The Hardball Times:How fast should a fastball be?
6
|
The Hardball Times found this 26 days ago on www.hardballtimes.com [flag] |
Tags:
MLB
Comments (26)
-
-
walshj58 Thanks, Justin. I started to look at errors, but I didn't come up with satisfactory results in time for this piece. It's certainly something that should be added to the analysis.
I don't know how precise you can get with individual pitchers, but if you're going to try to deal with selection bias, which some readers have mentioned, you'll have to compare pitchers to themselves, eventually.
-
-
David Gassko John,
Please quit your job and focus on writing about baseball. This is absolutely fantastic work.
-
MGL Yup, I am apparently not the only one who thinks that THIS is finally (not that the other stuff was bad - it wasn't) the kind of fantastic stuff we've all been waiting almost a year for. Anyway, see all my comments on The Book blog, www.insidethebook.com. -
ekogan What about selection effects?
It seems that on the lower end of the velocity scale only those pitchers that have enough extras (deception and/or movement) on their fastballs that their fastballs don't get hit hard in at least some part of the strike zone would survive in the majors, so you would have a big selection effect on the outside corner.
Also, what is your definition of the fastball? Does it include any varieties like cutter, sinker, 2-seamer, which ought to be harder to hit that a 4-seamer of the same speed?
-
walshj58 You make some good points. I tried to address the first point in the article: I attempt to look at pitches with similar movement by considering only fastballs. Not perfect, as I wrote, but it's a start.
Deception is a whole other beast, of course, which I also mentioned in the piece. I have no idea of how to measure deception.
My fastballs include sinkers and some fraction of cutters (some cutters are labeled sliders by my classification scheme). Again, that is not perfect, but if you slice and dice too much, you really run into sample size issues.
-
-
philly Fascinating article and excellent presentation of the data.
However, I'm confused about the 3rd conclusion. The conclusion that velocity doesn't matter in terms of HR percentage looks to be true for the middle right box, but down and away 80-88 mph fastballs look noticably higher than the other three intervals and up and away the slower two intervals are higher than the faster two intervals.
Is it that the middle box is a much larger sample such that it overwhelms the other two boxes?
-
walshj58 Thanks.
I was painting with a broader brush than you, not so much looking at individual points, but trying to see the general trend.
My general point is that the dependence on speed of HR pct is much greater for inside pitches than for outside ones. The 3 outside zones don't show perfectly flat behavior, but they are much flatter than the 3 inside zones, that was my point.
-
-
Redsauce Looking at the 2nd graph, with the full plot as the strike zone. It looks (to the naked eye) like the lower left zone shown in graph 1 matches up with the middle left side of the full zone, not the lower left side. Am I crazy?
Also, I know I'm a nerd because this @#$% is awesome.
-
walshj58 Redsauce,
You're not crazy. But, neither am I: the plots are correct :-)
Plot 1 looks different in the 3x3 plot, because I've combined pitches into fewer bins and the far-right point in Plot 1 has very few pitches, so its low value sort of gets washed out in the second plot. Does that make sense?
thanks for the kind words.
-
Redsauce Yup, that makes perfect sense. Thanks very much for the explanation.
-
-
-
GuyM Really fine work, John. I had the same selective sampling thought as ekogan, that guys who can survive with 80-89 mph heaters ("lukewarmers"?) probably differ in important ways that could impact these results, whether it's movement, changing pace, and/or pitch sequence. One way to deal with this is to put pitchers into, say, three groups based on pitcher's average FB velocity, and then re-run the analysis for each group. The speed variation will of course be less for each group, but might still be sufficient to learn something. It looks to me like you could also improve sample size by looking at three zones -- inside, middle -- outside, rather than 9 sectors (for this analysis; presumably not for other questions).-
walshj58 Thanks, Guy, for the suggestions. I will add them to my list of things to look at for the next iteration.
-
-
DisabledListInformer Awesome work with this article, John. It's no wonder that guys like Jamie Moyer can hang around for as long as they have, just nibbling the corners with the slow stuff.
I wonder how this data reflects the performance of stud pitchers who are struggling early on (Sabathia, Verlander, in particular)?
-
walshj58 Thanks.
You need a lot of data for this kind of study, so looking at a few starts for individual pitchers will not be enough.
-
-
notsellingjeans The future of a Major League front office:
A forward-thinking GM budgeting $500K annually to hire 8 guys do this type of analysis with Pitch F/X data on literally every major league hitter and pitcher (and elite minor league ones, too). These take several hours to do for each player, so you'd need a team of researchers (John Walsh, Sal B., Dan Fox, Josh Kalk, and Mike Fast would be an excellent start).
