“We are approaching a new age of synthesis. Knowledge cannot be merely a degree or a skill… it demands a broader vision, capabilities in critical thinking and logical deduction without which we cannot have constructive progress.”
—Li Ka Shing
“He showed why we have such high hopes for him for our future. It was on display tonight. We’ll get him back and [see] if he can continue on the journey.”
—Yankees general manager Brian Cashman, after the rookie Philip Hughes threw six and third innings of no-hit ball on May 1st.
With the start of Matt DeSalvo on Monday (a game which began well but ended badly at the hands of Adrian Beltre after a blown call at second base) the Yankees became the first team in major league history to start ten different pitchers in their first thirty games. The Yankees’ roster has included starters Andy Pettitte, Mike Mussina, Chien-Ming Wang, Kei Igawa, Jeff Karstens, Carl Pavano, Chase Wright, Darrell Rasner, DeSalvo, and Phil Hughes. As a reflection of that struggle, the Yankees also racked up a string of ten consecutive games in which they used five or more pitchers. And during those first 30 games the team won only 14, setting the stage for the return of Roger Clemens in four to six weeks.
As Yankee fans know only too painfully, the other circumstance that certainly didn’t dampen the desire to acquire The Rocket was the strained hamstring that landed Hughes on the disabled list, a hurt suffered after pitching six and third no-hit innings against the Rangers on May 1st. After a fairly rough major league debut against the Blue Jays on April 26th, Hughes looked very sharp against the Rangers. He recorded 53 strikes in his 83 pitches while walking three and striking out six. After Hughes threw on flat ground last Sunday, Will Carroll reported on Monday that the injury may not be as serious as first thought, putting him on schedule to return to the rotation just before Clemens enters it.
Beyond the never-ending speculation about the Yankees and their rotation’s ups and downs, calling up of Hughes gives us the opportunity to touch on the recent discussions of the so-called biomechanics revolution. For those who haven’t been paying attention, an article on Slate.com recently stirred the pot by noting the increasing availability of analysis like that of our own Will Carroll on MLB.com, in which he offers a closer look at the mechanics of Daisuke Matsuzaka.
While these articles may not be loaded with hard science in the traditional sense and include a fair bit of interpretation, that’s not in and of itself a reason to reject this kind of analysis. What they provide are valuable additional data points to assist in evaluating players and understanding a little bit better what transpired on the field. That can never be a bad thing.
As discussed in this space back in November, the good folks at Major League Baseball Advanced Media (MLBAM) are beginning to collect the kinds of information that can be combined with biomechanical evaluations in order to either verify and support or contradict those conclusions. When this information becomes readily available, what we’ll see is a synthesis where it will become commonplace to talk about release points and arm angles, with an assertion about a pitcher’s arm slot being lower today than it was last season getting backed by data, and with comments on how effective his pitches are in given circumstances illustrated with information. The beginnings of this kind of work have been highlighted in two fine articles written by Joe P. Sheehan (note the P.–the original Joe Sheehan is sans initial) providing yet another example of the power of combining two complementary ways of looking at the world.
Today we’ll illustrate the beginnings of the synthesis with a graphical representation of each of the 21 batters Hughes faced in that May 1st start. We’ll then dig just a little deeper to demonstrate where this is all going.
The key for each at-bat is shown below, with the addition that a strike with a red box around it represents a swinging strike.
On to the first inning. Batter up!
Kenny Lofton led off the game with a walk on six pitches. You can see in the upper right hand corner each pitch with its velocity recorded at 55 feet from the plate along with its velocity just before it reached the plate. In looking at the radar gun readings on the telecast, the speeds reported are pretty much an average of these two values, with perhaps a slight weighting towards the pitch’s initial velocity. The only curveball, pitch three, skimmed the dirt.
Michael Young erases Lofton by hitting into a deuce, having swung at the first and third fastballs. Although it’s not obvious until you put the previous two graphics side by side, unlike my previous interpretation of this kind of data, the strike zone is now adjusted for each hitter and each at-bat. In this case, Lofton’s zone starts a few inches higher than Young’s, and ends a bit higher as well, even though Lofton is supposedly close to two inches shorter than Young.
To end the inning, Hughes catches Mark Teixeira looking on the third consecutive offspeed pitch after starting the at-bat with what would prove to be his fastest pitch of the night (colored in pink)–94.1 mph out of his hand. There were two pitches later in the game that reached the plate at the same speed of 85.1 mph.
In the second inning Hughes again works his offspeed pitches, throwing them in 8 out of 11 pitches. After walking Hank Blalock he is helped once more by a double play.
