Historical Monthly Velocity Trends
Over the last week, I took an early look at those pitchers whose fastball velocity has increased most since last season, and those whose velocity has declined. In the comments, several wondered how velocity trended during the season. Intuitively, we would expect that velocity would gradually rise into the summer months as pitchers build their arm strength throughout April and battle through the cooler months. If this is the case, then we would become more optimistic and less panicky about the early season velocity dippers. Let’s see if this happens.
Using the FanGraphs league stats tab, I looked at the average fastball velocity by starting pitchers over the past five years for each month of the season. Below is a table with the data and then a graph to visualize it better.
| Mar/Apr | May | Jun | Jul | Aug | Sept/Oct | Average | |
|---|---|---|---|---|---|---|---|
| 2011 | 90.6 | 90.9 | 91.0 | 91.1 | 91.2 | 91.0 | 91.0 |
| 2010 | 90.5 | 90.7 | 90.7 | 90.8 | 90.9 | 90.7 | 90.7 |
| 2009 | 90.4 | 90.5 | 90.7 | 90.9 | 91.2 | 91.0 | 90.8 |
| 2008 | 89.8 | 90.2 | 90.4 | 90.6 | 90.4 | 90.6 | 90.3 |
| 2007 | 89.4 | 89.7 | 89.7 | 89.7 | 89.9 | 90.3 | 89.8 |
| Average | 90.1 | 90.4 | 90.5 | 90.6 | 90.7 | 90.7 | 90.5 |
There are several interesting observations we can make with this data. First is that velocity does indeed increase throughout the season. In the first two years, velocity also continued to rise throughout September and October (when there were actually any regular season games). However, in the most recent three seasons, velocity declined heading into the final month, which does make logical sense.
On average, we can expect a starting pitcher’s velocity to increase 0.6 miles per hour from the beginning until the end of the season. That is pretty significant. Not surprisingly, the biggest jump is from Mar/Apr to May, which represents half the overall increase a pitcher should experience. Of course, for a starter whose velocity is down 2.0-2.5 miles per hour at this point in the season, this impending jump is still not nearly enough to make us feel any better.
Another trend worth pointing out is how fastball velocity has steadily increased every single year. In fact, if you go all the way back to 2002, the first season FanGraphs has velocity data for, it has nearly risen in a straight line. Beginning at 89.5 miles per hour in 2002, average fastball velocity has only declined in two years, while rising in the other eight. Are teams becoming smarter about the importance of velocity and its correlation with strikeouts? I wouldn’t think so, as I would figure teams have always placed significant weight on it.
The data above suggests that pitchers whose fastball velocity is down 0.5 to 1.0 miles per hour should be of little concern at the moment. With an expected bump as the season pushes on, these guys will end up finishing pretty close to where they had been the previous year. However, this also tells us that on average, pitchers do not suddenly regain two miles per hour on their fastball in the middle of the season. This is an ominous sign for aces like Tim Lincecum and Felix Hernandez.
Mike Podhorzer is the founder of ProjectingX IQ, an advanced fantasy baseball analytics platform that transforms projection data and in-season performance signals into actionable intelligence. He is the 2015 Fantasy Sports Writers Association Baseball Writer of the Year and three-time Tout Wars champion. He is the author of the eBook Projecting X 2.0: How to Forecast Baseball Player Performance, which teaches you how to project players yourself. Follow Mike on X@MikePodhorzer and contact him via email.
2007 is curious, as well. I wonder what caused the late season spike we don’t see in any other season… who got called up in late 2007?
Great article. Very interesting.
“However, this also tells us that on average, pitchers do not suddenly regain two miles per hour on their fastball in the middle of the season.”
Not quite true, based on what you wrote at least. The data tells us that on average velocity tends to increase by 0.6 mph over the season. What is presented here says nothing about the standard deviation around that trend, which is what we need to know in order to evaluate that statement. I.e., just because the average pitcher only increases 0.6 mph doesn’t mean it is “rare” to see 1.5-2.0mph over a season.
What standard deviation did you find in the data?
This. It could be true, but this data could well mean that some starters have big velocity boosts but man have small ones. The standard deviation would tell you how much variance there is in velocity gain.
This.