That would take scouting the opposition to a new level. And the first team to do it will reap benefits far exceeding their $500K investment.
Terrific stuff as usual, John.
-
GuyM $500K would probably get you two or three guys, not 8 (if you figure salary, benefits, workspace, equipment, etc.), but the point stands.
Dan Fox, btw, just went to work for the Pirates.
-
-
obsessivegiantscompulsive Terrific stuff John, keep up the good work! While I'm not saying that it is on the same level of importantce as Newton or Gallileo, I think that there was a lot of the work that they did that can be considered "duh" discoveries of experential observations of nature but that was verified by their science. Similarly, the experential observations of baseball is great knowledge to be had, but the growth of sabermetrics is directly because not all those observations are correct or optimal. Data and science will help us figure out which ones are "duh" and which ones are "Eureka!" Obviously, as you noted, the latter ones are the more exciting ones, but like a nice 3-hit shutout, the former ones are exciting in their own way too because they do confirm our observations. I look forward to your next article (though the dots are still dancing in my head :^). -
stopworth I think the BABIP graph is the most interesting part--it's a shame you left it out of the main article. The fact that BABIP drops 100+ points for slower pitches when thrown high- and low-away seems really remarkable. It really seems that people like Glavine who make their living on hitting that outside corner are actually better off putting less on their fastball.
Selection effects might have something to do with this, as previously mentioned, but a variation of this size would seem to throw something at the BABIP = "mostly just luck" hypothesis.
Are the differences there statistically significant?
-
walshj58 These are good observations -- I sort of gave babip short shrift here, in part because the piece was running long.
I checked the statistical significance of the first two points in the low-and-away zone: I get the difference in babip of those 2 points at 57 +- 24, which puts it as marginally significant.
If I consider the first and last points of the up-and-away zone, though, I get a babip difference of 120 +- 25, which appears very significant.
Pitch-f/x will shed a lot of light on the nature of babip, I think, once we start asking the right questions and getting down to answering them.
-
-
edward471 "You make some good points. I tried to address the first point in the article: I attempt to look at pitches with similar movement by considering only fastballs. Not perfect, as I wrote, but it's a start.
Deception is a whole other beast, of course, which I also mentioned in the piece. I have no idea of how to measure deception.
My fastballs include sinkers and some fraction of cutters (some cutters are labeled sliders by my classification scheme). Again, that is not perfect, but if you slice and dice too much, you really run into sample size issues."
I'm surprised that someone with your physics (scientific training) background fails to understand the real danger of making ignoring these (and other) variables.
Each pitcher-hitter confrontation is a unique event. Attempting to say that a low outside fastball by a specific pitcher to a specific batter in a specific game with a specific count somehow averages out such that the only dependence is this be the pitch on the outside corner is in my opinion not realistic. In other words the process is not linear and therefore is subject to the peculiarities of nonlinear systems. Which I don't believe is captured in your analysis.
For example take the Yankees pitcher Chen-Ming Wang who throws essentially one pitch, a 92-95 mph sinking fastball. If on a given day he does not have movement, his 95 mph fastball on the outside corner has a greater chance of being hit safely than if he does have movement on is 95 mph. His 95 mph fastball is chaotic in behavior because Wang nor his coaches understand exactly why he does or does not have movement on any specific day.
Or in the case of Greg Maddux, if he does not have movement and location is 86 mph fastball has a much greater probability of failure.
Attempts to use statistics (sabermetrics?) such as you have done in this exercise does a great disservice to the "real" word of scouting and player development as in how is one to actually use and apply this information??
-
stopworth Gotta disagree with you here--I think the consequence of what you're talking about is the extent to which inferences can be drawn from this data about future events.
Certainly, I don't think anybody would advise any pitcher to just throw fastballs down and away all game. It wouldn't take long for hitters to adjust and without any guesswork, they would undoubtedly start getting hits in bunches.
That said, looking at data like this seems to have real use in that it provides insight into statisical variations in hitters' abilities to handle similar pitches at different speeds and in different areas of the strike zone. This statistical variation provides grounds for a number of new hypotheses, some testable, some not.
I am of the opinion that this doesn't to a disservice to anyone, really--quite the opposite--and perhaps with similar types of analyses, Chen-Ming Wang and his coaches can figure out what it is that accounts for the movement on his fastball.
-
walshj58 I didn't say that the things I wasn't able to control average out, I wrote that I must assume that they average out. There is a salient difference there, which you seem not to have appreciated. Any kind of study such as this needs to make some assumptions, otherwise the study itself would be impossible. I have stated clearly the assumptions (that were necessary) and each reader is free to accept them or not. You do not accept them, fine -- but it has nothing to to with scientific integrity or failure to understand.
