If you hadn’t noticed, pitchers get hurt a lot. Throwing baseballs for a living puts a lot of repetitive stress on one’s arm, and every year this manifests into a pile of pitchers hitting the disabled list throughout the season. Most of the work involved in preventing injuries comes outside of a game itself, in the form of conditioning and practicing mechanics.

Within a game, pitchers need to be able to trust their mechanics and focus on the task at hand. It stands to reason, though, that pitchers may seek out any small opportunities within the game itself to save energy. This reserve of energy then can be used in situations where it may provide the most benefit to the pitcher in the context of the game.

Pitchers, as a group, appear to do this by adjusting velocity based on the relative importance of the pitch at hand. Pitches potentially leading to a walk or strikeout are one example. Pitches to a batter that could make the final out of the inning are another. While this topic has been covered before, by Max Marchi, for example, in this article, I have calculated the average fastball velocity for a variety of game states to investigate this behavior of pitchers and determine whether it happens consistently.

I have selected the fastest 25 percent of all pitches for each pitcher within each game for this study. I also ensured that each pitch included was tagged as one of the variants of fastball as classified by MLBAM. To determine the consistency of such a phenomenon, I calculated the various velocities for each of the 2011, 2012 and 2013 seasons independently.

Game State | Velocity | |||||
---|---|---|---|---|---|---|

Balls | 2013 | Delta | 2012 | Delta | 2011 | Delta |

0 | 91.12 | - | 90.90 | - | 90.84 | - |

1 | 91.18 | 0.06 | 90.97 | 0.07 | 90.92 | 0.08 |

2 | 91.37 | 0.19 | 91.19 | 0.22 | 91.14 | 0.22 |

3 | 91.48 | 0.11 | 91.34 | 0.15 | 91.26 | 0.11 |

Delta/Ball | - | 0.13 | - | 0.15 | - | 0.15 |

Strikes | 2013 | Delta | 2012 | Delta | 2011 | Delta |

0 | 91.03 | - | 90.82 | - | 90.74 | - |

1 | 91.13 | 0.10 | 90.93 | 0.11 | 90.89 | 0.15 |

2 | 91.65 | 0.52 | 91.47 | 0.54 | 91.46 | 0.58 |

Delta/Strike | - | 0.31 | - | 0.32 | - | 0.36 |

Outs | 2013 | Delta | 2012 | Delta | 2011 | Delta |

0 | 90.92 | - | 90.71 | - | 90.65 | - |

1 | 91.28 | 0.36 | 91.09 | 0.38 | 91.02 | 0.37 |

2 | 91.47 | 0.19 | 91.27 | 0.18 | 91.23 | 0.21 |

Delta/Out | - | 0.28 | - | 0.28 | - | 0.29 |

Runners | 2013 | Delta | 2012 | Delta | 2011 | Delta |

0 | 91.16 | - | 90.95 | - | 90.89 | - |

1 | 91.18 | 0.02 | 90.99 | 0.04 | 90.93 | 0.04 |

2 | 91.48 | 0.30 | 91.29 | 0.30 | 91.22 | 0.29 |

3 | 91.75 | 0.27 | 91.72 | 0.43 | 91.63 | 0.41 |

Delta/Runner | - | 0.21 | - | 0.26 | - | 0.25 |

**Fastball velocity by game state, 2011-2013**

Of course, there is a lot of data when an entire season is considered across all pitchers. Nevertheless, I am surprised at how repeatable the relative differences in velocity are for all of the game states studied across the three most recent regular seasons.

Notice how it is not just the average deltas in the categories that are almost identical, but each individual delta. The most notable outlier is the bases-loaded case, which of course has the smallest sample size of all in the study, so it is not surprising that it has the most variability.

The fact that velocity rises with each additional ball is somewhat surprising to me, as pitchers do tend to ease off on pitches thrown at 3-0 in particular just to get one across. The other three-ball counts obviously more than make up for this fact. The largest incremental velocity increase happens in two-ball counts. Perhaps this is a concerted effort by pitchers to avoid pushing the count to where the next ball leads to a walk. In general, velocity climbs less per additional ball than all of the other game states assessed.

The most significant single event change identified here occurs when a count reaches two strikes. Pitchers on average pump their fastballs more than half a mile per hour faster than in one-strike counts. This likely would be due to the knowledge that one more strike would end the plate appearance, so pitchers dial it up to another notch in anticipation.

The effect of the number of outs in the inning is also remarkably consistent year to year. Velocity jumps the most from nobody out to the one-out situation.