Even more than just looking at the standard deviation of the velocity change though, why not just restrict the data set to players who saw a drop of ~2 mph in April relative to the previous year. I would be much more interested in seeing how the velocity of these players trends throughout the year than just the league average.
Hm. I am suspicious. Increases in velocity league wide could simply be an artifact of injured pitcher attrition through the year.
That’s an excellent point.
I would agree. At very least, the author should be looking at the same pitcher population each month. It sounds like he is using raw league averages, ignoring the fact that soft-tossers (esp. unexpected soft-tossers) may be getting weeded out as the season progresses.
Still, the effect appears to be real according to a more detailed study by Fast (below). Although it may be driven partly by air temps.
http://www.baseballprospectus.com/article.php?articleid=13380
Is that true? Its an interesting hypothesis, but I’m not sure why we should assume attrition creates an early selection bias. Are pitchers more likely to get injured in spring training or April than they are in June? Shouldn’t that attrition be a consistent factor throughout the year rather than one that’s extreme early and diminishes as the season goes along?
The assumption would be that there is a population of pitchers who represent the “walking wounded.” They start the season but the underlying ailment is reflected in lower pitch speeds and, eventually, a season ending DL trip/demotion/DFA.
Right, but my question is why that is something that should be more prevalent in April than any other point? Pitchers who get hurt on May 15 and make a couple starts before it gets revealed should have the same effect on the sample, no? Why do we assume that injuries are less likely to be immediately detected in April than later in the year?
Because there is an off-season where these theoretically moderately injured pitchers are not being measured. If you stop measuring something for 6 months, and then start measuring it again monthly, that first month (April) will reveal more hidden information than the fifth month (August). Anecdotally, Pineda this year would be an example of this possible effect.
@Mark: The fact that velocity seems to rise throughout the season in this study is completely consistent with attrition being a consistent factor, not just an April issue. Velocity tends to increase every month in this finding, albeit at a lower rate. And attrition need not simply be due to injury, but also to ineffectiveness: journeyman pitcher starts season in rotation, has lower velocity than last season, is ineffective, gets relegated to long relief, is replaced in rotation by harder-throwing youngster… average velocity of sample increases.
@MP
If we were measuring something and and then decided to simply stop, your point would be dead on, but it’s not that the measurement stops in September, it’s the measurable, so I’m not sure its so meaningful in this case. Pitchers have less opportunity to hurt themsevles at times when performance is not being measured because they’re not pitching, so its not as if a bunch of pitchers all get injured in the offseason, we watch them for a month, and then weed them out. And besides, we have all of spring training to do that even if it were the case, which is really what Pineda is an example of. He hasn’t thrown a single pitch in April. It’s not as if suddenly in April teams start measuring these things and filter through injured pitchers–that’s just when the data becomes consistent on a publically available basis. Teams and other insiders are measuring them for 4-6 weeks prior to that.
@mcbrown
I’m not sure if you’re agreeing with me or disagreeing, but the point about pitchers being replaced, whether due to ineffectiveness or injury, simply doesn’t strike me as meaningful in terms of it’s effect on average pitcher velocity. I’m not sure why its viable to assume replacement pitchers throw harder than the pitchers they replace. Just as often, it’s the journeyman pitcher doing the replacing of the injured hard-thrower. The harder throwers are more likely to be in the majors to begin with (with the exception of elite prospects, who represent a minority of the replacement class of pitchers in a given year), while the journeyman are not.
Mike Pelfrey, whose velocity was up this year from last year, busts his UCL and gets replaced by the soft-tossing Chris Schwinden, and if Schwindein is ineffective there’s a decent chance he’s ultimately replaced by the even-softer-tossing Chris Young who is currently on rehab.
Daniel Hudson goes on the DL and gets replaced by Wade Miley just as often as Josh Colmenter is ineffective and gets replaced by a harder-thrower.
Flamethrowing but ineffective Daniel Cabrera would routinely be given chances on his arm strength alone through the mid to late aughts, but only twice in his career did he make at least 30 starts. odds are the pitchers that filled in for him didn’t throw as hard as he did.
And even then, none of this strikes me as exclusive to or more prevalent in April. None of this is to say I don’t believe velocity correlates to effectiveness, but it is a scarcity, and I don’t buy the idea that replacement pitchers average higher velocity than the pitchers they’re replacing or the idea that there will be more pitchers pitching injured and below their “healthy velocity” in April than in June.
