I’m not suggesting that a one-game sample is conclusive proof of anything, but for those of us who have been nervously waiting for Josh Beckett to return to form, today’s start against the Tampa Bay Rays had to be encouraging. I’m not even talking about the 10 strikeouts or just allowing a pair of hits and three walks; what I’m looking at is his velocity.
Those of you who read Baseball Prospectus 2009 are aware of what transpired for Beckett late last year:
…his velocity was often well below past averages; this was noticeable during the playoffs, when Beckett was topping out in the low 90s. He spent time on the DL in August with right elbow inflammation, but except for the velocity issue he was effective upon his return.
Just how much did his velocity drop during those playoff games? Let’s take a look at his last start from 2008, which was also against the Rays. Beckett lasted five innings and just 78 pitches (his third straight start that failed to go more than five) and struck out three hitters. The average velocity on his fastball was 92.3 mph, with the fastest of those coming in at 93.6 mph:
This is for Beckett’s four-seam fastball, by the way. Early on Beckett’s velocity was between 92 and 93+, but after just 10 or so fastballs he slid down towards the lower end of the 90s, even throwing one pitch in the high 80s. This is reminiscent of the previous start, where his average fastball velocity was 92.1 mph (and his max was 93.5) or the start six days before that against the Angels, where his fastball averaged 92.8 mph. Beckett is the kind of pitcher that gets out of trouble by ratcheting up the velocity and blowing it by the hitter, so when he can’t do that, he struggles to succeed at the same level. Which brings us to today’s start:
This chart doesn’t even approach the 80s, and the lowest velocity fastballs this time around were some of his better ones from the previous chart and games. His best came in near 97 mph, and while his velocity was down later in the outing, he also threw 93 pitches this time as opposed to 78 and was still able to hit the mid 90s throughout the entirety of the start. We’re just talking about one start here, but it’s a good sign for those that were worried about his velocity.
[UPDATE] A few of you in the comments asked for one graph that shows both. Thanks to some assistance, I now know how to do that in Excel 2007:
one thing that sticks out though, is that in october, the fastball velocity appears stable, whereas you see the dropoff for Beckett pretty clearly on 4/7, despite the increased velocity just by running an eyeball regression on it. Is this/might this fatigue be an issue? Do we see a change in pitching motion?
I doubt that losing 1 mph - at most - on his average fastball over the course of 7 innings of pitching is abnormal, particularly since opening-day adrenaline was involved. Given that he still notched a few at 96 late in the game, it doesn't sound like it signifies a problem. Dude was probably just cruisin' at that point, what with 2 hits allowed and a 4 run lead.
That said, I didn't actually watch the game. Was there some other reason to suspect a change in motion?
i gather you see a steeper decline in 4/7 because of the spikes around pitch 20, but that could be situational, say getting into a jam. as you can see he dialed it up again later in the game for a few pitches, showing that he's holding something back in the regular ones.
A comparison of the charts is misleading. The top chart looks less scattered because the outlier compresses the remaining data points. Just from the naked eye, I would say that both show a slight declining trend. Also, as Marc said, the first chart only represents 78 pitches (less than 40 FB), while the second represents 91 pitches (just over 50 FB). Most of the decline in the second chart comes after 40 FB.
I don't think there is a significant difference between the two in terms of decline over the course of the game.
Agreed, they both show the same kind of spread--the first from around 93 mph down to 90 mph or so, while the second has him around 96.5 to 93.5. That's roughly a 3 mph range for both, just in different places on the radar gun.
If I may offer a piece of constructive criticism. Many of the issues brought up in the comments could have been avoided by simply taking the data from both starts and overlaying it onto one graph instead of two. Then issues like decline rates, and scattering would be very apparent. Its rare that combining two scatter-plots doesn't make the result too busy, but in this case it would have solved alot of problems. Of course the other solution would be to ensure that both plots had the same axes. But that is assured by combining them.
I couldn't get Excel to do what I wanted it to after reading this, so I made a line graph that combines them. If you know exactly how to make that happen in Excel, I'm all ears.
There are a couple ways to show two data sets on a scatter plot in excel. The easiest is to double click on the lines of the line graph and slect "none" for the type of line, which should erase the lines but leave the points so that it looks like a scatter plot. You can also go to chart>>source data>>series>>add to add another data set to a scatter plot.
I agree with Mike W.'s implied criticism. Some sort of systematic error (or just "difference") in the measuring apparatus would show exactly the same change. Maybe you have complete confidence in pitchFX data, but the average reader doesn't know enough to. Could readings be different at different stadiums, or on different days? Is there a calibration process? How, if you think about it, would you produce a "standard" 90mph fastball, to be sure your setup was calibrated right?
I know this is just a quick post rather than an article, but if you had a link to an article that covered the reliability of pitchFX it would have been useful to put it in.
And as for strongarming Excel into plotting several sets of data points on a scatterplot, davelew had that right - at the "source data" step, click on "series" and use that window to make sure you plot "game 1 fastballs vs. pitch count" and "game 2 fastballs vs. pitch count." You'll figure it out - you enter the x and Y ranges separately for each "series" (game in your example). Excel sucks for scatterplots but you *can* bend it to your will this way.
16 comments have been left for this post.
one thing that sticks out though, is that in october, the fastball velocity appears stable, whereas you see the dropoff for Beckett pretty clearly on 4/7, despite the increased velocity just by running an eyeball regression on it. Is this/might this fatigue be an issue? Do we see a change in pitching motion?
I doubt that losing 1 mph - at most - on his average fastball over the course of 7 innings of pitching is abnormal, particularly since opening-day adrenaline was involved. Given that he still notched a few at 96 late in the game, it doesn't sound like it signifies a problem. Dude was probably just cruisin' at that point, what with 2 hits allowed and a 4 run lead.
That said, I didn't actually watch the game. Was there some other reason to suspect a change in motion?
http://baseballprospectus.com/article.php?articleid=7641
The premise is a little different, but apparently drops in both velocity and in arm slot are to be expected. Makes sense to me!
Whew. You scared me, D.P.
i gather you see a steeper decline in 4/7 because of the spikes around pitch 20, but that could be situational, say getting into a jam. as you can see he dialed it up again later in the game for a few pitches, showing that he's holding something back in the regular ones.
A comparison of the charts is misleading. The top chart looks less scattered because the outlier compresses the remaining data points. Just from the naked eye, I would say that both show a slight declining trend. Also, as Marc said, the first chart only represents 78 pitches (less than 40 FB), while the second represents 91 pitches (just over 50 FB). Most of the decline in the second chart comes after 40 FB.
I don't think there is a significant difference between the two in terms of decline over the course of the game.
Agreed, they both show the same kind of spread--the first from around 93 mph down to 90 mph or so, while the second has him around 96.5 to 93.5. That's roughly a 3 mph range for both, just in different places on the radar gun.