Pitcher fastballs peak at 29, decline quickly
The greatest thing for me as a fan the last few years is the explosion of great data researchers can use, from play-by-play information that helped build better defensive measures to the amazing stuff you can get out of MLB’s Pitch f/x system.
Like this, “Preliminary aging curve for fastball speed” by Josh Kalk. It’s early, yes, and Kalk discusses some potential limitations of the data, but go look at that. That’s the kind of data people — any of us — can use to figure stuff out.
It appears that until pitchers reach 28 or 29, they increase the speed on their fastball by about 1.5 mph. After 29, there is a rather sharp decline in fastball speed.
During the next five years, pitchers lose just over four mph.
No one ever knew this before this article. You had to run a team and be willing to devote ridiculous resources to get this, or be an outsider willing to invest several times more than that.
Be excited.
Thanks to Alex, whose email bumped this up the reading queue
Comments
44 Responses to “Pitcher fastballs peak at 29, decline quickly”
Leave a Reply
You must be logged in to post a comment.
I am excited and can’t wait to read. Googling the article right now.
A link would be helpful though…
Well hell, I’ll just put a link here. You shouldn’t have to do all the work.
My fault, fixed
My guess would be pitchers compensate by learning to pitch and not just relying on blowing batters away.
Great stuff and can’t wait till we get a few more years of pitch/fx data under our belts for some really juicy analysis.
David Wells beat the M’s yesterday by switching arms and throwing right handed. How does that factor in… switching arms I mean.
From the article:
Good knowledgeable article, but this comment seems suspect to me. ERA is affected by so many things besides pitching velocity.
Also, how do these statistics measure up when comparing, say Pedro Martinez vs. Jamie Moyer? Different pitchers, different styles. Pedro is (was) a power pitcher, but Moyer has always been about location and speed changes.
Oh I get it, Colon’s fat like Wells…
What’s scary about the data is that it suggests Felix will ad about 2 miles an hour more to his fastball before he hits 29!
Although I imagine since the clock started so early on Felix he’s probably an anomaly as far as this curve is concerned. However, the thought that it could happen is mind blowing.
This post is about natural aging curves for pitchers.
Speculation about non-natural aging curves isn’t helpful.
10- Fangraphs has that data for the last three years for most players. 05,06,07. Not quite what you are after but may be helpful.
Moderation note: I’ve deleted the steroid comments and their replies. Everyone reasonable may proceed.
DMZ said – No one ever knew this before this article. You had to run a team and be willing to devote ridiculous resources to get this, or be an outsider willing to invest several times more than that.
Is this info that will aslaway be available to the public?
My question is about the pitch/fx data is are we to the point where we can examine the breaks/movement of pitches in one game vs. another. For example, can we look at the stuff Brian Bannister had early in the year and see if he has really suddenly had less stuff, or are the bats just hitting the ball better?
Did Jon Lester have awesome stuff the night of the no-hitter, or was he just suddenly more likely that night?
How would I go about answering those questions? Where is the pitch/fx data stored at?
Good to hear fastballs peak at 29 and decline rapidly the same year we trade the house for a 29 year old flame thrower who will now see his stuff begin to decline. Sweet!
I’m not sold on this at all. 143 total pitchers means you have a tiny sample size going year to year, relievers and starters aren’t separated so you’re not correcting for the effects of moving to the bullpen or the rotation (and even one pitcher in an age group is going to skew it badly due to the sample size), and using late ’07’s numbers compared to early ’08’s doesn’t strike me as a great way to build a age curve.
It’s a cool study, but I think we’re going to have to wait a few years before the data’s good enough to draw real conclusions out of.
I’m totally with Graham on this. It’s an interesting start, but not ready for conclusions to be drawn.
Using the fastball data off of Fangraphs, I see, for one, most people lose velocity all of the time, but velocity seems to be stable around age 26-27, then dropping more and more and more as the pitcher gets older from there. If I use only >= 400 BFP, I get an average of .35 mph lost per year, with it being ~.2 mph in their 20’s, .33 mph or so in their early to mid 30’s, and .5 mph in their mid to late 30’s, and .75+ per year after that. That’s a total of 206 year to year samples. If I increase it to 200+ BFP, I get .23 mph lost on average, but that undoubtedly includes moves to the bullpen, there is still over all loss (on average) for every age group, 23-24, 24-25, etc. There does appear to be less loss in a pitchers late 20’s that say in their late 30’s, but that’d be expected.
That’s using 2005-2007 data.
My hunch has always been Fastball velocities are downhill from a pitchers mid 20’s, but I never had the data before fangraphs and the pitch f/x data, which I STILL haven’t mined 🙁
14 & 15: Agreed — as individual results may vary (case in point: Nolan Ryan — who was throwing smoke well into his 30’s).
