The extinction of Zito’s hard curve

According to MLBAM pitch classification, last year Barry Zito threw 681 curveballs, accounting for 22 percent of his pitches. This year he has thrown 499 of them (as of Sept. 4)— that is 18 every 100 pitches. The big drop has occurred against left-handed hitters, who used to see a lot of Zito’s deuce, nearly 30 percent; this year, they have been fed it at a 16 percent clip.

The following graph depicts Zito’s curveballs (2008 version) by their movement.

You don’t need a software that performs cluster analysis to notice that there are two distinct types of pitch in the above plot—you don’t even have to know what cluster analysis is, for that matter. Separating those two pitches looks as easy as putting the curves with a vertical movement greater than minus five in one bucket and the remaining into another one. If you calculate the average speeds, you get 77 mph for bucket one and 71 mph for bucket two.

So Zito used (in 2008) two different curves: one faster, the other with more vertical break. This is consistent with his scouting report at 60 ft 6 in, that states “A couple years ago he added a tighter breaking ball that he uses like a slider to LHs”. The reference is to his 77 mph curve, that I’ll call the hard curve from now on (as opposed to the slow curve).

Last year, Zito delivered 202 curveballs to left-handed batters, 98 of them (49 percent) of the hard kind. He also used the pitch against righties, but to a lesser degree (123 of 479 total curves, 26 percent).

Let’s look at the curveball scatter plot for 2009.

It’s gone! Zito is not using the hard curve anymore, save for an occasional pitch here and there. He throws the slow one with the same frequency he did in the past—15 percent to lefties (identical to 2008), 18 percent to righties (3 percent more than last year).

To make up for the discarded pitch, he is using a slider extensively for the first time in his career. According to MLBAM classification algorithm, he used to throw one slider every 100 pitches; now he is using it 13 percent of the time. FanGraphs has different values (from 10 percent to 19 percent), but I’ll stick with MLBAM’s because with them I’ll be able to present lefty/righty splits.

His new weapon is used mostly against left-handed hitters, the same target that used to be fed the extinct hard curve. Zito delivers the slider once every four pitches against lefties, and once every 10 to righties. Other than filling up the void left by putting the hard curve in the closet, the new pitch is eating up some of the change-up share: both left-handed and right-handed batters have seen a drop of 10 percentage points in that pitch, meaning that the former are hardly seeing it anymore.

Nick Kapur at UmpBump wrote, at the All-Star break, that the slider gives Zito “another option to get a strike over without always having to groove his 86 mph fastball right down the pipe”. That seems to be hardly the case, since the slider is the pitch Zito is throwing less over the plate (41 in-zone percentage). And that’s for a reason, since batters are chasing the pitch out of the zone (42 percent swings on balls).

In fact, Zito is placing the other pitches more frequently over the plate: The change-up, now used only versus opposite-handed batters, has gone from a 42 in-zone percentage to 47, the fastball from 40 to 44, and the slow curve from 46 to 51. Moreover, batters used to chase the slow curve out of the zone 20 percent of the time; this year they are fishing a lot more, at a 31 percent clip.

As a result, Zito is falling behind in the count less frequently this year: he has decreased the 1-0 counts by nine percent and the plate appearances in which he goes on three balls counts by six percent, while increasing by the same six percent the PAs where he goes on two strikes counts.

His strikeout-to-walk ratio is back to MLB average for the first time since 2005, thanks in part to the 3.75 ratio against lefties (in 163 PAs, so let’s not expect it to stay so high in the future).

Okay, Barry. Now learn another pitch for right-handed batters during the offseason and you’re back to Cy Young form!

References & Resources
PitchF/x data from Sportsvision / MLBAM.

Barry Zito’s secret weapon by Nick Kapur @ UmpBump.

A Hardball Times Update
Goodbye for now.

Barry Zito’s pages at FanGraphs and Baseball-Reference.

Newest Most Voted
Inline Feedbacks
View all comments
Tom M. Tango
12 years ago

I was thinking exactly like John.  I really wanted to see what his slider looked like.  Indeed, I’d like to see all the pitches, just to make sure that it’s not an algorithm issue.

John Walsh
12 years ago

Actually, I went a looked at the movement plots (at Fangraphs) for several starts in 2008 and 2009.  For 2008, all Zito’s pitches are classified as fastball, curve or change.  But, there are clearly 4 different pitches there: there is a cluster of pitches in the slider zone which are sometimes called curves and sometimes changeups.

For 2009, Zito has 5 pitches (slider and cutter have been added), but the plots don’t look materially different.  It looks like in 2008 the classifier was forced to use only 3 pitches.  That’s my guess, anyway.

Mike Fast
12 years ago

The biggest difference I notice between 2008 and 2009 for Zito is that he added a couple mph back to fastball speed.

Max Marchi
12 years ago

I was also concerned about the algorithm change issue, so I did some checking before posting the article.

I’ll try to produce some more graphs during the weekend; meanwhile here are some data about the two pitch (They can’t be considered conclusive, but they seem to me a hint that they are actually two different pitches).

2008 Hard curve
Speed 76.9 (1.48)
H-Mov -2.97 (2.21)
V-Mov -0.59 (1.58)

2009 Slider
Speed 80.0 (1.47)
H-Mov -3.13 (2.12)
V-Mov 2.81 (2.01)

(Those in paretheses are standard deviations).

Looking at the vertical movement they look like different beasts; there is also quite a gap in speed.

Mike Fast
12 years ago

The difference in vertical movement is pretty big.  The difference in speed may be partly explained by Zito throwing all of his pitches faster in 2009.

Tom M. Tango
12 years ago

“The biggest difference I notice between 2008 and 2009 for Zito is that he added a couple mph back to fastball speed. “

Isn’t it also possible that there’s a misclassification going on here?

Let’s say that Zito throws so many kinds of pitches, that it’s hard to cluster.  Let’s say we “know” his fastball is thrown at 83-85, but there are alot of pitches also thrown at 79-81.

One algorithm lumps the 79-81 with his fastballs, and another calls it a slider.  And the algorithm changes between the two seasons.

So, even though nothing about him changed, one year, his FB is 84mph, and another year, his FB is 82 mph.

This is why I suggest that a “fastball” classification, for the same pitcher, be more rigid, especially if the number of FB thrown changes.

Take, for example, all his pitches (FB, curve, slider, whatever), take the top 25% fastest thrown, and call that his “fastball”.  If you do that, did Zito actually throw faster now?

12 years ago

Max, those numbers still strike me as Zito throwing the same PITCH per se, but just throwing the hard curve in a way that it moves slightly differently.

This pitch still has the same horizontal movement as the hard-curve, and the 2-3 inch increase in “rise” may simply be the result of throwing harder with the same grip.  In essence, Zito is throwing the same hard curve, but it might be BETTER (not sure on this, i’d like to see run values on the new hard curve rather than the old hard curve). 

In essence, unless a pitch is completely different, i’d hesitate to be calling situations like this one of “dropping one pitch (to extinction) and adding another pitch” and rather fine-tuning and improving that one pitch.

John Walsh
12 years ago

Nice work, Max. 

I’m wondering if this is slider is really a different pitch from the old hard curve.  Maybe MLBAM’s classification algorithm has changed.  Perhaps Zito’s improvement this year is more due to how he uses the slurve and changeup, rather than actually ditching one pitch for a new one.

It would be interesting to see a movement plot of sliders and curveballs (together on the same plot) for both 2008 and 2009.

Or alternatively, could you provide the average movements and speed for Zito’s 2009 slider, so we could compare to the old hard curve?