A Look at Knuckleball Locations

In a previous article, I looked at the average locations of R.A. Dickey's knucklers and compared them to former Red Sox knuckler Tim Wakefield. I promised repeatedly to revisit the topic, and that is what I am doing here. I used Baseball Savant to gather all the knuckleballs thrown regardless of pitcher.

This first graph shows the average locations of every pitcher that threw a knuckleball since 2008 (I took out Brad Peacock, whose curveballs were being called knuckleballs for some reason by MLBAM, and while I don't know why, and I didn't realize it until I was almost done with the article, Josh Banks and Eddie Bonine were not included in the data set).

Ryan Franklin, who only occasionally mixed in a knuckleball, threw his the lowest on average and it is not that close. Dickey has the most glove side knuckleball on average. No lefties have thrown a knuckleball in the Pitch F/X era, not even any of the position players (there have been some released on the left side of the rubber by Dickey and Wakefield, Dickey more than Wakefield, but obviously these are at parks that measure release points slightly to the left than usual or they moved around on the rubber). When looking at results of all knucklers thrown, it is really difficult to find a difference between which knucklers turned into whiffs and which ones turned into balls put in play for the three standard results

The swinging strikes are higher and more glove side, but I don't know that it is significant enough for there to be a real difference.

Since I looked at Wakefield and Dickey in the previous article linked to above, I'll look at two pitchers who have a significant amount of knucklers thrown in the Pitch F/X era, both of them currently in the Boston Red Sox organization. Charlie Haeger hasn't pitched in the Majors since 2010, throwing 83 innings with the Dodgers and Padres. He is currently pitching in AAA, but struggling. His knuckler (114 wRC +) was not as effective as Dickey's (85 wRC +), or Wakefield's (98 wRC +). Here were his average result locations

As we saw in the first graph, no one liked to throw their knuckleballs higher in the zone than Haeger on average. The highest is where he gets his swinging strikes, with the lowest being balls put in play without any outs, so we can see why.

Steven Wright made his debut earlier this season, and while the sample size is small, no one has had a more successful knuckler according to whiff/swing % than him.

When he gets the ball glove side, he is more likely to get whiffs, but he gives up contact about equally glove side and arm side. When he keeps it arm side, it appears that hitters are less likely to swing.

I mentioned the position players who have thrown knuckleballs above (only 79 of the 20305 knucklers since 2008). The following graph shows where the position players located their knucklers on average compared to the standard pitchers

The pitchers were able to get the knuckler more glove side and higher in the zone on average, something that, if we are looking at trends, seems to be important.

Just for fun, here are where the individual position players' knuckleballs have been thrown. Of course, probably not all of them are knuckleballs, just slow pitches that MLBAM called knuckleballs.

Joe Mather's doesn't show up because he didn't throw it for a strike on average.

Slow Pitches in the Major Leagues

I have always been fascinated by the slow pitch in baseball, especially the Major Leagues. In a league where 95 MPH fastballs are not unusual, and velocity is important, some pitchers throw pitches in the 60s to unsuspecting batters. I wanted to look at why pitchers might do so and whether or not this is an effective strategy.

Not counting intentional balls or pitchouts, there have been 32323 pitches since 2008 tracked by Pitch F/X thrown under 70 MPH. The clear leader in slow pitches thrown is Tim Wakefield, as over 85 % of the pitches he threw from 2008 to the end of his career were below 70 MPH. Wakefield was well below average in effectiveness over that stretch, but was still an above replacement pitcher.

By far the most common pitch thrown under 70 MPH is the curveball. About 68 percent of pitches thrown under 70 MPH are curveballs (21982 times) and this doesn't include the 509 eephuses or the 231 knuckle curves.

The eephuses are slower at 59.27 MPH on average, while the knuckle curves under 70 average about 67.89 MPH and the curveballs 67.8 MPH. Strangely, there isn't much of a difference between the way the average eephus and slow knuckle curve is located, but the average slow curve is thrown much lower (the entire graph is the strike zone).

When you break the curveballs down by 5 MPH velocity ranges, we see that the slowest ones are located higher while the "harder" ones are located lower:

This explains the difference in eephus and slow curve locations, but not really the high knuckle curves. 8432 of the pitches under 70 MPH are knuckleballs, mostly from Wakefield. I won't look further at them because I plan on looking at knuckleball locations a little bit more in a future post.

