Knuckleballs and Grounders

News out of the Red Sox camp (h/t Neyer) is that the Red Sox might finally part ways with Tim Wakefield. The season is still is ways off, so there is time for injuries to come up and Wakefield find a way into the rotation. But if this is the end of the line for Wakefield with the Red Sox — only seven wins shy of 200 and 13 away from the franchise record — it would be too bad. For one thing, last year was a rare one when two knuckleballers got a substantial number of innings: Wakefield and R.A. Dickey.

Looking at the two pitchers’ numbers I was struck by their very different ground-ball rates, 55% for Dickey versus just 37% for Wakefield. My main frame of reference for a knuckleballer has been Tim Wakefield, so l always assumed that there was something about the knuckleball which led to lots of fly balls. But with Dickey’s high ground-ball rate maybe it is just Wakefield’s knuckleball.

The most likely culprit with differences in ground-ball rates is pitch height. So here, on the left, are histograms of knuckleball height for the two pitchers. On the right is ground-ball rate by knuckleball height.

So, yeah, Dickey’s knuckleballs are a little bit lower in the zone than Wakefield’s. But that alone is not enough to account for the difference: no matter the height Dickey’s knuckleballs get more grounders than Wakefield’s.

Looking elsewhere, the big difference between the two pitches is that Dickey’s is about 10 mph faster: averaging 76 mph versus 66 mph for Wakefield. It looks like this plays a big part in the difference between the ground-ball rates:

For the 65-70mph range where they both throw knuckleballs (though Dickey rarely and mostly earlier in his career) they get roughly the same ground-ball rate. But once Dickey’s knuckleballs get up to the mid-70s they get tons of grounders. It seems the additional speed on Dickey’s knuckleballs don’t lead to any more whiffs (whiff rate on Dickey’s knuckleballs in 8.2% compared to 8.4% for Wakefield), but rather more ground balls.

Having two knuckleballers gives a nice opportunity to compare what is the case about knuckleballs generally versus what is unique to specific pitchers. I personally hope that Wakefield finds a way to stick around for 2011.

Quick note: Garik16 has a great three part series on the knuckle ball that looks at both Wakefield and Dickey. Part 1. Part 2. Part 3.

Print This Post

Dave Allen's other baseball work can be found at Baseball Analysts.

44 Responses to “Knuckleballs and Grounders”

You can follow any responses to this entry through the RSS 2.0 feed.
  1. Millsy says:

    Interesting. Mid-70’s (to 80, WOW) are no small feat, especially considering his fastball velocity is only around 84 mph (88-89 when he wasn’t a knuckleballer, I guess).

    From personal experience, my fastball was only in that low-to-mid 80’s range and I can’t imagine ever being able to put a knuckle grip on the ball and throw it anywhere near 70, let alone 80.

    I’d be very curious to see some stuff about his actual grip on the ball.

    Vote -1 Vote +1

  2. Telo says:

    Wow. That last graph really nails it home. Cool piece.

    Vote -1 Vote +1

    • Telo says:

      Though, I can’t help but think that, on average, knucklers with higher velocity are going to break more, and not knowing the answer, I would guess that “down” is the primary direction for a knuckler to break.

      Therefore more speed = more drop (on average, we all know knucklers can go more or less in any direction), which clearly would lead to more grounders. Maybe not so surprising then?

      Vote -1 Vote +1

      • garik16 says:

        Not quite. There’s no difference between the velo really and range of movement shown by the pitch. Seems virtually identical.

        The change is entirely speed based.

        Vote -1 Vote +1

      • Telo says:

        Yea, that early theory/intuition was wrong.

        Vote -1 Vote +1

      • Guy says:

        If the break of the pitch is identical for the two speeds then it makes sense that the faster would be harder to hit, less time to register it. Plus Dickey’s got a good fastball for a knuckleballer to keep the hitters off balance. I always thought that the slower knuckler had more lateral movement.

        Vote -1 Vote +1

      • eastsider says:

        @Garik16, Dave Allen, others

        First off, I love this post.

        So, I looked at the 2010 pfx_x and pfx_z graphs on Dickey’s and Wakefield’s Fangraphs pages and came to the same conclusion as you – that there is no difference in horizontal or vertical movement as a result of speed. Of course, I couldn’t leave that alone, so I looked at pitchfx data (my first “analysis”!). Indeed, there is no correlation for Dickey – r2 values of 0.0017 and .000097 for pfx_x and pfx_z, respectively (only looked at Dickey). So far so good.