Something that just dawned on me is that Fangraphs does not seperate 2 seamers and 4 seamers, which could affect the results, if you use fangraphs data. It seems to me, in their late 20’s and early 30’s, a lot of pitchers sart moving toward using a 2-seam fastball more. That of course is something pitch f/x would be better suited toward seperating, but I agree with Graham, the sample is too small both in number of pitchers, and in total pitchers for the 2008 bucket.
By pitchers, I mean year to year end points.
I guess the Giants management wasn’t aware of this when they were stupid enough to offer Zito that ridiculous contract they gave him.
Forgive me for asking something totally off-topic, but [deleted, totally off topic]
Zito’s fastball was already gone when they signed him. The scouts should have known that. Zito doesn’t have good control, and at 88/89 mph (as in, when he was good), it was all about deception in his delivery, and his big hook. Fangraphs has his velocity at 85.8 in his last year with the A’s. The Giants gave $$$$ to a pitcher that was in rapid decline.
[email moderation complaints please]
If you think it’s too early, that’s perfectly fine — the thing I want to focus on is that if you want, say, two, three more seasons of data, we can wait and get it. It’s only a matter of time and sample size now — that’s what’s amazingly cool about this whole thing, that this data set exists, is accessible, and can give us things like this.
The other thing I’d like to see is how movement ages. Do pitchers start throwing with more movement (2-seamers or otherwise) as they age? The problem there would be a selective sampling bias, as in, pitchers who are able to offset decline survive, those who don’t, don’t survive. I think you definitely want more years before drawing firm conclusions about any of this, but agree with DMZ, this is cool stuff, and I feel more enlightened than I did before.
This article is like one of those moments in science where someone says, “Because we finally have the ability to collect and process data about X, we can now tell you something about X we never knew before….”
Yeah, maybe we need 10 years of longitudinal data. But at least we have a beginning.
27- Along with movement increasing as pitchers age- What about movement over the duration of the season. I wonder if some pitchers peak and digress throughout the course of a season while others remain more constant. Lots of good info. Who knows what it means yet.
It can go both ways.
Some pitchers learn new grips that generate more movement and it can happen at any random point of their career, because it can be something they learn from a coach or teammate, something they teach themselves over the winter, or simply the natural progression of a pitch they already have.
On the other side of the coin, some pitchers might decide to sacrifice movement for better control. When Jake Peavy first came up, he threw a fastball with a ridiculous amount of movement but there were days when he struggled to locate the pitch (by Peavy standards). He centered his grip, refined his control, and slowly worked the movement back into the pitch. His fastball still doesn’t move as much as it once did, but he has reached a near perfect balance of speed, movement and control.
Felix currently has a similar issue. He gets a ridiculous amount of movement on his entire repertoire, but in time, I believe he will either refine his control with the same amount of movement or sacrifice some of the movement for better control. If you could plug intangibles into an excel spread sheet, perhaps it would go something like: (Ability * Experience) / Cowboy Mentality.
For Mariners starters, this should be interpreted not as “age 29” but “pitch 29” or possibly “game 29”.
Y’know, combining this article with the Beane interview, I find myself wondering: did he accidentally stumble onto a formula that works out of necessity? In other words, his pitchers are usually gone to other organizations by the time they hit the velocity drop years. So, when he has more money in Fremont and can retain more of his players (as he suggested he would in the interview), will that actually work against him? I had that thought as I was reading the interview, and this article only seems to bolster that idea.
Actually, I think it’s if they see the number 29. Beltre can’t ever turn his back on the pitcher…
Absolutely. Of course, you can always draw conclusions on data, even small sample sized, but the confidence level will be pretty low. For me, the really cool thing will be when we can start seeing prototype aging curves, e.g. the Nolan Ryan curve, or the Greg Maddux curve (or the Barry Zito curve for that matter, but alas, we probably can’t name them after Ryan and Maddux since the data won’t cover their careers – maybe the Felix curve and the Verlander curve then).
Making a wild-assed guess, I would bet there will be certain forks in the road where pitchers go one way or another.
Ding! Ding! Ding!
That’s an absolutely fabulous insight. As you said, need more data, but this has so much common sense, yet we weren’t able to quantify it (or come close to it) until now.
The bulk of Beane’s success hasn’t been one particular theory of talent evaluation or another. It’s really been about identifying and exploiting value – what do other teams undervalue (sign them, they’re a bargain) and what do teams overvalue (trade them, they bring a lot in return). Buy low, sell high. Sure, the buy low part was forced on him by his team’s budget woes, but unlike GMs in other financially strapped markets, he didn’t assume he had to buy low-quality versions of the same things the rich teams bought. FA pitchers have been horribly overvalued by most teams, regardless of aging curves, so it was a natural fit for Beane’s value method.
The real test for Billy Beane will be whether he can continue to find value as the rest of the league catches onto and eliminates the market inefficiencies Beane first exploited.
And it’s free!