So is the slow pitch more effective than a normal pitch? Here are how the slow pitches break down:

2980 whiffs 9.2 %

5954 contact plays 18.4 %

204 home runs .63 %

Compared to league average (just using 2013 for league average for our purposes), the slow pitches actually have a slightly smaller home run rate (by about .03 %) and a league average whiff rate. They also had a .2 % smaller contact rate. So the slow pitches seem to be slightly better than a league average pitch, but probably not significantly. The only thing you can complain about them is that they are harder to throw strikes. The slow pitches have been strikes 61.33 % of the time, while 2013 league average is 63.53 % (63.38 % for relievers, 63.61 % for starters).

What about pitchers that throw a lot of slow pitches? After all, throwing a slow pitch seems to be about deception and tricking the batter. If a pitcher throws a lot of slow pitches, then wouldn't the effect wear off?

This is where looking at Wakefield and others can be helpful. Unfortunately, Wakefield is the only one with a significant amount of pitches with over 50% slow pitches. In fact, only Daniel Herrera and Charlie Haeger are over 20 % with a significant amount of pitches.

With Wakefield, you knew what was coming. He was going to throw the knuckleball, and then another knuckleball, and then a few more. Here are his pitches under 70 MPH broken down by result:

8158 pitches

1771 contact plays (21.7 %)

73 home runs (.89 %)

718 whiffs (8.8 %)

Wakefield was below average across the board with his slow pitches, not surprising considering that he was a below average pitcher and featured the slow pitch heavily. Looking at the two most prominent non-knuckleball examples may be more helpful though.

Herrera:

418 pitches

107 contact plays (25.6 %)

2 home runs (.47 %)

62 whiffs (14.8 %)

I guess since it was a screwball, it isn't surprising to see weird results. He gave up a lot of contact, but at the same time got a lot of whiffs and still didn't give up home runs with it.

Justin Duchscherer:

420 pitches

70 contact plays (16.67 %)

52 whiffs (12.38 %)

3 home runs (.71 %)

The results here are almost as weird as Herrera's. Duchsherer gave up less contact than average with his slow pitches, getting more whiffs, but also giving up more homers (even though he pitched for Oakland, who plays in a pitcher friendly park).

Randy Wolf and Dave Bush are two more pitchers who featured slow curves more than 15 % of the time. Wolf had a well below average whiff rate with it, and a slightly lower than league average home run rate. Bush had an above average whiff rate and a slightly above average home run rate.

What I was really interested in was guys who have good stuff, but still throw the occasional slow pitch. Yu Darvish is probably the best example, getting up to 97 MPH with a fastball with one of the top sliders in baseball, and also throwing an occasional curveball in the 60s.

On a whole, Darvish is better than league average at limiting homers (.57 %), much better at getting whiffs (12.2 %), and gives up less contact than average (15 %). He has thrown 198 pitches under 70 MPH, all curves, about 3.55 % of pitches thrown. They still have an above average whiff rate (10.1 %), but is below his average. He has given up 3 homers with them, a well above league average rate, but gives up contact with it just 11.1 %, much less than league average or his average rate. So for him, it isn't a pitch he misses a lot of bats with, and he has gotten burned with it, but it is thrown mainly as a shock pitch, an attempt to get a called strike. It is a pitch on average, he throws arm side, and about equally high and low. He is trying to throw it for a strike intentionally, just to steal an occasional strike. Considering the rest of his pitches, it is risky.

Vicente Padilla is another example of a guy who throws hard but also occasionally throws a really slow pitch (and ironically, he went from the MLB to the NPB, where he still throws the slow pitch). Sometimes labelled as an eephus and sometimes labelled as a curve, it is more of an aggressively slow pitch than Darvish's, averaging 58.3 MPH versus Darvish's 67.1 MPH. The pitch had a laughable whiff rate, short of 3 percent, but he gave up contact with it two percent less than average and only gave up 1 homer in over 400 pitches with it. Perhaps even more dramatically than Darvish's Padilla's slow pitch was designed as a take pitch for a strike. Padilla locates his slow curve equally arm side as Darvish, but a little higher. It isn't supposed to generate swings, it is supposed to just fall into the strike zone while the hitter stares at it.

Zack Greinke's slow curve is a little different, as his home run rate with it is low, but his whiff and contact rates are high. It is designed to be swung at, and not surprisingly, it is thrown glove side and arm side about equally, and in the low part of the strike zone on average. It is thrown about .3 MPH harder than Darvish's, so it is a slow curve, but he throws it like a classic curve.

There are also some unique pitches such as Yoshi Tateyama and Herrera's screwball/changeups. Herrera located low in the zone on average about equally glove and arm side. Tateyama on the other hand, has both a screwball and a curveball that he throws under 70 MPH. His curve is located arm side and a little lower than the middle of the zone on average. It is clearly a take pitch, with well below average contact and whiff rates. His average screwball on the other hand, is located slightly lower, and so far arm side on average that it isn't in the strike zone. It is more of a contact pitch, getting hardly any whiffs, and a slightly higher than average contact rate.

