I implement a machine learning technique called nearest neighbours to accurately answer this question and find that a mound visit decreases the probability of the batter reaching base or a run being scored by 65 points!
Charlie Brown says it's the loneliest place in the world.
But several times each game, the pitcher's mound becomes an area of congregation for the pitcher, the catcher, the manager, and infielders alike. So what the heck are they all talking about out there?!
Candlesticks, probably. (Kevin Costner reference number 1).
A mound visit is the baseball equivalent to the time-out in other major sports. It can be used as a stall tactic for the next pitcher to warm up before entering the game; a way to break up the hitting team's momentum; or a time for a catcher to remind a pitcher, 'we don't stink right now.' (Kevin Costner reference number 2). Theoretically, the mound visit is a chance for the fielding team to regroup and focus on turning the next at bat into an out.
But does a mound visit actually improve the odds of the fielding team? To evaluate this question, I start by gathering all MLB plate appearances from the 2016 to 2018 seasons. I then calculate the number of times an at bat ends with the batter reaching base or a run scoring (let's call these 'negative outcomes') and whether a mound visit occurs before the at bat. Below is a summary of my findings:
Situation |
Batter Reaches or Run Scores |
---|---|
No mound visit | .346 |
Mound visit | .433 |
So, this means that the impact of a time-out ...
... 433 minus 346 ... carry the one .... and ...
Increases the likelihood of a negative outcome by 87 points?!
Wait a minute, if the probability the batter reaches base increases after a mound visit, why would the fielding team ever use this tactic?
What the table above is missing is that the situations that induce a mound visit are the same situations that negative outcomes are very likely to occur. The table presents unconditional probabilities which, unfairly, do not compare apples to apples.
In order to do an accurate comparison, I implement a machine learning technique called Nearest Neighbours. In a very high-level explanation, I first estimate the probability a mound visit occurs before a given at bat using a logistic regression. I then pair each observation with a mound visit to a nearly identical observation without a mound visit (its nearest neighbour). What is left is a balanced set of observations with different mound visit strategies and potentially different outcomes. I can now use this balanced dataset to estimate the impact a mound visit has on the outcome of the next at bat.
Below is a graphical depiction of the results. On the left is the unconditional probabilities of the batter reaching base or a run being scored as summarised in the table above. On the right is the conditional probabilities resulting from the nearest neighbour technique:
With the correct comparison, we see that a mound visit decreases the probability of the batter reaching base or a run being scored by 65 points. It is important to remember that there is a high probability that a negative outcome occurs in a situation where a mound visit is likely. While, the mound visit gives some relief to the fielding team, a negative outcome is still more likely to occur after a mound visit than in an average situation (as demonstrated by the unconditional probability). But when the probability of a negative outcome is dire, a mound visit can have quite a large impact to help alleviate some pressure.
So there you have it: mound visits, when executed at the right time, can have a non-negligible impact on the outcome of the game. This now concludes the main discussion on mound visits. Below I share some other insights I came across while conducting my preliminary analyses. Thanks for stopping by and if you enjoyed what you read, please consider commenting, sharing, subscribing, and clicking on some ads!
Hello again! Like a good baseball game going into extra innings, enjoy some exciting extra Sports and Economics (in that order)!
While researching this topic, I came across a number of interesting findings: First, I find the best predictor for whether a mound visit is about to occur is the number of pitches thrown without recording an out. Below is a graph of how the probability of a mound visit changes as the pitch count between outs increases.
Another interesting find is the number of mound visit by the team at bat. The percent of times a mound visit is called against a team correlates rather well to the team's On-Base Percentage. Unsurprisingly, the top three teams who see the most mound visits are the same three teams featured in the last two World Series, while the bottom three teams are, well, also MLB teams.
Wait a minute, if the probability the batter reaches base increases after a mound visit, why would the fielding team ever use this tactic?
What the table above is missing is that the situations that induce a mound visit are the same situations that negative outcomes are very likely to occur. The table presents unconditional probabilities which, unfairly, do not compare apples to apples.
In order to do an accurate comparison, I implement a machine learning technique called Nearest Neighbours. In a very high-level explanation, I first estimate the probability a mound visit occurs before a given at bat using a logistic regression. I then pair each observation with a mound visit to a nearly identical observation without a mound visit (its nearest neighbour). What is left is a balanced set of observations with different mound visit strategies and potentially different outcomes. I can now use this balanced dataset to estimate the impact a mound visit has on the outcome of the next at bat.
Below is a graphical depiction of the results. On the left is the unconditional probabilities of the batter reaching base or a run being scored as summarised in the table above. On the right is the conditional probabilities resulting from the nearest neighbour technique:
With the correct comparison, we see that a mound visit decreases the probability of the batter reaching base or a run being scored by 65 points. It is important to remember that there is a high probability that a negative outcome occurs in a situation where a mound visit is likely. While, the mound visit gives some relief to the fielding team, a negative outcome is still more likely to occur after a mound visit than in an average situation (as demonstrated by the unconditional probability). But when the probability of a negative outcome is dire, a mound visit can have quite a large impact to help alleviate some pressure.
So there you have it: mound visits, when executed at the right time, can have a non-negligible impact on the outcome of the game. This now concludes the main discussion on mound visits. Below I share some other insights I came across while conducting my preliminary analyses. Thanks for stopping by and if you enjoyed what you read, please consider commenting, sharing, subscribing, and clicking on some ads!
Hello again! Like a good baseball game going into extra innings, enjoy some exciting extra Sports and Economics (in that order)!
While researching this topic, I came across a number of interesting findings: First, I find the best predictor for whether a mound visit is about to occur is the number of pitches thrown without recording an out. Below is a graph of how the probability of a mound visit changes as the pitch count between outs increases.
Another interesting find is the number of mound visit by the team at bat. The percent of times a mound visit is called against a team correlates rather well to the team's On-Base Percentage. Unsurprisingly, the top three teams who see the most mound visits are the same three teams featured in the last two World Series, while the bottom three teams are, well, also MLB teams.
Hitting Team | % of At Bats with a Mound Visit |
On-Base Percentage |
|||
1 | Houston Astros | 3.74% | 0.331 | ||
2 | Boston Red Sox | 3.66% | 0.339 | ||
3 | Los Angeles Dodgers | 3.65% | 0.329 | ||
… | |||||
28 | Baltimore Orioles | 2.98% | 0.309 | ||
29 | Chicago White Sox | 2.95% | 0.311 | ||
30 | Kansas City Royals | 2.72% | 0.309 |
Don't forget to comment, share, rinse, and repeat.
No comments:
Post a Comment