|Data from JoeLefkowitz.com|
While I've dabbled in horizontal and vertical pitch break analysis in the past, I realized that despite looking at the numbers, I still wasn't entirely clear on what constituted a "good" or "bad" amount of pitch break. Fortunately there are several prominent sabermetricians, including Mike Fast, Dave Allen, Josh Kalk and Sky Kalkman, who have done yeoman's work in helping amateur statisticians like myself in better understanding how to interpret PitchFX data. I'd also like to take a moment to thank SG of RLYW for providing feedback and guidance for this post.
For this particular post I wanted to acquire a data set that would ideally enable me to take a look at a pitcher's H-break and V-break for a given pitch on either an individual game or seasonal basis and be able to state with some confidence that said pitcher had "good," "bad" or "league average" H-break and/or V-break on said pitch. The end result is the table you see at the top of this post.
I wound up tallying the cumulative seasonal H-break and V-breaks for all of 239 pitchers that threw at least 10 innings as a starter, to ensure that the sample was large enough, binning them as righties or lefties, logging their 2010 Pitch Values per Fangraphs to see who had the best and worst pitches for each pitch type so I could rank the associated pitch break accordingly, and subsequently logging the average seasonal H-break and V-break for all 239 pitchers in the sample using Joe Lefkowitz's site.
As it turns out, I didn't have to spend the countless hours that I did compiling this data, as, had I looked just a little deeper, Fangraphs apparently carries individual PitchFX data for each big leaguer, and also enables you to toggle the league average for a given pitch's pitch break. I'm not entirely sure how I've been using Fangraphs on a daily basis for the last 18 months and somehow never managed to uncover this treasure trove of data, but it happened, I still ended up spending 30 hours creating a spreadsheet, and there's no way to undo all that time. If you are interested in reviewing my work, please click here to download the spreadsheet I created to tabulate all of the data for this post. (Note: Turns out TexasLeaguers.com carries an average pitch break table as well; although theirs tabulates every pitcher in MLB -- and includes postseason data -- whereas the one found at the top of this post if for starting pitchers only during the 2010 regular season).
Here are the H-break and V-breaks for the top and bottom 10 Fastball pitchers in MLB, per Fangraphs, which doesn't isolate Four-Seamers, Fastballs, Two-Seamers and Sinkers in its weighted pitch data but rather appears to utilize whichever fastball type a given pitcher throws the most (e.g., Tim Hudson's wFB value is for his sinker).
Interestingly, among the lefthanded Four-Seamer batch, the discrepancy between H-break for the Top 10, MLB average and Bottom 10 is almost nonexistent; however in terms of V-break the Top 10 threw their four-seamers nearly an inch higher than the MLB average. The top 10 lefties all have fairly comparable V-breaks on their Four-Seamers except C.J. Wilson, whose Four-Seamer broke nearly two inches lower than the rest of the pitchers in that chart. Wilson also threw his Four-Seamer less often than the other nine, and favored his Two-Seamer, which breaks more than two inches closer to the plate than the MLB average -- seven of the other 8 on the list with Two-Seamers actually broke further away from the plate than the MLB average.
I'm not going to dive too much deeper into the weeds on the three other fastball types, as there's already way too much data to digest and likely a decent amount of noise in the data set regarding the differentiations in pitch classifications, so feel free to draw your own conclusions.
Here's the data for Sliders:
Here's the data for Cutters:
two inches closer to the plate (obviously getting too much of it).
There's a bit more of a difference in reviewing the lefthanded Cutter, but there's also a significantly smaller sample size, as there simply aren't that many lefties that throw Cutters, so I'm not sure if the data is tremendously helpful. The top lefty Cutters average -1.44 H-break compared to the MLB average of 0.11, and 4.18 inches of V-break (exactly an inch closer to the plate than the MLB average).
Here's the data for Curveballs:
The Top 10 lefties saw their Curves also break significantly further than the MLB average, with -4.69 inches (compared to -3.16) of horizontal break and nearly -7 inches (against -5.45) of vertical break. Curiously, the Bottom 10 Curveballs were actually closer in break to the Top 10 than the MLB average Curves were -- clearly one of the reasons for the discrepancy in pitch value was velocity, as the average speed of the Top 10 was 75.4, while the Bottom 10 averaged 72.1.
And here's the data for Changeups:
The Top 10 lefthanded Changeups saw almost no difference in H-break from the league average (7.86 to 7.80), but broke about half-an-inch less higher vertically (4.90 to 5.43). The least effective lefty Changeups curiously had nearly identical average breaks compared to the league average, and so I assumed the discrepancy would be due to a significant difference in velocity as was the case for the lefthanded Curveball. However, the Top 10 Curves averaged 82mph, while the bottom were 80.6, and so this is the one pitch type of the five we've discussed herein where the average breaks really aren't helping distinguish the best from the worst. Of course, if you examine each pitcher in the Top and Bottom 10 individually, you'll see a pretty dramatic difference between the pitch breaks of the good and bad curveball throwers and it would appear that the breaks between Top and Bottom pitchers are so erratic that they both managed to average out to unexpectedly similar figures.