I count seven other pitchers with two different pitch types thrown under 70 MPH at a significant amount. R.J. Swindle a slider in the high 60s with a curve in the 50s. Mike Mussina change and knuckle curve both in the high 60s. Donovan Hand has a few changeups in the high 60s to go with his high 60s curveball.
Charlie Haeger has a few high 60s curves to go with his knuckleball. R.A. Dickey is a similar story (he has three if you include the eephus and curve differential in the tags). Wakefield had a few 4-seam fastball tags under 70 MPH, along with his curveball and knuckleballs. Dallas Braden threw an occasional slow curve to go with his unusually slow change.

There were also some obvious errors, such as a 4-seam fastball by Frank Francisco and a moving fastball by John Danks. Some of the really slow changeups, including some thrown in the 40s, are really hard to believe (especially one thrown by Radhames Liz, someone who can throw 100 MPH with his fastball). 

So the slow pitch can be effective, and it can be used in different ways by different pitchers. It can be an interesting tool to have, but it doesn't guarantee whiffs. Personally, I thought the home run rates and the whiff rates would be higher than they were. 

Why Pitchers Struggle in the First Inning

It is no secret that starting pitchers struggle in the first inning compared to the later innings of the ballgame. So far in 2013, starting pitchers have a 4.31 ERA in the first inning, with an opposing OPS of .742. In the second inning, the OPS drops to .720, with an ERA of 3.74. Part of this probably has something to do with quality of hitter, though so far this year spots 4-6 in the order have an OPS of .752, while 1-3 in the order has an OPS of .750.

So I looked at the all pitches thrown in the 1st inning of 2013, along with all the pitches thrown in the 2nd inning of 2013. I took out all the unknown pitch types. I also ignored things like pitchouts, intentional walks, and the FA tag (though these last three things weren't removed until after, so they still played in the percentages). I sorted the pitches by MLBAM pitch type and looked at their average locations and velocities, comparing all the pitches thrown in the two innings to see if there was a discernible difference.

Here is the first inning graph, showing where each individual pitch was located on average, regardless of platoon and handedness:

By contrast, the second inning:

The average vertical release point is the same in total from inning one to inning two, so there doesn't appear to be anything obvious going on with pitchers' deliveries, or at least how they end.

From the first inning to the second inning, there is a drop in four seam fastball usage of nearly four percent, and there is an increase in slider usage of two percent. Moving fastball usage drops about a percent, while sinker usage stays about the same and cutter usage is the same. Changeups see an increased usage of about a percent, while curveballs saw an increase of over a percent.

As far as average location difference goes, the 4-seam fastball is located better in the 2nd inning, as it is closer to the center of the strike zone slightly (meaning it is thrown to both sides of the plate more on average), and slightly higher. It isn't a dramatic change, but with the large sample size, any small change is pretty significant. The cutter is located a little lower on average, while the slider is about the same, and the curve is actually thrown a little higher on average. The moving fastball is similar to the 4-seamer, in that it is located closer to the center of the plate on average and a little higher.

Dickey, the only one to throw knuckleballs in the first two innings of a ballgame this season, improved from the first to the second just like league average. The knuckle is actually a tick softer in the 2nd inning, and he locates it more arm side on average (though it is slightly higher).

Hector Santiago is the only one that throws a screwball, and he locates it, on average in the first two innings, outside of the strike zone, and that is why it doesn't show up in the graphs. He is actually a lot better than in the first inning than the second inning. In the second inning, his screwball is actually lower on average and closer to the strike zone.

To me, it seems that the biggest difference is fastball usage and fastball location. The fastball is the pitch most pitchers throw the most, but it usually isn't as effective as off-speed pitches. For example, curveballs this year have a 10.8 % swinging strike rate, and changeups have a 14.8 % swinging strike rate, both of these rates much higher than fastballs, so it would make sense that pitchers would have more success by throwing the fastball a little less and more breaking balls if they can locate them (of course, fastball velocity is still very important, and there is a point of no return, but there is no correlation between fastball usage and overall success, as qualified pitchers have a -.097 between fastball percentage and FIP -).

Talking About R.A. Dickey's Knuckleball Locations

R.A. Dickey has not been able to repeat his Cy Young season after being traded to the Blue Jays. Even adjusting for league change (that is, facing tougher competition on average), Dickey has regressed to a level that is closer to how he pitched with the Twins than the one he pitched at with the Mets.

