The effect of seeing pitches

With the World Series in full swing, there has been ample opportunity to analyze in great depth the decisions of the two managers. A popular topic has been the reluctance to pull starting pitchers as they are set to face the opposing lineup for the third time, with well-rested relievers waiting in the pen. I have written before about the typical approach and success rate for pitchers across times through the lineup, and MGL has most recently highlighted the relative penalty expected for starters as the game progresses regardless of the pitcher’s performance to that point in the game.

While we know that each time a pitcher and hitter square off in a game, the hitter gains more of an advantage, can we see this at a more granular level? To do so, I have used PITCHf/x data from the past three seasons and broken down batter performance based on the number of pitches seen from the same pitcher in a given game. Does offensive performance improve on each additional pitch seen by a hitter?

In addition, while I could believe seeing any offerings from a pitcher could help a hitter pick up timing and release point information, it seems to me that a hitter might gain at least part of this advantage through seeing a particular pitch type multiple times. In other words, if someone threw me four pitches to look at and asked me to hit the fifth (aside from the fact that I could not touch any big league offerings) I would have a much better chance of hitting the fifth if all five pitches were fastballs than if I got to watch four fastballs and then was thrown a curveball. I would have no prior experience in this game with the kind of velocity and movement that accompanies a curveball, so the first time I see it would be baffling.

This is actually what I wanted to examine the most in this study—the typical improvement experienced by batters for every instance of a given type of pitch that they are thrown by a given pitcher in a game.

To start, consider the following graph showing the weighted on base average (wOBA) of hitters per pitch seen as well as per pitch of the same type faced from the same pitcher in the same game. In this case I have collapsed all pitch types as classified by MLBAM into an “average pitch” to determine first if a trend is visible.

From the graph (which is zoomed in so presents a non-zero starting point for the y-axis!), it is clear that hitters are building on their levels of success upon each occasion that they face a particular type of pitch from a pitcher within the same game. Also presented is the wOBA based on all pitches seen, which shows a steep decline until the third and fourth pitches before also following a rising pattern. I have seen this pattern exposed previously, with one example being David W. Smith’s work. The decline is really just a result of batters not being able to strike out until at least the third overall pitch that they see, so by definition any plate appearances that end within the first two pitches are at the whims of BABIP (or else a home run or a hit by pitch). Taking all strikeouts out of a set of possible outcomes clearly drives expected offensive output in a positive direction.

So we now know that in general, hitters improve their fortunes every time they face a given type of pitch from a particular pitcher in a game. The next question is, which pitches tend to hold up the best over the course of a game, and which fare the worst? I split the study by MLBAM pitch type, and calculated the typical increase in wOBA per additional pitch thrown again.

wOBA increase per pitch to given hitter from given pitcher in given game, by pitch type (2011-2013)

I can make a few observations about these results. The first is that in general, breaking balls and offspeed pitches decay at a faster rate than fastballs. Overall, “soft” pitches as a group show an average .006 wOBA increase per pitch to the same hitter, while “hard” pitches as a group typically inflate wOBA by .003 per pitch. This two-to-one ratio lines up with standard pitch usage, which over the past three seasons has been 64 percent “hard” and 36 percent “soft.”

The results also speak to the reason that starting pitchers typically need to have a more varied repertoire than relievers. Facing the same hitters three or even four times in the same game, starters must routinely throw a dozen or more pitches to the same batter. A more diverse arsenal allows a starting pitcher to architect pitch selection to keep the number of instances of each pitch type lower than if he had only two pitches to mix, lowering the expected wOBA- against. In effect, varying pitch types allows a pitcher to move along the lower red wOBA-against line in the graph above as they face hitters multiple times rather than the blue wOBA against line if they were forced to throw a solitary pitch type again and again.

The curveball that baffled me in my example above appears to be the type of pitch that professional hitters adjust to the quickest, outpacing the other secondary offerings. While not included in the table due to small sample sizes, both the knuckle-curveball and knuckleball demonstrated the best performance over repeated use, with each of them posting negative wOBA trends over the first several instances within a game. The knuckleball, of course, is an extreme outlier pitch type, as it tends to be a pitch that a hitter would see none of in nearly every game and then practically nothing but when facing a knuckleballer.

One final point: while I have published the expected wOBA steps per pitch above, each pitch type has its own initial wOBA-against for the first time it is thrown. Although breaking and offspeed pitches rate poorly in terms of rapid loss of performance per instance used to a given hitter, they still in general fare much better than fastballs in absolute terms. There are likely many reasons for this, including the fact that pitchers will tend to turn to their fastballs when in hitters’ counts to get one over but offer up a secondary pitch out of the zone when ahead in the count.

The times through the order penalty realized by starting pitchers is real. Now we can see that each additional pitch makes things generally worse for pitchers, so managers get your bullpens warming early!

References & Resources
Credit and thanks to Baseball Heat Maps for making available the PITCHf/x database upon which this analysis is based.