        But (you knew that was coming) – if you look at break_length there is a bit of a correlation, 0.304 (the graph looks more impressive). I’m new to this and looked at what define break length as – “largest deviation from a straight line between the release point and the point at which the ball crosses the plate.” It seems that it might be possible that the net effect of the vertical and horizontal break washes out with speed, but that it doesn’t mean that overall movement isn’t affected.

        pitchfx data from Did not do any quality checks – simply grabbed all pitches of type “KN” pitched by Dickey.

        Vote -1 Vote +1

  3. Telo says:

    What would be interesting to see is the plot of where Wakefield knucklers were caught by the catcher, in relation it’s projected path, then compared to Dickey.

    My guess is that Wakefield has more randomly distributed movement, whereas Dickey has more consistent sink, over other directions, which leads to his higher GB rate. And when these respective guys throw harder, they get more break, therefore more grounders.

    Vote -1 Vote +1

  4. Saul says:

    Great piece, Dave. As a Mets fan, it’s exciting to see some evidence on his spontaneous success, especially how it relates to a known, proven knuckleballer. Thanks for this.

    Vote -1 Vote +1

    • Telo says:

      No offense, but it’s not evidence of anything. He’s been throwing the knuckler that hard for 3 years, and just this year seemed to get about %8 more GB.

      He was fortunate on BIP and HR/FB. Even assuming his gains in the BB/9 and K/9 as a knuckleballer are real, and he posts something like 5.4 K 2.5 BB/9, if his BIP and HR% come back to reality, he’s going to be a 4.25ish FIP pitcher. Not useless, but nothing close to what his ERA might lead you to believe he was last year.

      Vote -1 Vote +1

      • Telo says:

        E: it looks like he did throw MORE knucklers last year, which would perfectly align with the increase of GB. So you could presume his GB rate is close to reality. However, that doesn’t change anything else in my post… he still outperformed his expected BIP and HR/FB rates significantly. Oh, and he left a lot of guys on base. He’s not very good.

        Vote -1 Vote +1

      • RC says:

        FIP isn’t a very good stat.

        Vote -1 Vote +1

      • Saul says:

        He pitches in CitiField, so I personally wouldn’t be so fast to say that his HR/FB rate is going to balloon… Sure, I’ll take into account that his style’s been seen and scouted, but I’m not expecting him to be a total failure. He threw a career high 174 IP, so the sample size is small. But even if you’d compare it to 2008, when he tossed 112 for the Mariners at Safeco, his HR/FB was 10.8%… Park Factors figures Citi holds the ball marginally better than Safeco anyway, and I’m faithful that he’s starting to hone his craft.

        Vote -1 Vote +1

      • garik16 says:

        Telo, the increaes in GB rate comes entirely from the fact that he threw the fast knuckler at a much greater rate than previously. Go check out the part 2 article linked in this article and you’ll see it’s pretty clear.

        The GB rate on the pitch is pretty stable each year if you control for velocity. He just threw the fast one more often last year.

        Vote -1 Vote +1

      • Millsy says:

        I think it’s a stretch to say he’s been throwing the knuckle “that hard for 3 years”. From 2005 to 2010, average speed at least looks like:


        So, over the past 3 years, he’s actually increased the average velocity of the knuckle by almost 3 mph. Over the past 5, it’s approaching 10 mph increase. That’s a lot in my book.

        Doesn’t mean he’ll repeat 2010. But it seems reasonable to say that faster pitches (at least for knuckleballs) produce more GBs. I’m not sure the article says anything specifically about ERA regression in 2011.

        Vote -1 Vote +1

      • Millsy says:


        2005 and 2006 don’t really matter, as it was a total of 32 innings.

        Vote -1 Vote +1

      • Telo says:

        @ Garik

        Yea, I caught that and addressed it in my second post. It still doesn’t change the fact he was luck on BIP, HR, and LOB.


        There’s not nearly enough data to project his HR/FB anywhere below the league average.


        So, he finished up last year throwing his KN 76 MPH. On the graph above that represents about a 52% GB rate. Considering he was above that mark last year at 55%, maybe we should expect his GB rate to actually come down. Regardless, his “luck peripherals” (BIP, LOB, HR/FB) are unsustainable, even if you believe he has some small amount of control over one or more of them.



        Vote -1 Vote +1

      • garik16 says:

        “So, he finished up last year throwing his KN 76 MPH. On the graph above that represents about a 52% GB rate. Considering he was above that mark last year at 55%, maybe we should expect his GB rate to actually come down.

        This is incorrect. This would be correct if the balls in play came entirely on knuckleballs at random speeds.

        But Dickey throws the fast knuckleball more frequently (drastically) on counts where batters are more likely to swing. The end result is that 49.8% of balls in play against Dickey were against knuckleballs over 76 MPH, often much over. 34.1% were against slow knucklers below this rate.

