When the Wind Blows

The Arizona Diamondbacks finished the 2006 season with the fourth-lowest home run total in the National League. But on July 31, this was the team that launched six home runs before the seventh inning stretch at Chicago’s Wrigley Field on the way to winning 15-4.

What was to blame for the Cubs pitchers’ trouble with the long ball? The wind, of course.

While Diamondbacks starting pitcher Brandon Webb generally kept the ball on the ground and allowed only one home run, Cubs starter Mark Prior struggled despite following his game plan: “I did exactly what I wanted to do, make them hit it to the big part of the ballpark. Unfortunately the wind was howling out tonight and I can’t do much about it.”

Dusty Baker pulled Prior after 106 pitches. “I wish Mark could have gotten farther in the game,” Baker said, “but the wind was blowing out quite heavily. They got some balls out and we didn’t.” Arizona manager Bob Melvin concurred, “It was a good day to get the ball in the air, nice and warm with the wind blowing out a little bit.”

Do atmospheric conditions, particularly wind, really have a significant effect on home run production? Hit Tracker’s estimates suggest it isn’t uncommon for wind to increase or decrease a fly ball’s distance by 40 feet or more. I used game time weather reports for all major league baseball games from the 2005 and 2006 seasons to evaluate the effect of wind on fly balls across all major league ballparks. I coded any game with wind speeds of 10mph or more in toward home plate or out to the outfield as “In” and “Out” respectively. When there was no wind (<3mph in any direction) the conditions are coded as "Still." The various other possible combinations of wind speed and direction, including side winds, are coded as "other." Here are the results for all regular season MLB games during the past two years:

           FB     HR     HR/FB   
  Out      9,487  1,331  14.03% 
  In       7,356  1,019  13.85%
  Other    40,681 5,674  13.95%
  Still    16,107 2,307  14.32%

A Chi-Square test suggests the differences are not statistically significant. That is, the frequency of home runs within each group is about what we would expect if the results were randomly distributed.

Does this mean wind doesn’t matter? Probably not. The categories are crudely constructed. It’s also likely that I am underestimating the effect of wind because conditions can change over the course of a game. Also, retractable roof stadiums complicate matters. Wind conditions might matter more in some places than others, and I tested this idea by calculating home run rates across wind conditions in all 28 open air major league ballparks.

             Frequencies              HR Rates
        Out    In   Oth   Stl     Out   In    Oth   Stl
ARI    3.1%  2.0% 27.3% 67.6%  14.6% 17.3% 16.3% 15.2%
ATL    2.5%  4.9% 90.1%  2.4%  15.9%  9.6% 14.0%  5.2%
BAL    7.5%  4.9% 73.9% 13.7%  18.5% 12.7% 16.1% 13.2%
BOS   26.6% 16.7% 51.8%  4.9%  13.6% 11.3% 11.4% 10.6%
CHC   17.5% 27.3% 52.1%  3.1%  23.2% 13.7% 17.0% 16.3%
CHW    9.2% 19.3% 68.0%  3.6%  24.5% 15.5% 17.1% 14.6%
CIN    9.5%  7.4% 70.3% 12.7%  18.0% 20.3% 19.6% 22.1%
CLE   20.0% 15.0% 52.7% 12.3%  14.0% 11.2% 13.5% 15.3%
COL    6.1% 12.9% 68.1% 12.9%  11.3% 13.2% 14.3% 16.5%
DET   11.7% 12.3% 70.5%  5.5%  12.6% 14.2% 13.5% 19.0%
FLA    6.8% 40.4% 44.8%  8.0%  10.1% 11.9% 12.4% 13.8%
HOU   14.0%  1.9% 15.3% 68.9%  15.7% 17.3% 14.7% 14.2%
KC     5.0%  4.3% 75.7% 15.0%  11.4% 14.2% 11.9% 15.0%
LAA    5.0%  0.6% 84.4% 10.0%   9.3%  5.3%  9.4% 12.2%
LAD    3.7%  0.0% 81.2% 15.1%   5.6%       13.8% 18.2%
MIL    6.2%  8.8% 42.3% 42.8%  19.7% 14.6% 14.3% 16.9%
NYM   19.3% 14.1% 62.9%  3.7%  13.2% 13.1% 12.2% 12.5%
NYY   19.2% 14.2% 63.0%  3.6%  16.4% 14.2% 15.6% 13.1%
OAK   29.9% 14.9% 52.0%  3.1%  11.8% 11.9% 12.0% 12.2%
PHI    9.4%  4.3% 80.8%  5.5%  17.9% 19.3% 19.4% 20.7%
PIT   10.3%  4.8% 73.8% 11.1%  10.5% 15.1% 10.5% 12.9%
SD    27.8%  0.0% 71.6%  0.6%   9.4%       11.6% 20.0%
SEA    1.2%  1.8% 38.7% 58.3%  23.3% 21.4% 11.3% 12.5%
SF    66.5%  3.9% 26.5%  3.1%  10.8%  7.0% 12.7% 10.0%
STL   16.5%  7.1% 71.4%  5.0%  16.7% 23.4% 15.3% 14.6%
TEX    8.9% 48.2% 33.3%  9.5%  18.7% 15.1% 16.9% 16.5%
TOR   14.8%  8.4% 36.1% 40.7%  14.2% 17.5% 16.9% 18.7%
WAS   11.1%  9.3% 66.5% 13.1%  12.7% 10.2% 11.5% 10.5%

