Pitch sequencingby Josh Kalk
February 10, 2009
|Being a craft vet, Andy Pettitte can throw just about any pitch. So what should he throw after a slider? (Icon/SMI)|
Pitching has been called an art and that really is true. One thing that continually amazes me looking through PITCHf/x data is how many pitchers with amazing stuff have rather mediocre results and how many pitchers with relatively ordinary stuff have very good results.
Throwing 96 mph with a devastating curveball is great, but if you can't throw strikes or if you are tipping your pitches, major league hitters will eventually notice and you will be sunk. We are starting to get a handle on what attributes correlate to success (e.g., speed differential between a fastball and change-up and a slider that moves away from the hitter) but what about more artistic quantities? How much better will a pitcher do if he sets up the hitter properly by throwing a fastball in then a slider low and away? With good pitch recognition and the sheer volume of PITCHf/x data, we can begin to look at some of these questions.
Last week, we looked at how much better a curveball does when it is preceded by a high fastball. We found that changing the hitter's eye level has a big impact. This result backed up conventional pitching wisdom, but major league pitchers don't throw a lot of curveballs and precede their curveballs with high fastballs even less. So while this result was interesting, it isn't very powerful. But now that we can look at pitch sequences, we can generalize the procedure and calculate the benefit, or harm, from a certain pitch sequence.
We are going to use the general template from last week, this time asking for all sliders that follow a change-up and every such combination. Again, we will use a slightly modified measure of runs100 that normalizes data depending on the count. This is important because most off-speed pitches are thrown later in the count, many times when the pitcher has a big advantage.
What we will be calculating is the normalized runs100 for a certain pitch compared to that pitch when it is preceded by a particular pitch. So last week we found that curveballs preceded by high fastballs decreased a pitcher's runs100 by 0.62 (remember lower is better for runs100 just like ERA). We are going to look at every combination for the following pitches; fastball (fb), sinker (si), curveball (cv), slider (sl), change-up (ch), splitter (sp), and cutter (cu). We will be ignoring knuckleballs because most knuckleballers only throw a fastball and a knuckleball. Because PITCHf/x tracked almost 700,000 pitches almost all of these combinations had very large statistics but some of the combinations that included the splitter do suffer from low statistics. When that is the case, a star will be placed by the combination.
What follows is a list of the seven pitches and how each of the seven pitches fare when preceded by the pitch we are looking at.
Everything works off the fastball. While the fastball grades out as the worst pitch by runs100 standards, any pitch that you throw after you throw a fastball will be enhanced. You can't say that for any other pitch. This shows the importance for a pitcher to establish his fastball. If he can do that, he has the hitters right where he wants them.
While the high fastball-then-curveball combination is excellent, any fastball followed by a curveball is nice. Curveballs gain a bit more oomph than sliders and that makes sense. Fastballs tend to be more over the top and curveballs tend to fall off, mimicking the fastball for the beginning part of the flight. The slider slides off, which doesn't match the fastball quite as well.
I was very surprised to see how relatively poorly change-ups did after fastballs. especially considering how much splitters were after the fastball. Splitters and change-ups have somewhat similar movement and both seem to have the same approach; get the hitter thinking fastball and throw him something the looks like the fastball but is about 10 mph slower. In both cases, pitchers are throwing most of their change-ups and splitters after a fastball, so I don't have a good explanation for what is going on here.
A fastball followed by another fastball isn't a bad combination either, but a fastball followed by a sinker is even better. Pitchers like Fausto Carmona who throw both a fastball and a sinker seem to have a real advantage here. They can come back with a pitch that is very similar to the one they just threw but slightly different. The hitter might make contact but he isn't likely to square the ball up because of the differences in the pitches.
Cutters, however, don't seem to receive much help. Cutters tend to look like fastballs but move away from a similarly handed batter. This might be causing a hitter who was jammed by the fastball to now get the barrel of the bat on the ball on the next pitch more often.
sp 1.47 *
Sinkers, like fastballs, generally improve pitches that follow. As with fastballs, a sinker followed by another sinker is good, but a sinker followed by a fastball is even better. The best off-speed pitch to follow a sinker with is a slider. This shouldn't be any surprise when you think about all the sinker/slider pitchers out there. These two pitches work very much the same way the fastball/curveball combination works, only now it is the horizontal axis that is being lined up.
Curveballs are much more rare than sliders for sinkerballs, and when they are thrown the results aren't great. Sinkers naturally have less rise than a fastball, so the differential between the curveball and the sinker is much less than that between a fastball and a curveball. Change-ups, too, lose some potency and this is likely for the same reason. The hitter has just seen something moving downward, so when the change-up drops it doesn't have the same effectiveness.
Lastly, cutters and sinkers have a good deal of synergy. Cutters, like sliders, tend to be on the same horizontal plane as the sinker but the cutter is also thrown at a closer speed to the sinker. Pitchers like Roy Halladay and Chris Carpenter have gotten a lot of mileage from this combination.
Curveballs, like other off-speed pitches, tend to be a mixed bag for the next pitch. Going back to the hard stuff doesn't produce great results, especially with a sinker but most of the soft stuff isn't too good either. This makes some sense to me as curveballs tend to be thrown later in the count. If the curveball doesn't get rid of the hitter the pitcher probably has used most of his pitches already. The main exception appears to be curveball/curveball, which appears surprisingly good. Hitters must not be expecting a second curveball. Maybe they got in a hole early and then when they laid off the first curveball they were expecting the pitcher to come back with something hard.
Sliders are much like curveballs with one notable exception. Both sinkers and cutters benefit from the pitcher having just thrown a slider. Cutters are mostly harder sliders that don't move quite as much and we have already looked at the good fortune that sliders get from sinkers. Sliders also tend to do well back to back though not quite as good as back to back curveballs.
sp -1.72 *
Not only do change-ups do worse after a fastball than I would expect, but they also don't appear to give pitchers a good choice as to where to go next. The trend does continue, however, as a change-up/change-up is rather productive. Because splitters and change-ups tend to serve the same purpose, few pitchers throw both. Manny Parra is an exception but even he doesn't use his change-up to left-handed batters, so this sample is by far the smallest of the study.
si 0.41 *
ch 0.59 *
cu 1.81 *
Splitters can be absolutely devastating pitches despite the relatively large wear they can do to an arm. Splitters tend to move just like fastballs horizontally but drop severely when they approach home plate. Many pitchers use their splitters exclusively in pitcher counts, but when that doesn't get the hitter out, the choices as what to throw next are severely limited. Splitters are the worst pitch as far as setting up the next pitch, though coming back with a fastball definitely appears to be the best bet. Splitter/splitter is the first back-to-back combination that produces negative results, though I can't say why splitters behave like this when other pitches don't.
sp -1.28 *
Cut fastballs like the one made famous by Mariano Rivera move away from a similarly handed batter, so many pitchers will use this pitch to bust an opposite handed batter inside. The classic followup to busting a hitter inside is soft stuff that moves away and in this case that generally means change-ups. Curves and sliders will move in to a similarly handed batter but the change-up will move away and can be started on the outside corner only to move off the plate. Of course, another alternative is back inside with the hard stuff. Throwing another cutter also is a good choice.
As you can see, certain pitch combinations are far superior to others and a pitcher who just throws his pitches at random is not going to be very successful. Setting things up with a fastball is the way to go for a pitcher. Pitching backwards has become somewhat in vogue recently, but very few of the off-speed-then-fastball combinations are very good. Once you have shown your best off-speed pitch, it is hard to come up with a good plan for later in the count.
Throwing the same pitch consecutively was far better than I would have thought. These combinations tend not to be used very often and probably should be called more often.