        Then there’s the 16% off fastballs….which had a 55% GB rate (it’s got massive sink, even if it’s not fast).

        “Regardless, his “luck peripherals” (BIP, LOB, HR/FB) are unsustainable, even if you believe he has some small amount of control over one or more of them.”

        Well first of all, knuckleballers are known for lower than usual BABIP results. So that’s somewhat explainable (his BABIP in 2 years as a knuckler comes out to below .300, though I haven’t added it up). There might be some regression there, but it’s incredibly hard to predict.

        The LOB% also is very possibly sustainable. 77% isn’t amazingly high (and is sustainable by good pitchers), especially when you consider as a knuckler his BABIP should be lower than we’d otherwise expect, it’s quite possible that he can equal that rate. It also helps he pitches half his games at Citifield, which suppresses HRs.

        And of course, the HR/FB rate isn’t really that low.

        Predicting knucklers is insanely difficult, and regression ought to be expected. But nowhere near as bad as you think it should be.

        Vote -1 Vote +1

      • Telo says:

        But Dickey throws the fast knuckleball more frequently (drastically) on counts where batters are more likely to swing. The end result is that 49.8% of balls in play against Dickey were against knuckleballs over 76 MPH, often much over. 34.1% were against slow knucklers below this rate.

        Interesting. Where are you getting those numbers from? Show me how you would project his GB% rate going forward.

        RE: BABIP

        When you have 400 IP of data, you can’t ignore it. He looks (eyeballed) to be around .285 as a knuckleballer. Wake is obviously the only other comparable, and we’ve determined his is much more a FB, so he is inherently going to have a lower true BABIP. .285 would be the LOWEST you could comfortably project Dickey’s BABIP.


        LOB% is function of skill, it’s not a skill itself. He can’t have a “knack” for stranding guys (tell that to 2008 Dice-K). Dickey is a middling enough type of guy that his true LOB is probably around 72%. He doesn’t K enough guys for it to be any higher.

        RE: HR/FB

        Last year he gave up 30% less HR than his career average, and 20% less than his career as knuckleballer (which is consequently approx league avg.) He will very likely regress there as well.

        Vote -1 Vote +1

      • kick me in the GO NATS says:

        I thought I once read that FIP did not relate to Knuckle ball pitchers. They were the one exception.

        Vote -1 Vote +1

      • garik16 says:

        @Telo: The numbers are from my pitchf/x work on Dickey, a piece of which is linked in the above article.

        Re: BABIP: .285 sounds about right. Obvious caveat being that his knuckler changed drastically during the last 3 years, but .285 sounds very reasonable.

        Regarding LOB%, I still disagree. His strand rate isn’t amazingly high, and while he doesn’t have a huge K rate, I think it’s high enough in combination with his lower KN babip and GB rate (suppressing HRs at the least) to maintain.

        RE: HR/FB
        Last year he gave up 30% less HR than his career average, and 20% less than his career as knuckleballer (which is consequently approx league avg.) He will very likely regress there as well.”

        Except his knuckler is clearly different (faster one used more frequently, and it does have more sink on average, though I don’t actually think that’s much of a big deal to be honest), which allows for the possibility that the change is real.

        And of course, check out the HR/FB home/away splits. On the road, it’s normal (11.5%). At home, it’s really low. (4.5%). Citifield is a major factor here that will remain the same.


        Listen, I’m not going to pretend to try and predict knuckleballers..the pitch is too inconsistent and many of our normal tenets of BABIP and such don’t apply as usual. But I do think you’re expecting too much of a regression. A regression of his ERA back to his xFIP in 2010 – 3.88 – is a more reasonable expectation.

        Vote -1 Vote +1

  5. MikeS says:

    Is there anybody that doesn’t love a knuckleballer? I think it’s because we all know we can’t snap off curves like Blyleven or throw as hard as Ryan but we see these guys throwing soft and making major league hitters look silly and we think “hey, maybe….”

    I seem to remember somebody, somewhere (maybe here) coming up with a law of conservation of knuckleballers. They are sort of like Vampire Slayers. There can be only one at a time and when one goes down, another gets called up to the big leagues to take his place. So clearly Wakefield must leave so Dickey can enter.

    Wilbur Wood was my favorite player growing up, probably because he was always pitching whenever my dad took me to Comiskey Park.

    Vote -1 Vote +1

    • Bob B. says:

      The ‘conservation of knuckleballers’ you mention makes me miss the days of two Niekros and Hough all throwing it! I to hope Wakefield can make a comeback from last year, I hope Dickey has continued success and I hope that Dodgers pitcher from last year can ‘make it’ as well!