The first four columns following the abbreviated team name describe the frequency of games played in each of the four conditions. In San Diego, for example, the wind is never strongly moving in toward home plate and is strongly moving out to the outfield at the start of about 28% of games played there. The final four columns describe the proportion of fly balls resulting in home runs in each condition. For example, at Camden Yards, 18.5% of fly balls result in home runs when the wind is headed toward the outfield.

Although wind conditions were not significantly associated with changes in homers per flyball rates across major league baseball, wind conditions are significant in some parks. Wrigley Field, the site of the six-homer effort by the Diamondbacks, is one of those parks. More than 23% of fly balls result in home runs when the wind is blowing strongly toward the ivy-covered outfield walls, but only 13% of fly balls result in home runs when the wind is blowing in from the outfield. The difference is even more striking on the South Side of Chicago. About once every 10 games, when the wind is blowing strongly toward the outfield, nearly a quarter of all fly balls leave the White Sox home park. The difference in proportion of fly balls resulting in home runs is statistically significant (p<.01) in both Chicago ballparks. In many settings, there are enough games in each condition to be sure that the difference in home run rates cannot be attributed to differing quality of pitchers. For example, the cumulative ERA of pitchers involved in the "Out" games at the park formerly known as New Comiskey was not significantly higher than the cumulative ERA of the pitchers in the other games. In other places, such as Fenway Park, the wind conditions don't lead to significant differences in home run rates. More fine-grained coding of wind conditions, such as analysis of games when the wind is blowing out to right field, might lead to notable results in a place like Fenway. At the moment, there simply aren't enough data available to draw strong conclusions regarding such specific conditions. Why do these results matter? Context is everything in baseball. Simple park effects that consider the difference between home and road outcomes are no longer adequate for describing the nuances of the game. Atmospheric conditions can provide useful information in assessing player abilities and making sound decisions during a game. For example, understanding typical changes in atmospheric conditions during a summer can help us understand which players' values will be maximized during the warm and windy mid-summer games in Chicago. Also, wind direction and speed aren't the only variables that can guide in-game decisions. Not only does the value of a run change as the temperature and wind shape the offensive environment in subtle ways, but the value of a fly ball in particular changes with the weather in many major league parks. This means baseball strategists may want to consider the effect of the atmosphere when calculating win expectancies for game situations. For example, small ball might be more useful during a cool day with the wind blowing in at Cleveland than it would be during a gusty day in Arlington. In sum, the wind plays a role in the value of a fly ball in some major league baseball parks, and we should consider this information in understanding how games are won and assessing the abilities of players. Next week, I'll describe the pitchers and hitters who benefited most from favorable atmospheric conditions throughout the major and minor leagues in 2006.