A beautiful day at the ballpark in the summer is unrivaled. This is especially true in an outdoor venue when the temperatures are in the 65-75 range. But a hot summer’s day can be brutal for players, fans, and umpires alike. That begs the question: do games with bad weather end up being shorter than games in more ideal conditions?
To do so, the first set of data I took a look at was the full 2016 season through July 4th. I excluded box scores where temperature was not given (e.g. games played indoors), extra inning games (to be able to compare 9 inning games against one another without an 18 inning six hour epic skewing the data), and discounted rain delay times. This should give us a relatively clean data set of 980 games thus far this season.
In the course of my investigation, I expected “nice” temperature games (60-80 or so) to be the longest games, the psychology of a nice summer’s day providing no incentive for umpires to call quicker games, for batters to swing at bad pitches because it’s cold/hot, and pitchers opting to pitch around more batters than they would in colder temperatures. Imagine my surprise when I took the 2016 numbers and threw them into a graph:
Of course, you can see how the lines are quite variable on the tails with few data points, as one would expect. In the middle, despite decent sample sizes of 25-35 games, you still see quite a bit of variance between games, but the general trend seems to hover around three hours flat (180 minutes). But let me clean up the data a little bit to try and eliminate some of the noise:
Now this graph is a little misleading due to the left axis showing an incredibly small range, but what it seems to show is that temperatures in the “ideal” range tend to be a little bit faster than their hotter or colder counterparts. Though with a net variance of just over three minutes, I would hardly call this statistically significant.
So if I were a quitter, that would be the end, and the answer would be no. There’s no real connection between temperature and game time, at least in 2016.
Lucky for you folks, I’m not a quitter. With the help of the Retrosheet database from 1952-2015, I was able to pull 73,354 games with temperature data, and it looks like this:
Of course, this is still surprisingly messy, despite having >1000 games of data for 60-90 degree temperatures. Interestingly, you can see huge peaks at 68, 70, and 72 degrees: My guess is that these are temperatures reported for games that took place in a dome. The scale on this is much larger, from a 23 degree game in Colorado in 2013 to a pair of (box 11, box 2) 109 degree games in the old Arlington Stadium (you’ll note that the 109 degree day is actually off the chart . Cleaning up the data as before, we get this:
To me, this seems a bit more statistically significant. With a sample size of nearly 75,000 games, a difference of 3-4 minutes seems to me that it is rather dramatic. And from the graph, we can see that the games from 40-69 degrees are between 1.8-3.2 minutes faster than games outside of that temperature range.
This flies in the face of my initial assumptions, where I valued psychology to play a major factor, but what I had ignored was the physiological aspects of the game. My theories:
- Colder baseball games lead to more pitching changes, as pitchers struggle to stay warm and loose. I would imagine a pitcher sitting on the bench in, say, the 6th inning for a half hour in freezing temperatures is more likely to lead to him being pulled than that same situation in 70 degrees.
- Hot baseball games lead to more overall substitutions, as fatigue increases due to dehydration. Pitchers are the main cause here, with pitching changes slowing down the game more than pinch hitters, but they may factor in as well.
- The ball travels farther in hot games, as air density is less (especially with high humidity). This turns what may have been 395 foot outs in 40 degrees into home runs in 90 degrees, leading to higher scoring and thus longer games.
Of course, those are just theories, but the data for 50 years of baseball seems clear: if you want to increase your chances of getting home at a reasonable time, check that weather forecast and head to the ballpark when it’s 55 degrees out.
Got a theory yourself as to why the data skews as it does? I’d love to hear your ideas in the comment section below.