####
By: Michael Bogaty, Tyler Duncan, and Luke Benz

######
June 30th, 2017

Pitchers try any number of strategies to control the other team’s running game, including slide steps, pitchouts, and pickoff throws. What we wanted to investigate is whether pitchers changed what pitches they chose to throw because of the runners on base. To do this, we examined data from every pitch for the 2016 season. While pitch-fx data is not perfect, it is generally pretty good at, at a minimum, distinguishing fastballs from off speed pitches, and more often determining exact pitch types.

We categorized pitches into four categories: Four Seam Fastballs, Moving Fast Balls (2-seamers and cutters), Changeups, and Breaking Balls. This allows us to do a more uniform comparison given the different pitches in different pitchers repertoires while ensuring the sample size in each category remains large. We excluded knuckle balls from the analysis, since the pitchers that throw them are unique and their tendencies likely can’t be applied to other pitchers. Below is a breakdown of pitch-type by count.

Count | Breaking Ball | Change-Up | 4-Seam Fastball | Moving Fastball |
---|---|---|---|---|

0-0 | 0.252 | 0.076 | 0.476 | 0.197 |

0-1 | 0.297 | 0.118 | 0.406 | 0.180 |

0-2 | 0.369 | 0.089 | 0.407 | 0.115 |

1-0 | 0.198 | 0.127 | 0.465 | 0.211 |

1-1 | 0.279 | 0.135 | 0.407 | 0.179 |

1-2 | 0.369 | 0.108 | 0.383 | 0.140 |

2-0 | 0.112 | 0.088 | 0.561 | 0.239 |

2-1 | 0.198 | 0.128 | 0.466 | 0.208 |

2-2 | 0.318 | 0.121 | 0.405 | 0.157 |

3-0 | 0.023 | 0.022 | 0.744 | 0.211 |

3-1 | 0.094 | 0.059 | 0.600 | 0.247 |

3-2 | 0.210 | 0.099 | 0.497 | 0.194 |

These results are rather intuitive. In “hitter’s counts” such as 3-1 the pitcher needs to throw a strike and will throw nearly 85% fastballs, while in “pitcher’s counts”, the pitcher is looking for a strike out and more likely to go off-speed.

To begin investigating the effect of baserunners, we compared the pitch selection data based on purely whether the bases were empty or there was at least one runner on base. This includes both stolen base situations, such as runner on first, and situations where a stolen base is incredibly unlikely, such as runners on second and third. Because pitchers are pitching from the stretch and perhaps worried about wild pitches, it is certainly plausible there would be differences merely based on the presence of a base runner. However, we find that this is not the case to any meaningful degree. The percentages of pitches in each category are below:

Runner(s) on Base | Breaking Ball | Change-Up | 4-Seam Fastball | Moving Fastball |
---|---|---|---|---|

No | 0.261 | 0.104 | 0.454 | 0.181 |

Yes | 0.266 | 0.103 | 0.441 | 0.190 |

While this is significant based on a chi-squared test, it is not meaningful in any real way (with such a big data set even minor differences will be significant). A batter knowing that a breaking ball has a 0.5 percent greater chance of coming with runners on base won’t change his approach at the plate.

Next, we looked only at situations where a runner was on first base and second base was unoccupied. These are the situations in which a stolen base is the biggest threat. Here, there is a somewhat meaningful effect on pitch selection. Pitchers throw about 2.5 percent fewer breaking balls in steal situations.

Runner on First | Breaking Ball | Change-Up | 4-Seam Fastball | Moving Fastball |
---|---|---|---|---|

No | 0.269 | 0.104 | 0.447 | 0.180 |

Yes | 0.241 | 0.103 | 0.455 | 0.202 |

The difference was significant based on a chi-squared test (p<10^-5). This suggests that the cause of the difference is most likely the threat of a steal. It is nothing intrinsic to having runners on base that causes pitchers to throw (slightly) fewer breaking balls but rather the runner being in a position to steal.

If it is indeed the threat of a steal that changes pitch selection, then we should see an increase in fastballs when it is a very speedy runner on base. So, we examined the top five fastest runners in the league based on Statcast data (Billy Burns, Dee Gordon, Billy Hamilton, Delino DeShields, and Jose Altuve). With one of these runners on first base and second base empty, there is approximately double the difference in fastball percentage as with merely any runner on first base and second base empty.

Speedster on First | Breaking Ball | Change-Up | 4-Seam Fastball | Moving Fastball |
---|---|---|---|---|

No | 0.263 | 0.103 | 0.449 | 0.185 |

Yes | 0.210 | 0.105 | 0.476 | 0.209 |

The difference was significant based on a chi-squared test (p < 10^-5). With one of these runners on base, the difference in breaking ball percentage is substantial enough that we could imagine hitters could benefit from incorporating this information. Interestingly, in all of these circumstances the percentage of fastballs that are straight and the percentage of changeups thrown does not change substantially.

We then ran the analysis instead looking at the top 5 base stealers in 2016(Jonathan Villar, Billy Hamilton, Starling Marte, Rajai Davis, Eduardo Nunez). It would make sense that, if stolen bases are what worries the pitcher and causes the change in behavior, then how often a runner steals would matter more than his pure speed.

Stealer on First | Breaking Ball | Change-Up | 4-Seam Fastball | Moving Fastball |
---|---|---|---|---|

No | 0.263 | 0.103 | 0.449 | 0.185 |

Yes | 0.202 | 0.085 | 0.495 | 0.219 |

This is indeed the case, as the difference in fastball percentage is now around 8%.

Thus, our analysis has confirmed the intuition that base runners change the behavior of pitchers. The reason for this seems to be solely the threat of a stolen base. Having a runner on first, one of the fastest runners in the league on first, and one of the top base stealers increase the percentage of fastballs 2.5%, 5%, and 8% respectively. In the latter case, pitchers throw nearly a quarter fewer breaking balls. This could be another hidden value of fast runners. If they give more hittable pitches to the following batters they can have an impact on the game by merely standing on first, even when they don’t attempt to steal.

**Authors' Note:** This analysis was conducted before the creation of Statcast Sprint Speed earlier this week. Adding an additional analysis in light of this new baserunning metric would be a natural extension of this work.