I think we already have examples of this happening. The Haren trade with Arizona is a perfect example where he wasn’t necessarily exploiting market efficiencies (batting average preference vs OBA, pitchers who get outs vs. power pitchers, athletes vs. winners) but simply proper talent evaluation and the willingness to pull the trigger. Arizona’s GM Byrnes is very well respected and he obviously made a trade that will benefit his team this year (and probably the next three) while the A’s made a very profitable trade because they got a lot of high talent players.
Saying that Beane is exploiting ‘market inefficiencies’ is a little simple in my opinion. He simply does a better job identifying players that will be successful. In his ‘system’ or any other system.
Also, simply because the players fastball velocity decreases at the age of 29 (small sample size!) does not necessarily mean they are anymore inferior pitchers.
So far…. 😛
But I do think the bulk of his success was exploiting the market inefficiencies. He didn’t have a markedly different talent evaluation mechanism for, for example, Mulder and Hudson, he just knew they weren’t likely to be worth the money they were going to cost in free agency. He also wasn’t under any misconception that guys like Scott Hatteberg were superior talents, just that they could be made into useful pieces of a team at minimal cost. From a talent standpoint, I’m sure Beane would’ve rather kept Giambi Major at 1B in 2002, but knew that the talent differential between Giambi and Hatteberg was much, much less than the salary differential when Giambi hit free agency.
True, especially considering a fastball is not much of a weapon if you don’t have anything to mix in with it (c.f Putz, JJ, circa 2004). FB velocity is just one data point in the evolution of pitchers. Like hitters who compensate for lost bat speed, pitchers can compensate for lost velocity (though like hitters, I suspect certain types of pitchers do better than others, a la my comment about prototype curves above).
Likewise for Beane, just because other GMs catch onto his old tricks doesn’t mean he can’t come up with new ones to compensate. But, like a pitcher losing speed on his FB, Beane will need to refine his approach.
I’m w/Graham. It’s really cool that Josh is doing this, and his pitch FX cards blew me away the first time I saw ’em, but this strikes me as highly questionable.
Be careful about things like this:
“Because we finally have the ability to collect and process data about X, we can now tell you something about X we never knew before….â€
Yes, even a small sample size can tell you something, but you can’t be sure it tells you anything *correct*.
To me, the conclusion is flat-out bizarre. It would be quite a thing if it’s eventually corroborated in a 10 year study or whatever.
None of this is meant as a jab at Kalk, who’s done more work with these data than anyone. I’m just not buying it, even as more of a ‘don’t apply to individual pitchers’ sort of a thing.
Just to be clear, the finding that I think is bizarre isn’t the FB velocity declines after 29 – that’s more of a ‘sun rises in east’ sort of a thing – it’s that it increases before that. Chris Miller’s brief study makes more intuitive sense to me, although even there, I’d expect the peak and decline to occur very early. Of course, this has nothing to do with pitcher success.
Hi everyone. I am glad that you generally enjoyed the article. As I pointed out, this is far from precise science but the amount of data is more than most people think. This is why I put the statistical error bars on the bins. You can visually see the range that the correct aging curve should land. You will notice that all of the years between 26 and 29 the data is far above the baseline even if you assume the lower range. There might be systematic errors like using the data from the end of 2007 to compare against 2008 which could still have a big effect but the statistics are really there for the middle ages at least.
I agree with you Marc this isn’t what I expected to see from the data at all. I was shocked when I saw the results. It is absolutely counterintuitive but that is what the data is saying. It was very counterintuitive when Voros found that pitchers don’t really control BABIP yet now pretty much everyone accepts that. You shouldn’t disregard the results just because they aren’t what you expected.
Also, speed of the fastball absolutely does have an impact on pitcher success. You are right that it is far from the only thing that matters but I am currently studying that too. If you check my last two THT articles you can see what things make up a good curveball and slider.
Thanks for stopping by and giving us the additional insight, Josh; I think you did a good job of saying “hey, there’s some reasons not to necessarily think this is the be-all and end-all just yet, but you might find this interesting anyway”.
Unfortunately, my fastball speed curve dropped straight down many, many years ago. 🙁
“You shouldn’t disregard the results just because they aren’t what you expected.”
Wouldn’t dream of it, but we’re Mariner fans. Small sample sizes have produced such wonders as Willie Bloomquist, good MLB hitter. You’ve acknowledged it, but the end-of-07 vs. beginning-of-08 thing could be big. The sample size at each single year of age seems like it might be susceptible to swings if a pitcher moves from a 4-seamer to a 2-seamer, or is just looking off like Verlander.
Also, did you correct for innings pitched in a career or season, too. I’d expect Felix’s FB to be down from where it is now when he’s 29, whereas someone like JJ Putz would be stable (or increase).
To be clear, I wasn’t saying that fastball velocity isn’t correlated *at all* with success, but that you can lose velocity and still be improve. That is, that a decrease in velo need not cause a drop in runs prevented or whatever.
Finally, thanks again for the study and for coming over here to discuss it Josh! Keep up the great work.