Assume that all injuries are preceded by a drop in velocity, and this drop can occur at any time in the calendar year. Assume that all pitchers called up to replace injured pitchers are selected for in part because of higher than normal velocity. If these two assumptions are even partly true there will be at least some selection bias when trying to compare MLB avg. April velocity with MLB avg. Sep. velocity. Only if these two assumptions could be shown to be completely false could you safely ignore the effect without testing it.
I’m not ignoring them, I’m debating them on philosophical grounds. You guys are presenting an objection to the conclusion Mike has drawn from the data on philosophical grounds, and I’m presenting an objection to your objection on philosophical grounds. Seems a bit of a double-standard to lay the burden of proof at my feet as a supporting argument for your claims, you could just as easily do the research as I could and clearly have the stronger objection than I do (and its not easy to do said research, but that’s not a viable argument against a philosophical objection).
There’s still a problem with your argument, even if I go with your assumptions (which, of course, I don’t, but for argument’s sake): why does velocity reset in April? If replacement pitchers persistently feature above average velocity, and the net effect of attrition+replacement is positive, why doesn’t the trend continue upwards into the subsequent season?
Obviously, this is absurd, but it is the logical conclusion of your assumptions. If attrition is persistent throughout the year (part of your assumption, you state it explicitly), but the net effect of attrition plus replacement velocity is positive, there should never be a sharp decrease. At worst, the data should level off as the net increase “catches up” to the average replacement velocity (which you assume is higher than the average April velocity), and the effect of attrition+replacement becomes neutral.
You’re missing an important premise. For your argument to function, there has to be a reason that average pitcher velocity decreases during the offseason. Maybe it’s just me, but that screams violation of ockham’s razor. You need an extra explanation to explain the first explanation.
I’m not outright rejecting your hypothesis. I’m just suggesting Mike’s initial conclusion seems more plausible. If it could be justified for me to invest the time and effort in testing both hypotheses, I’d be happy to do so, as I assume you would as well, so let’s not play the “burden of proof” game when we’ve implicitly agreed by simply having this discussion in a rotographs comment section that neither of us is prepared to go that far.
The other potential point here is that perhaps we should place additional weight on April velocity gainers. Should we treat those guys as if they’d actually gained about half a mph more on their fastball than they actually have? If so, that’s potentially extremely promising.
Where is the Fangraphs league stats tab?
Nevermind, I found it.
Hard to make year to year comparisons without taking into consideration the variation in guns used to measure velo from park to park and especially from year to year. With current technology, you can measure the ball at any point from release to it crossing the plate. Where do you take the reading? 6 inches, one foot, two feet, five feet? The old ray guns used to measure it across the plate. you had to make sure you got a “roll down”. 84-86 used to be an average FB on a “ray” gun, provided it “rolled down,” back in the 70’s to early 90’s. That was equal to around 89-91 on the “jugs” at that time. With the current technology, 84-86 is sometimes 91-93. it can be a 7 mph bump, depending on the gun. If you look at the difference between home and away velocities, you can get an idea of which stadiums are “juiced.”
I’d highly doubt that technology has produced a steady increase in velocity over the past 5 years.
The advancing technology of measurement does indeed change the data set. You get one set of numbers when you measure with an hourglass, and another set of numbers when you measure with a micronometer.
I think you missed the part of my explination: “With current technology, you can measure the ball at any point from release to it crossing the plate. Where do you take the reading? 6 inches, one foot, two feet, five feet?”
There is not an industry standard as to where along that point you take the measurement. 5-10 years ago, that wasn’t even an option.
Here’s another point: If you have five guns recording the same pitch, you might have five different readings with a variation of .1 to 1.7 MPH. Who decides which one is valid?
Pitch velocities continue to rise. I looked this article up because it appeared to me that overall pitch velocities seemed to be down, although I don’t think the data seems to bear that out.
I’ve been looking for a way to assess statistically where the PED era began and ended (if it did). I thought that pitch velocity might be able to tell us that. But that assumes a lot – consistent velocity measurement and assuming that MLB employed a similar percentage of hard throwers over the years. So, maybe my idea is flawed.