In the third Hughes, throws all of eight pitches in making quick work of the bottom third of the Rangers lineup:
The fourth inning sees Hughes going back to his fastball, working both Lofton and Teixeira on the outer half of the plate, and getting three ground outs.
You’ll notice that the first pitch to Lofton is recorded at 78.4 mph, and was therefore an off-speed or breaking pitch. One of the goals in analyzing this data is to be able to detect whether the pitch was, in Hughes’ case, a curveball or a changeup. MLBAM has not developed the software to do that today, but since break angle and break length are captured in the data, it would be a matter of creating profiles of each pitcher’s repertoire and then developing an algorithm that would identify the pitches with some level of certainty given his pitch profiles. (When a pitch doesn’t match a profile, it could be flagged and examined, with the result that an existing pitch profile is updated or a new one entered into the repertoire database.)
In the case of Hughes’ outing, he threw 24 off-speed pitches among his 83 pitches. Six of the 24 were changeups, and the other 18 were curveballs. The average speed of the changeups was 79.3 mph, whereas the curveballs were 73.3 mph, but the measurement of break length on the changeups was anywhere from 40% to 60% of what it was on the curves. These are two data points that would go into building the profiles.
The fifth would prove to be Hughes’ longest inning. He threw 20 pitches, including giving up a walk to Ian Kinsler after eight pitches. With confidence in his fastball increasing, Hughes dealt just three off-speed pitches in the inning to Ranger batters.
In the sixth inning Hughes recorded two of his six strikeouts, both of them coming on swinging strikes on fastballs dealt to Gerald Laird and Lofton.
Hughes would only throw six pitches in the seventh inning, but the fifth, a fastball to Teixeira would be among his fastest of the night. The final pitch of the evening was a curve, and after hopping on his left leg he called Jorge Posada out to the mound. That was all for Hughes.
The rookie’s night can be summed up in the two graphs below. The graph shows all of his pitches against left- and right-handed batters, and the table summarizes pitch outcomes and velocities.
Outcome Velocity SS CS Ball Foul InPlay Total 100+ 0 0 0 0 0 0 94+ 0 0 1 0 0 1 89+ 10 12 19 9 8 58 83+ 0 0 0 0 0 0 <83 5 4 10 2 3 24 Total 15 16 30 11 11 83
Although the data obviously tracks other outcomes, I eliminated those from the list, since there were no swinging strikes in the dirt or intentional balls thrown. In all, Hughes threw 58 pitches that were clocked from 89 to 93 mph (using the first data point, velocity at delivery), and one pitch at 94+, to go along with 24 total off-speed pitches.
As a side note, it’s probably occurred to many of you that like the QuesTec system, this data could be used to track the accurracy of umpires. Although the representation of the ball I’ve used on these graphs is slightly larger than it should be (20 pixels versus about 17), and I’m using a strict 17-inch plate, my summary of the 46 pitches that were not swung at indicates that home plate umpire Derryl Cousins was correct on 40 of those pitches, good for 87%. (Cousins doesn’t generate a high number of strikeouts, relative to his peers–for more, check out our Umpires Report.)
As mentioned at the outset, this data can also be useful as a companion to biomechanical analysis. For example, the release point data is included and using it we can create graphs like the one below.
Here we see a two-dimensional representation of the release point on all of Hughes’ pitches. Although the release points appear to be fairly different, keep in mind that the horizontal axis is more spread out than the vertical axis. As a result, the actual area from which all of these pitches come from is an area seven inches wide by five and 1/4 inches tall–about the size of a postcard.
The area is consistent, although a bit smaller than what Joe P. Sheehan found when looking at the data for other pitchers like John Lackey and Felix Hernandez. If we had comparison data, we could then determine if perhaps his release point has changed over time or between starts. You can also get a feel for how this changes by pitch type. The following graph shows only the offspeed pitches.
Although the sample is smaller, the area is even further restricted, even though he threw two types of off-speed pitches.
Finally, we can take a quick look at how Hughes’ velocity was–or in this case was not–affected by fatigue as the game went on. The next graph shows the origin and destination velocities of each fastball he threw as the game progressed.
While you can detect a slight downward trend as the game wore on, the difference between the first inning amounts to less than two-tenths of a mile per hour in destination velocity from the first to the seventh inning. Once again, this sort of data (along with break length and angle) could be compared start-to-start in an effort to detect what triggers changes in effectiveness.
We’ve only begun to scratch the surface. The ever-increasing power of information technology coupled with the passion of baseball fans promises to usher in new ways of analyzing our shared distraction. It’s a great time, for the game, and to be a fan.