The circle graph is definitely the most intuitive way of displaying the data I've seen although he needs to throw some quantitative (by angle measures?) break x/y numbers on there...that and the pitch filter must suck as I doubt Smoltzie threw a 92 mph slider that ran into a righty.

Repoz, thanks for the link!

The circle graph is definitely the most intuitive way of displaying the data I've seen although he needs to throw some quantitative (by angle measures?) break x/y numbers on there

I'm glad you like it. I'll see about putting some angle numbers on there. I'm still working out the best way to do polar plots in Excel. Good suggestion. The break in terms of x-z is also dependent on the spin rate, and there's just no great way to plot three parameters on a 2-dimensional graph.

the pitch filter must suck as I doubt Smoltzie threw a 92 mph slider that ran into a righty.

I hope it doesn't suck too much since the pitch filter is me. :)

One of the reasons I looked at Smoltz is because I am investigating the slider and what that term is used to mean in baseball today. Some pitchers use it to describe something that moves more like a cutter, and Smoltz borders on that, but when he talks about his own repertoire, he calls it a slider, not a cutter (see http://mlb.mlb.com/news/article.jsp?ymd=20070515&content_id=1966438&vkey=news_mlb&fext;=.jsp&c_id=mlb).

The 91.8-mph "slider" in question was the 1-2 pitch to Chris Duncan in the bottom of the 1st inning in St. Louis on August 24. The PITCHf/x system records Smoltz as being fast in that game. I don't know whether that's accurate or a result of miscalibration, but it has his fastball running 95 mph to start that game, in which case a 92-mph slider is maybe a little more realistic. The Fox gun clocked that pitch at 90 mph. It was definitely one of his more cutter-ish pitches, but it moved much more like his sliders than it did his fastballs in that game, so I classified it with the sliders.

Haha I assumed you were using an algorithm. Maybe the Pitch F/X data sucked then haha.

I'd say the biggest issue with classifying slider vs. cutter vs. whatever is that pitchers "cut" their fastball but I wouldn't call it a cutter. #1 it's really hard to see from the behind the pitcher perspective that they did anything and then #2 just looking at the data it's not going to show up as a distinct cluster as the difference is tiny in numbers but big in missing the barrel i.e. it's going to be a continuum not a real separate pitch. Then you have the different types of sliders (power, slurvy, Lidge etc) and you really end up with a bunch of pitches that will be tough to say 'this is what a slider is' and such just from the data.

Just on my own visual observations, Homer Bailey comes to mind as a guy that really cuts his fastball and he should probably be easiest to detect with the data if it's possible (the Pitch F/X tool didn't register any cutters over at Baseball Analysts. I guess if I'm coming up with my definitions: Carpenter throws/threw the most distinguishable cutter in the non-Rivera division, Smoltz is the definition of a power slider and I dunno Andrew Miller's slider gets slurvy sometimes when it drops into the lower 80s (at least when he was at UNC).

I haven't found any algorithm that works very well unless you baby it a whole lot and are willing to do a lot of adjustments by hand, in which case, I'm not really sure it's an algorithm any more. I don't know if you're familiar with Josh Kalk's work--I'm guessing maybe yes, if that's what you mean by the PITCHf/x tool which Josh has running at Hardball Times. Josh is the only one who's published the results of an algorithm applied to every pitcher with a decent sample size, and I'd say his old algorithm misclassified about 30% of pitches. He doesn't have the player cards up for his new algorithm, but it appears to be doing a little better, so maybe there's some hope on that front. For example, his old algorithm completely missed both the curveball and the split-finger for Smoltz, now at least his new one finds the curveball.

When I do stuff by hand, I can hopefully knock the errors down under 5-10% in the worst case and sometimes very close to zero if the pitcher has only two or three very distinct pitches, like Joba Chamberlain, for instance.

With Smoltz, I'm pretty sure I'm missing the boat on some of his split-finger fastballs. I looked at the data on a game-by-game basis, but in most cases I did not look at individual pitches. If Smoltz threw a hard splitter early in the game, I may have confused it with a fastball from the end of the game where he's starting to lose a few mph. If I go back and check each pitch by hand, I can usually figure that out, or there are some automated methods I can apply to help me, but I don't always dig that deep depending on what I want from the data and how much time I want to spend. I'm also quite sure that I've confused some splitters and changeups since they move so similarly for Smoltz.

I like the idea of an algorithm in principle since who's going to have the time to look at every major league pitcher individually, but in practice the results of an algorithm are a very rough approximation and generally not good enough when you want to evaluate a specific pitcher.

I completely agree with your comments about there being different types of sliders and how many pitchers will sometimes put a "cut" on their regular fastball without throwing a distinct cutter as a separate pitch. That definitely lines up with what I see in the PITCHf/x data.

Thanks for your observations about who throws a distinct cutter. Those are some guys that I can look into. Rivera's cutter is obvious, if only because that's the pitch he throws 90% of the time and he can crank it up to 94. Other than Rivera, Greg Maddux was the only other guy I'd run across so far who threw the cutter as a distinct pitch.

The pitch Smoltz throws to lefties that rides in on their hands is more cutter than slider. His slider to righties has way more downward tilt, whereas the pitch to lefties flatter; not completely flat but more so than his standard slider. Very Rivera-esque. He also occasionally throws the Maddux cutter that starts way inside and comes back over the inside corner against lefties. IMO, that's his least effective pitch, command wise. He often doesn't start it out far enough inside and it runs right out over the middle of the plate. Eep.

As someone who watched pretty much all of his starts last year, I can tell you he went away from the split sometime around midseason when his elbow was acting up (he always stops throwing his split when he's in pain, it's a dead giveaway and has been for years). He threw a lot more changeups as the season progressed.

Your description of the difference between his split and change-up is dead on. Well done.

Corn, thanks for your comments. It's very helpful to have insight from someone who's seen many games by a pitcher and paid attention to which pitches he was throwing.

My initial pass through the data didn't show much difference between lefties and righties in the pitch group I called the slider, except that the lefties didn't see it as often. Since I also didn't see a cutter mentioned in scouting reports on Smoltz or in comments he himself had made, I discounted the possibility of a cutter hiding in there.

However, now that I go back and look specifically for the pitches you are talking about, those that are thrown inside to lefties, I think I may see a signature of a cutter. I'll have to give it a more detailed look, but thanks much for your help.

Given this info, the 92-mph "slider" pitch that was mentioned previously was almost certainly a cutter, and many of the other 86+ mph "sliders" to lefties may also really have been cutters.