I hadn't touched Dickey yet because I knew I would have nothing to add to the great research already done on knuckleballs. However, I had found no good articles talking about and trying to quantify Dickey's pitch locations from this season. The best articles I have read on Dickey's 2013 have been about movement (and release point) and velocity. There was some thought that maybe he was back, but his horrible start against the Twins on Saturday suggests otherwise.

Even Dickey himself said his dominant start against the Rays was more about location (though nearly all pitchers say that). So,. using average locations, I thought it would be helpful to take a look at how he is locating the knuckleball, and whether that gives us a better look and understanding as to why Dickey is struggling. When it comes to location, I am only going to look at the knuckleball in this post, as that is his feature pitch and what everyone talks about when discussing Dickey. First, let's compare his average knuckleball location in 2012 and 2013 (entire graph is the strike zone):

Dickey is throwing the knuckler slightly more arm side and lower on average than he was in 2012. I originally got the idea for this post because of how high the average knuckler swinging strike was on average. I thought that the knuckler being higher would be a bad thing, but his dominant 2012 had an average knuckler very high in the zone. Just to make sure that the higher knuckler is a better knuckler, the following graphs break down his average knuckleball locations sorted by result. First, here is 2013

Dickey's whiffs this year come when he gets the knuckler glove side on average and keeps the ball high. When he gives up contact (no real difference in average location between contact and homers for him this season), it is on pitches lower in the zone and more arm side. We get similar results when looking at 2012

So it would appear that a high knuckler is a good thing. I was curious to see if this was unique to Dickey, so I looked at the other pitcher with a high volume of knucklers in the Pitch F/X era

While getting it glove side was better for Wakefield (just like Dickey), he succeeded more when he got the knuckler low. So perhaps this is more unique to Dickey's "hard" knuckleball. However, Dickey is still having success up despite the velocity drop. The problem is that he just isn't getting it up quite as much.

Despite the location changes, his actual release point change has been very minor

One would expect (at least I would expect), that if he was hurting, he would have a harder time getting the ball down. Instead, the problem is that he is not getting the ball up. It is not as if a softer knuckler can't stay up, as Wakefield's appears to have stayed up on average, throwing it in the top half of the strike zone. Perhaps there is a slight mechanical issue I am missing, as it does not appear to be Rogers Centre, as he is locating the knuckleball higher at home (2.6 feet on average versus 2.45 feet on the road).

The knuckleball is fun, and the Mets agree

R.A. Dickey's story is now relatively familiar across baseball-land. He was a first round pick by the Texas Rangers, but then received a much smaller bonus when a doctor saw a picture of him, and determined that he was missing a ligament in his shoulder. Then he took some lumps in the big leagues, changed into a knuckleball pitcher, struggled more. He was then let go by the Rangers, bounced around, and then found himself pitching for the Mets. Then last year, he had a stellar year, and has become either the Mets number 1 or number 2 starter. This year he has been phenomenal, posting a 4.6 WAR and making on $2.75 million. This gives him an incredible 598 WASP. His WHIP is a solid 1.250, and a very good ERA of 3.35. When he gets ahead in the count, its basically all over, as his opponent batting average is .225, with a .168 average with 2 strikes. Key divisional opponents aren't doing much against him either, as the Braves are hitting just .190 off him, and the Phillies just .222. He is averaging 5.8 strikeouts per 9 innings, which is higher than he has ever had in the big leagues, and in his two years as a Met he has kept his home runs allowed per 9 innings under 1, something he had never done before. He is getting more groundballs than ever before, and throwing twice as many strikes as balls (with 75% of them being the knuckler). It is hard to pinpoint what exactly finally went right for Dickey, it may just have been gaining familiarity with the knuckleball, but things are going extremely right, and the usually inefficient, unlucky Mets somehow have this guy.

Wakefield, 200, and comparisons

On Wednesday, Red Sox Pitcher Tim Wakefield failed for the 7th straight time to get his 200th win (when he finally does, he will become the 111th pitcher to do so, quite an accomplishment considering how many people have been in the majors in the past century plus). I am biased towards the knuckleball, so I think Wakefield is the best thing ever, but lets look at some stats. Here I will compare him to the two pitchers who won exactly 200 games (Chuck Finley) and 201 games (Charlie Root). Wakefield has a career ERA of 4.41, Finley 3.85, and Root 3.59. They all have winning records, but in my opinion, wins and losses are extremely unhelpful in rating a pitcher (take Matt Cain, who has a career era of 3.35, but has a losing record). Wakefield also has a lower wins above replacement player (31.8, with negatives the past two years. I would be shocked if he pitches for the Sox next year) than Root (37.9), and Finley (55). Wakefield does have a better WHIP than Finley, but Root's is better.