The Stats Go Marching In: Reaching Back for a Little Extra, Part Two

In my previous installment, I explored pitch speeds in several situations and discovered that pitchers can add some gas to their offerings in certain spots. Both here at Baseball Prospectus and at The Book Blog, readers made insightful comments on the subject, suggesting possible biases and ways to expand on the analysis.

This time, I’ll go over some of those points.

​Starters and relievers
As we have reiterated many times, starters, who have to pitch several innings at a time, approach the game differently than relievers, who might be on for as little as one batter. Starters need to hold something back when they have the luxury of doing so, in order to conserve their strength for the long run. That being the case, repeating the analysis from part one for separate groups should lead to higher values for starters.

Last time we saw that pitchers go full speed in two-strike counts, while being more careful when behind in the count (and also on the first pitch of the at-bat.) The split analysis shows the same results. Starters and relievers follow a very similar pattern, but the former are able to increase their speed more.

That’s what we were talking about just before the table. Relievers are always throwing close to 100 percent, so they don’t have much of a margin for dialing up their velocity; conversely, starters who pace themselves throughout the game have more room to maneuver.

If more confirmation is needed, look at the 1-0, 2-0 and 3-0 counts. The decreases are higher for relievers, since they are the pitchers who have more room on the opposite side of the speed spectrum. (I suppose you can’t hold back too much on 1-0 and 2-0 counts, or a tentative fastball can become a fat pitch.)

For completeness, below is the comparison by base/out situation. Once more, the pattern is quite similar for the two groups, but the magnitude is higher for starters.

One more thing on the starters/relievers issue. One reader pointed to a possible source of bias that could be leading to the results in the table above.

Suppose a pitcher comes back out for one more inning, but it becomes apparent that he doesn’t have it anymore. He has suddenly lost some speed and, as a result, he can’t retire any opponents. He is removed before getting anyone out in the inning. Conversely, if the starter still “has it,” he will stay in the game and put batters away.

This situation could artificially lead to the observed effect of pitchers throwing harder as they put opponents away. My gut feeling (yes, we statheads have gut feelings too!) was that the effect of this would be minimal. However, I repeated the analysis of part one anyway, including only completed innings (that is, innings where no pitching change occurred). The results were unchanged.

Flamethrowers and control artists
Hard throwers like Stephen Strasburg and David Price deliver their heaters in the mid-to-high 90s, while soft tossers like Barry Zito and Mark Buehrle usually pitch in the 80s. Does a situational velocity difference similar to the one we saw between starters and relievers apply to hard versus soft throwers?

I limited the analysis to half the starters in the 2010-11 PITCHf/x database. I labeled the fastest 25 percent of pitchers as the flamethrowers (the cut-off was around 93 mph of average fastball speed) and the slowest 25 percent as the soft tossers (the line here was around 89 mph.) The middle half was discarded from this analysis.

Below are the usual two tables, by count and by base/out situation.

The differences are quite small across the board, but nearly all of them are in the same direction—higher for hard-throwing pitchers.

It seems that flamethrowers can work (or need to work) on a larger speed spectrum. I’m going to use Justin Verlander’s words (from David Laurila’s interview at FanGraphs) to address this issue:

I’ve learned how to slow down—how to slow my velocity down. I’ve always been able to amp it up when I need to, but the hardest adjustment was to slow it down and locate. It’s a location and pitch-saving thing. As a power pitcher, you get a lot of foul balls and not a lot of contact. Guys don’t put the ball in play all that much. When it’s 90-91 [mph], guys tend to put it in play a little more. If I hit my spot, they put it in play weakly. That‘s the biggest adjustment I‘ve made.

Batters’ discipline
In part one, I found that pitchers were likely to use some extra speed when facing their most productive opponents. However, as was suggested in the comments, I did not consider the opponent’s selectivity.

I decided to explore the matter again in three different ways.

First, as I did a few weeks ago for batters’ production, I compared the speed pitchers used against a certain hitter with his proclivity for swinging at pitches outside the strike zone. While a mild correlation existed between run production and speed of pitches seen, none emerged between tendency to chase and pitch velocity.

Then I inserted the discipline variable (the percentage of swings outside the zone) into the regression model. Again, it did not come out as having an effect on pitch speed.

Finally, I repeated what I did above for hard versus soft tossers. Running split models on the 25 percent most-disciplined hitters and on the 25 percent least-disciplined hitters, I did not find any difference worth mentioning.

So, while the difference by count seem to indicate that pitchers in some instances might be willing to sacrifice control in favor of velocity, they do not seem to alter their behavior according to the opponent’s discipline (or lack of it.)

However, percentage of swings outside the zone might not be the best choice for looking at this issue.

In fact, as I have investigated in a few trivial articles in the past (see the Yogi Berra’s award series at The Hardball Times), there are hitters who can do some damage even on pitches way out of the zone. Thus one would probably want to control by production on bad pitches rather than tendency to swing at them.

Exchanging speed for control
I’m leaving this issue open, at least for now. I’m positive that with some good effort, we can come up with some decent estimate of the ability of a pitcher to hit the intended spot, even if the intended spot is unknown. But that would not be attainable on a pitch-per-pitch level (and that is what we need for evaluating how much control, if any, a pitcher has to surrender in order to throw harder.)

Sportvision has COMMANDf/x data going back to 2010 at least, but they have not released it for public use. Measuring intended location is not a perfect science, though I think teams could conduct experimental studies in bullpen sessions if they had PITCHf/x cameras installed in the pen. Using the target set by the catcher’s mitt as the indicator for intended location (as COMMANDf/x does) is far from perfect, but judging by the brief analyses I conducted on those data for the 2011 PITCHf/x Summit, I think it can be a reasonable proxy.

Thus, my take is that with COMMANDf/x data, a first rough measure of the tradeoff between speed and control can be calculated. Maybe Mike Fast already working on it.