      Vote -1 Vote +1

    • Bryz says:

      Which explains why guys like Charlie Zink and Charlie Haeger couldn’t stick it in the majors.

      Vote -1 Vote +1

  6. garik16 says:


    You can see the trend of velocity here:

    Vote -1 Vote +1

    • Choo says:

      Cool graph. It appears Dickey mastered two different knuckleballs in 2010, at least in terms of speed.

      Vote -1 Vote +1

  7. B N says:

    I’d be pretty surprised to see Wakefield not even be kept around as a longman/reliever/spot starter. In my opinion, he’s still got some value so unless the farm system is more chock full of ML ready talent, I’m hoping they’ll keep him around for one more turn. At 4m, it seems like a great insurance policy at the least.

    Vote -1 Vote +1

  8. Barkey Walker says:

    I LOVE these graphs. Are the error bars from loess (well, form predict after making a loess fit)? At any rate, the way they are plotted is really nice.

    There are lots of nice touches like making the names themselves be the colors of the lines, cuts down on clutter and is even easier to understand than putting the colored lines next to the names.

    Vote -1 Vote +1

    • Dave Allen says:

      Brakey, yeah those are standard errors from the prediction of the loess fit. Thanks for your comments about the graph.

      Vote -1 Vote +1

      • Barkey Walker says:

        The second plot is a great plot for making your point. Is it possible to test the joint significance of the whole loess fit? I’m not sure you would get it for Wakefield in either of the plots while it looks like you would for Dickey in both cases.

        Vote -1 Vote +1

  9. chris says:

    pitches move more the slower they’re thrown, not the other way around.

    Vote -1 Vote +1

  10. Choo says:

    The average knuckleball for any knuckler falls within an inch of horizontal axis point 0, plus or minus, Dickey included (0.1). However, Dickey gets way more sink (5.3) and way more run (-9.9) on his fastball than any of the other knucklers* and he throws it 15%-20% of the time, enough to make hitters aware of the pitch. So with his knuckleball at 0, and his fastball at -10, the difference is +10 which, when combined with the 6-10 mph differential, rates his knuckleball about the same as a roundhouse slider. Crazy!

    Of course that’s only on paper and any professional hitter worth his weight in sunflower seeds will identify a knuckleball in less time than it takes to identify a decent slider, but the reaction time (6-10 mph differential) / plate coverage (10” differential) ratio might create a slider-like challenge for hitters. Only those hitters who have faced Dickey know for sure, but nonetheless, it’s an unusual feature for an already unusual pitch. Add that to the unpredictable movement of the knuckleball, Dickey’s knack for making it sink and his ability to locate it low in the zone, not to mention a sneaky little fastball, and the high GB% makes sense.

    *No RHP of any ilk within 3 mph of Dickey’s fastball throws a comparable pitch. Dave Bush and his 86.6 mph 2-seamer was the closest RHP I could find, and the list of lefty comps features the funktastic trio of Jaime Moyer, R.J. Swindle, and Daniel Ray Herrera.

    Vote -1 Vote +1

  11. Mr.MojoRisin says:

    I sure did like watching RA pitch last year. He was just fun to watch throw! I live in Connecticut so I have seen Wakefield pitch many many times, and he and RA are very different in their approach. Wake seems to just push the ball up there while getting a lot more drop than RA. But RA throws his with less lob action in his motion and his ball finishes with a quicker, but smaller break at the end.

    RA’s pitching motion also creates better deception than Wake when they throw their fastball. I can’t count how many times opposing batters were shaking there heads after getting locked up on the fastball. But that also leads to my biggest fear for RA, is that he picked the right time to throw his fastball every time last year. What if he doesn’t do that this season.

    As a Met fan I hope he can mirror last season, but realistically I’m looking at a normal Wakefieldesque year of 11-10 -ish, a low 4.00 ERA, and 200 IP. I don’t think those numbers are that unreasonable.

    Vote -1 Vote +1

  12. Johnb says:

    Dave — great graphics. What function in R do you use to plot the shaded confidence intervals?

    Vote -1 Vote +1

  13. james wilson says:

    The hard knuckler often causes hitters to be jammed, where the slower knuckleball only causes him to look foolish. Perhaps jamming leads to more ground balls. Dickey, unlike Wakefield, can throw the pitch high in the zone to good effect.
    I’ve caught four knuckleballers. The softest tosser had an amazingly still ball, which gave the greatest break at the plate. The hardest thrower was actually throwing a no rotation forkball in the low eighties, which broke little but looked frightful and shimmied all the way in. It also ruined his elbow.

    Vote -1 Vote +1