Finally, pitchers also tend to throw their fastballs harder the more runners there are on the basepaths. Notice how when just one runner reaches, velocity barely changes, but allowing a second and then again a third runner to reach leads to much more notable velocity increases.

Interestingly, these velocity deltas per game state appear to work out quite well in an additive fashion. In other words, if I calculate an estimated velocity increase relative to the 0-0 count with no outs and nobody aboard for every base-out-count state by summing the appropriate deltas from each of the four categories in the table, on the whole the velocities estimated average out to within one tenth of a mile per hour of the actual velocity differences.

In some ways, this makes sense, as these four categories are pretty well independent of one another. For example, the number of balls in the count should not be influenced by the number of outs in the inning, so if these two things are independent, we could expect a random distribution of ball counts for each out state.

While these velocity difference numbers are very consistent across the three seasons studied, they are of course not representative of every major league pitcher. Some pitchers ramp up velocity much more than average under some or all of these game situations, and some much less.

To give a few topical examples, American League Cy Young Award candidates Yu Darvish and Max Scherzer both sported velocity increases of over one mile per hour for each additional strike in the count, well above league average. They both also posted high velocity gains per additional out, with Scherzer also upping his fastball speed for every extra baserunner by noticeably more than the league norm.

Before you scroll down to the comments to type, “You should check if any of these increases correlate with higher strikeout rates,” I did, in fact, try this upon seeing those two strikeout kings near the top of the list. Unfortunately, no such luck.

Rounding out the AL Cy Young finalists, Hisashi Iwakuma actually posted a negative velocity delta per extra strike while hovering around the league norms otherwise.

The National League Cy Young Award final three are less interesting in this regard, with Jose Fernandez, Clayton Kershaw and Adam Wainwright all operating closer to the league averages than the noted cases of their AL counterparts.

In a previous study, I had observed a potential link between relative velocity increases per strike and pitcher disabled list trips, but unfortunately, the 2013 injury database is not available at this time. I hope to revisit this when it becomes available to see if the same trend extends across another season.

While somewhat unrelated to the specific purpose of this study, it is also apparent from these values that league-wide velocity rose on the order of 0.06 miles per hour in 2012, and then roughly another 0.20 miles per hour in 2013. Where and when will the increases stop?!?

From what I can see, there is no magic formula here with respect to increasing velocity in certain game states to generate success as a pitcher. It seems clear to me that, overall, pitchers tend to throw at a speed that is not maxed out, such that they can reach back when they feel the situation demands it and add a little extra spice.

Now, the question is, do hitters *swing* harder depending on the game state? Hmm…

Gil Tex said...

Jon, you beat me to the punch. The variance that jumped off the page to me is the increase in average fastball velocity over your 3-year study period. Is that a trend that is likely to continue? Is this 3-year period an aberration? I’d like to know if there’s a correlation between increase in average velocity and control (command).

Regarding game state velocity variances, I’m curious if you believe that the type of fastball thrown can impact statistical outcome? For example, more 2-seem sinkers may be thrown with a runner on 1st base and less than two outs in an effort to induce a double play ground ball. Maybe there are a higher percentage of cut fastballs thrown in less than 2 strike situations, therefore creating a greater mix of 4-seem fastballs thrown in 2 strike counts and in 3 ball counts. Darvish especially comes to mind as there is a significant velocity spread among his three fastballs; cutter, sinker and 4-seem. I’d like to hear your thoughts on fastball-type and its potential relevance in game state velocity variances.

Jon Roegele said...

Thanks for the comments.

I suspect the increasing velocity trend will continue, yes. I’m not sure where it will stop, but I see several reasons that could allow velocity to continue to climb. For one, pitchers are being used for shorter and shorter periods within a game, as specialization increases. Knowing that you’re in to face fewer hitters can let pitchers throw closer to their maximum speed, on average, I would think. Also teams continue to covet prospects with big arms, and likely conditioning is better and more well understood, so pitchers are likely on average able to throw harder just from training methods.

As for the fastball split, yes I think this would play a role for some pitchers, like Darvish as you suggest. I had to decide whether to look purely at 4-seam fastballs or all fastballs in doing the study. They both had issues. Often there are pitches classified as cutters or 2-seam fastballs that are faster than 4-seamers. So leaving this out didn’t feel right, as obviously pitches can and are throwing harder in some situation that I would not be counting. The downside is that pitch mix can confuse the results somewhat, but I was okay with having the different pitch types intermixed here as I thought that it is still telling us something about how pitchers tend to pitch in these situations, regardless whether they are switching to a more controllable fastball or not. Does that make sense to you?