graph above, the movement on Dickey’s knuckleball has changed over the years, while Wakefield’s is from 2008 to 2011.)

 

We see that Tim Wakefield’s knuckleball averaged over a foot more vertical drop than Dickey’s has even in his best season, 2012. But the clear difference is velocity as Dickey throws his knuckleball more than 10 MPH fastball than Wakefield did on average. Wakefield threw only a handful of knuckleballs since 2008 as fast as Dickey’s have been consistently his entire career. How does this manifest itself in terms of performance? Looking at career numbers since 2008, Wakefield actually threw more of his knuckleballs for strikes than Dickey, 65.7% compared to 63.7%. The increased velocity does not seem to help Dickey all so much in terms on command. But the big difference is control. Wakefield forced just a 1.13 to 1 groundball to flyball ratio on his knuckleball as he consistently lost command of it and left it up in the zone for flyballs. Dickey, meanwhile, has managed a much better 2.07 to 1 groundball to flyball ratio, keeping the ball on the ground and although missing the zone, missing more down than up.

The knuckleball is a crazy pitch. You let go of it and you can’t be sure what’s going to happen from the time the pitcher releases it from his hand to the time it arrives at the catcher’s glove. What R.A. Dickey has done has been to keep the variability, keep himself and especially the hitter guessing, but exert some type of command over what’s going to happen. He has managed to keep the unknown of horizontal movement, releasing his knuckleballs not knowing whether they’ll move towards the batter or away, but he has kept his knuckleball velocity faster and varied within that in order to get a better grasp of the pitch’s vertical movement. The result of that has been more swings-and-misses and more weak contact than any knuckleballer has ever induced before. Can he keep this up? Can R.A. Dickey go down as one of the great knuckleballers baseball has ever seen? He very well may. But he’s not brushing the dirt off a pitch of his forebears in Major League Baseball- he is innovating, contributing his own take to a baseball tradition.

Next Rays Game View full schedule »

Tags: New York Mets Pitch F/X R.A. Dickey Tim Wakefield

  • Pingback: Great Tampa Bay Rays Links For June 24, 2012 | Tampa Bay Rays Dugout Online | Tampa Bay Rays Blog

  • halejon

    I love pitch f/x and am glad to see someone taking a stab at Dickey, but I think some of this is a little misleading. Especially the graphs that show the smooth arcs of the ball and the talk about more or less “movement”. Pitch f/x isn’t made to deal with a pitch that can change it’s movement, acceleration, etc, multiple times on the way to the plate, since it’s just looking at the release point and where it crosses the plate and then applying equations to extrapolate the path. If Dickey threw a knuckleball that broke a foot left and then back 11 inches right, and another one that went 1 inch left, it would dilligently report them as having the exact same “movement”, since it’s just about comparing the final position to where a “straight” pitch would have gone (also, the projected arcs would be the same).
     
    It’s a technique that works great for every other kind of pitch that has standard acceleration, but for a knuckler the whole point is the somewhat-random back and forth fluttering it does from release to crossing the plate. I would strongly suspect that the dominant factor that has made Dickey the best pitcher on the planet for his last 5 stars is that the standard deviation of that wobble has gone way up, but there’s no way to get a handle on it with this data. If you had never seen a real knuckleball and read this article, you would come away with the idea that it’s what it looks like on TV – a really slow pitch that drops and maybe moves a little towards the hitter. So much more to it. 

  • http://rayscoloredglasses.com/ RCG_Robbie

     @halejon The big problem is the inherent limitations of using average movement. The knuckleballs move all over the place and the average movement shows so little horizontally because the movement in both directions (towards and away from the hitter) basically cancels out. I agree that Pitch F/X can’t accurately display a knuckle ball but I just wanted to see what Pitch F/X could actually tell us.
    I hope that people reading this article wouldn’t just look at the graphs because without the explanations involved, the graphs are truly misleading. 

  • halejon

     @RCG_Robbie  Right…sorry if sounded really negative, that’s just something I think most people don’t quite understand about the system, and the knuckle. The rest of what pitch f/x CAN tell us is interesting, and well done. 
     
    (Incidentally, average movement is problematic for other pitches too, eh? A slow change that goes up and comes back down and a hard splitter get called the same pitch all the time unless you look at the ‘Breaklength’ variable or something…Funny how hard it is to describe 3D movement with 1D numbers…;))

  • http://rayscoloredglasses.com/ RCG_Robbie

     @halejon See the misidentified knuckleballs graph for some prime examples of that. 
     
    But I do like average movement overall because every pitcher works to both sides of the plate with their pitches and the average movement (along with frequency and velocity) is a way to compare different pitchers. I’ve done a lot of work with a fun variable called WVANM (weighted velocitized absolute net movement) which has an interesting association with pitcher performance.