By CB Garrett
Abstract
Major League Baseball has one of the longest seasons in professional sports and players from all different backgrounds compete to play at the sport’s
highest level. The US is one of the largest countries in the world and with team’s scattered all across the country (and Canada), often times players are exposed to playing in weather conditions different from what they are used to. Players born in warm-weather climates such as Florida or the Dominican Republic might not be prepared to play a cold-weather outdoor game in Boston or Denver at the end of the season. The goal of this research is to determine if and how much where a player is born affects their performance in different weather conditions.
Data Collection
Data on every plate appearance from 2014-2017 was gathered from baseballsavant.com. This was then matched with the Chadwick Register to get the birthplace of every pitcher and hitter. Chadwick was also used to match the data on each plate appearance to data from Fangraphs.com on how each hitter performed over the course of that season. Data gathered by Bill Petti was also downloaded on the specifics of each game, such as game start time and stadium. Finally, both hourly and daily weather data for each city that was purchased from AccuWeather was added into the data set.
Enter Google Map Chart data for Figure 1. The number of plate appearances by US state from 2014-2017…Methodology and Modeling
Plate appearances were grouped by the birth country and birth state of the hitters and separate models were run on each US state and each country.
This was done to capture the effect of performance in different weather conditions by each birthplace. The model type that was used was a onefactor fixed-effects model. This was done so the model could control for the overall talent and performance of each individual hitter across the sample. Models were then run by country and by US state so the different locations could be compared.
Models were run in R with different Statcast metrics as dependent variables that designate a player’s ability to perform such as: isolated power (ISO), weighted on base average (wOBA), Exit Velocity, among others. Weather variables that were run as independent variables included humidity, precipitation, lowest daily wind chill, average daily temperature, highest daily wind gust, average daily pressure, and average daily visibility.
US Regression Results
| State | Highest Temp | Coldest Wind Chill | Avg. Humidity | If Precipitation |
|---|---|---|---|---|
| California | 6.965e-04** | -5.29e-04* | -3.58e-04** | -7.886e-04 |
| Florida | 0.00132*** | -0.000709* | -0.0002040 | 0.0049746 |
| Texas | 0.001053** | -0.0005676 | -0.00051** | 0.0046378 |
| New York | 0.0013661 | -0.0007522 | -0.000264 | 0.0225331** |
| North Carolina | 0.0013684 | -0.0012057 | 0.0001576 | 0.0024048 |
| Washington | 0.0015120 | -0.0012445 | -0.000709 | 0.0076429 |
| Illinois | 0.0009808 | -0.0002992 | -0.000191 | -0.0125090 |
| Pennsylvania | 0.00389* | -0.0023 | 0.0003489 | -0.0114362 |
World Regression Results
| Country | Highest Temp | Coldest Wind Chill | Avg. Humidity | If Precipitation |
|---|---|---|---|---|
| USA | 1.02e-03* | -5.9e-04* | -3.2e-04*** | 2.779e-03 |
| Dom. Rep. | 0.000782** | -0.0002818 | -0.0001897 | -0.0006510 |
| Venezuela | 0.00118* | -0.00069 | 0.0001543 | 0.0067673 |
| Puerto Rico | 0.0013711* | -0.0010044 | -0.0006611* | 0.0137135 |
| Cuba | 8.942e-04 | -6.315e-04 | -3.142e-05 | 8.949e-03 |
Results
A higher temperature is shown to lead hitters to hit for more power, while a colder wind chill and higher humidity cause power to fall at a statistically significant level. This can easily be seen in the USA, the country with by far the largest sample. Across countries and states with lower sample sizes, the signs are mostly the same, while level of significance and magnitude varies.
Warmer countries like the Dominican Republic, Venezuela, and Puerto Rico having significant coefficients with larger magnitude than the USA show that players from those countries have power affected by more than players from colder climates. In the state-by-state model, similar results are shown as warm climate states like Florida and Texas have significant coefficients.
Conclusions
Generally players performed better in warmer and less humid weather; however, players born in colder climates were less effected by the conditions. This is likely because they are more used to colder weather conditions and played games in similar situations. This could have a slight impact on roster construction because teams that play more games in colder climates (especially during playoff time) might prefer a player born in a colder climate, all other things being equal.
Future research would include use of hour-by-hour weather and historical weather on the birthplaces of the players, and a deeper dive into the effects of weather on pitchers. Additionally more Statcast data would be regressed to see how it is effected by weather and see if players from warm climates can get acclimated to colder weather over a season. Finally, injury data has been scraped from prosportstransactions.com and this will be run in a model to see if players from warmer climates are more susceptible to injuries while playing in colder weather.
References
- Nathan (2012)
- Koch and Panorska (2013)
- Eddy (2018)
- Blatt (2014)
- Swartz (2014)
- Ejermo and Hansen (2014)
- Baker et al (2009)
- Rossing et al (2018)
- Curtis and Birch (1987)
- Cote et al (2008)
- Baker and Logan (2007)
- Vigotti et al (2006)
- Watanabe et al (2017)
- fangraphs.com
- Chadwick Register
- baseballsavant.com
Contact
CB Garrett
Syracuse University
Email: chgarret@syr.edu
LinkedIn
Phone: 908-565-6556