In poring over this data, a few random non-pitch break observations couldn't help but pop out:
- Not that this is news, but James Shields, Scott Feldman and Brian Bannister had really terrible years. Bannister ranked below average in all four of his weighted pitch categories, and had the worst Cutter in the game. No wonder he's off to try to resuscitate his career in Japan. Hey, it worked for Colby Lewis.
- Ian Kennedy had the second-worst Slider among righthanders, but was in the Top Ten in both Curveballs and Changeups.
- Phil Hughes had the fifth-best Cutter among righthanders.
- C.J. Wilson had a fantastic year, appearing among the Top Ten lefthanders in wFB, wSL and wCT.
- Ted Lilly had a strange year. Despite having the eighth-worst negative delta between his ERA and FIP, he ended up recording the second-most-valuable lefthanded Fastball and 8th-best Cutter.
- Good news Brewers fans: Shaun Marcum's Changeup was the fifth-most valuable pitch among all pitches in this sample, behind only Tim Hudson's, Ubaldo Jimenez's, Trevor Cahill's and Cliff Lee's Fastballs. Felix Hernandez's Fastball rounded out the top six.
Anyway, if you are somehow still with me after all of this, to bring everything full circle, let's go back to those two posts I linked in the very first paragraph, one of which analyzed Cliff Lee's ALCS Game 3 start against the Yankees and the other looking at CC Sabathia's Game 5 start against the Rangers.
Here are Lee's and Sabathia's average pitch breaks from Game 3 and Game 5, respectively:
Cliff Lee's pitching was otherworldly that October night.
As for Sabathia, the break on his pitches wasn't quite as dramatic as Lee's, but he still significantly outperformed the league pitch break-wise with his Sinker, Changeup and Slider -- not coincidentally the three pitches that, according to Brooks, were the most effective per Linear Weights out of the five pitches Sabathia went to that night.
Anyway, there are several important caveats with all of this data -- (i) I realize that by including spot starters and pitchers who pitched significantly fewer innings than the primary starters for most teams that there could be some statistical bias in averaging the pitch breaks given the discrepancies in innings pitched, but I felt drawing from as large a sample as possible would ideally drown out most of the noise, and by including everyone that made a start the data set would be as close to a comprehensive compilation of average pitch breaks as one could find for all 2010 MLB starting pitchers.
(ii) As you can see, there are a few pitches I didn't bother with -- Screwballs, Forkballs, Eephus pitches, Knuckleballs and Knucklecurves were left out, since there simply aren't enough pitchers that throw them to draw any meaningful conclusions from averaging pitch breaks. I also left the Splitter out of this analysis -- even though Fangraphs carries wSF, there aren't enough pitchers that throw it to extrapolate meaningful conclusions. It also needs to be noted that Fangraphs' data and Lefkowitz's data, while pretty close, don't correspond perfectly. There are several instances whereby Fangraphs has weighted runs above average data for a specific pitch by a certain pitcher, but Lefkowitz doesn't have corresponding pitch break data for that pitcher having thrown said pitch.
(iii) And perhaps most important to keep in mind is that obviously every pitcher in baseball not only throws every pitch in their arsenal at different H- and V-breaks for every single pitch they throw, but at different speeds than everyone else does. And of course some pitchers also achieve later breaking action than others. Even if two random pitchers from a random distribution threw their fastballs at the same exact H-break and V-break, and one pitcher is among the best in the game and the other the worse, we have no idea whether either pitch will be considered "good" by looking solely at the pitch break -- other variables must factor in as well. In the Dave Allen articles linked below he assigns run values to a given pitch's H-break and V-break, which is massively insightful, although it's in graphical form which makes it difficult to review the actual numbers.
While I still have plenty more to learn about PitchFX data, and hopefully run values will soon be easily customizable, sortable and comparable to the league average on Fangraphs, my goal with this data was to at least provide some sort of benchmark or guideline so that when I look at a given pitcher's H-break and V-break data in a given game, I can at least try to get a sense of the level of effectiveness.
The following research was vital to compiling the information found in this post:
A PitchFX primer, by Mike Fast [Fast Balls]
The Breaking and the Knuckling: Run Value by Pitch Movement, by Dave Allen [The Baseball Analysts]
Fastball and changeup run value by movement, by Dave Allen [The Baseball Analysts]
Anatomy of a Player: League Average Pitcher, by Josh Kalk [The Hardball Times]
The Internet cried a little when you wrote that on it, by Mike Fast [The Hardball Times]
Understanding PitchFX graphs: Location and movement, by Sky Kalkman [Beyond the Boxscore]
PitchFX: Fangraphs' Sabermetric Library, by Steve Slowinski [Fangraphs]