As for your examples, Wang's first name is Chien-Ming and he does not throw "essentially one pitch", but throws almost 20% sliders and 10% changeups. I suppose that "real world" scouts and player development personnel would know that.
Furthermore, I have not seen any evidence that "movement" varies from start to start. We know that velocity is fairly consistent from start to start and, from my observations based on the pitch-f/x data, movement is also consistent.
Also, this:
In other words the process is not linear and therefore is subject to the peculiarities of nonlinear systems.
is a very bizarre comment. What linear system or model have I employed here? This sounds just like somebody trying to say something smart, but it, in fact, is totally irrelevant.
-
Redsauce Booya!
-
-
-
Digital Headbutt Excellent article. Great work, John. -
alannathan Excellent piece of research!
Links (6)
THT: Walsh: How fast should a fastball be?
Published 25 days ago at BBTF's Baseball Primer Newsblog
THT: Walsh: How fast should a fastball be? “Which is the better pitch: an 87-mph fastball on the outside corner or 96 down the middle? John has a look using (of course) PITCHf/x.” The first plot shows the percentage of swing-and-misses relative to the number of swings. We expect a greater percentage of misses as speed increases and we see that for both the low-and-away and the middle-in pitches. The trends aren’t that startling, although it appears that throwing above 94-mph really helps, especially on the inside pitches. Naturally, pitches on the outside portion of the plate ...
Do fastballs need to be fast?
Published 25 days ago by Bjoern at At Home Plate Blog
... that a large part of the dependence on speed for inside pitches comes from the ability of hitters to drive the slower pitches out of the park. Outside pitches seem more difficult to hit for homers, no matter what the speed.
Absolutely fascinating stuff! Those who have always wondered how guys like Jamie Moyer or Tom Glavine are doing it have to read that article. It also features a lot of graphics that really help understanding the findings.
Link: How fast should a fastball be?
98 On The Black
Published 25 days ago by FrostKing at Frost King Baseball
John Walsh at the HardballTimes wrote a very interesting article about the effectiveness of the fastball. He broke the strike-zone into sections and looked at the result of fastballs at various speeds. A strike thrown down and away from a righty pitcher to a right-handed batter was up first: "Let me say that in plain English: An 87-mph fastball in this part of the strike zone is every bit as effective as a 96-mph heater. " Here's the breakdown for each zone: ...
SAINT ANDREW'S NET: SEXY PEEKS!
Published 25 days ago at With Leather
... What are you complaining about? Your team last won a championship in . . . 1975?!? Ouch. Plus you live in Philly. Lion in Oil knows you can watch the NBA Playoffs with the Nets' coach. I wish the women I'm intimate with would "manufacture a sense of urgency". Best Week Ever presents Tim McCarver: Civil War buff. Let's nickname him "Stonewall" for his IQ level. The Hardball Times asks: How fast should a fastball be ...
Friday Links (25 Apr 08)
Published 21 days ago by Geoff Young at Ducksnorts
... John Walsh at Hardball Times examines the importance of a fastball’s speed (h/t ...
Phoning In a Friday: The Comments
Published 15 days ago by John at Twins Geek
... But, yes, break and changing speeds, as we all know, is important, too. I almost added that little computation to the story SBG. Great minds think alike. I didn’t because the more I thought about it, the less I cared. As a batter, swinging seem to be more about timing than distance. But that’s why I included the 4.5% thing. I’m glad someone brought it up. SL__72 said: And to follow up my last comment with something that is actually useful: Here is a really good article regarding fastball speeds. You’re right SL_72, that was ...
This is freaking fantastic work. And it's largely consistent with conventional wisdom about fastball location and velocity, which makes it all the more convincing. Even so, I can imagine that some pitchers could really benefit from paying close attention to this kind of work.
There's just so many directions you could go with this...One thing that might be interesting to do would be to attach standard error bars to those dots and see the degree to which the variation is consistent. I wonder, for example, if certain locations are big-risk, big-reward areas (i.e. high variation in outcome, "typical" mean outcome), whereas others are more reliable. I guess comparing the home run plot vs. the runs plot gets at the same issue...it tracks pretty well, except for at the top of the zone, and down and away.
Also, do you think that a season's worth of data would be enough to do this kind of thing on individual pitchers? It might be interesting to track frequency of pitches vs. outcomes and see if there are some gross indications that certain pitchers are pitching effectively/ineffectively given their talents.
Again, great work! -Justin