The Wondrous and Withering Weather

Sometimes, it's so hot that you sweat just from standing (via Matthew Straubmuller).

Sometimes, it’s so hot that you sweat just from standing (via Matthew Straubmuller).

At The Hardball Times, we spend a lot of time quantifying different events that occur on the baseball field. We don’t do this just because we can, but because we want to understand where value is created and lost. One of the areas where we don’t do a great job is with weather.

As somebody who played through college, I consider weather to be one of the three strongest influences on performance. I’ve played baseball in rain, snow, sub-freezing temperatures, oppressive humidity, and skin-melting heat. On occasion, those less than ideal conditions were combined (especially heat and humidity). When the field was wet, I regularly made wild throws. As a pitcher, my velocity would drop as much as 10 mph from its peak in cold conditions. Professionals obviously have many advantages, from climate controlled clubhouses to new baseballs for every pitch if needed. But if weather affects amateurs, then it follows that it probably affects professionals.

Not much work has been done to discover how weather affects the outcome of major league games. I’ve teamed up with data specialist Daren Willman — who runs the indispensable Baseball Savant website (and two others as well) — to further explore the relationship between weather and game outcomes. Daren developed an excellent tool for visualizing temperature by date. Our sample — derived through the power of Retrosheet — covers 1980-2013, and breaks down every game for each team by start temperature. You’ll notice that before 1997, there are gaps in the data. You’re not reading it wrong — the data from 1980-1996 is simply incomplete. Overall, the sample is 59,438 games played with documented weather data. Included in the data set are the teams that played, score, start time, game length, temperature, field condition, wind speed/direction, precipitation, and attendance. In many cases, the field condition, wind direction, and precipitation are unknown.

The calendar grid for all 30 teams is below. We set it to the Red Sox, since they are the reigning World Series champs. The temperatures go from blue (cold), to green (warm), to orange (hot). As an added bonus, if you click a specific game, a link for that day’s Baseball-Reference box score will pop out. Underneath the calendar grid, you’ll find a summary graph. The summary graph is static — i.e., it is the same no matter what team you are examining.

OK, we’ll hang up now and let you play.

One important caveat applies: the data appear to be local temperature rather than on-field temperature. When I used to attend Veterans Stadium as a child, a breezy 85 degree day could easily be over 105 degrees on the Astroturf. You can assume that the on-field temperature is higher than the reported temperature for the summer months.

Conditions vary greatly in different years and at different locations. Of our nearly 60,000-game sample, 24 percent were played at an ideal temperature—70 to 74 degrees. If we expand “ideal” to a more liberal range of 68 to 89 degrees, we find that 68 percent of games featured a pleasant temperature.

Some of those observations are for fixed domed stadiums. Without the domes, we have 54,341 observations, of which 19 percent are in the 70 to 74 degree range and 65 percent are in the 68 to 89 degree sweet spot. We have chosen to leave retractable roof stadiums in the sample since we do not know when the dome is open or closed.

At the extremes, 26 games were played in temperatures over 105 degrees. Guess who hosted 12 of those games? (Hint: they play outdoors in Texas). The Rangers won eight of those 12 games. On the other end of the scale, the Colorado Rockies have hosted four games under 30 degrees and won two of them. Don’t feel left out, Chicago fans, the Cubbies hosted 45 games under 40 degrees and the White Sox pitched in with another 38 games. Congratulations, your city is cold.

Last season was quite chilly; 19 games were played below 40 degrees. That includes two of those sub-30 degree games in the sample. Both Colorado and Minnesota hosted five of the sub-40 games. The Chicago teams combined for six and Pittsburgh, Detroit and Texas rounded out the list. By contrast only 11 such games were played in 2011 and 2012. That trend remains true of hot games too; 2011 and 2012 combined for 72 games above 100 degrees compared to just seven last season.

Hot and Cold

So which teams have played in the hottest weather and how did they perform? The following table is for all games that were at least 95 degrees.

Games Played, 95+ Degrees
Team Games Wins Losses
TEX 380 165 215
BAL 151 79 72
KCA 128 54 74
SLN 112 52 60
CHA 75 30 45
BOS 74 35 39
PHI 64 26 38
DET 64 32 32
TOR 63 28 35
LAN 59 30 29

Texas very nearly played as many 95 degree or higher games as the next three teams (391 games combined). Assuming the on-field conditions are even hotter, summer games in Texas must be a major burden on major league training staffs.

It would be interesting to learn what the Texas training staff does to keep players hydrated in extreme heat. Several years ago there was talk that the hot summer conditions were part of the reason why guys like Ian Kinsler and Nelson Cruz couldn’t stay off the disabled list. There is some intuitive sense in that; hydration is essential to peak performance. Dehydration has been found to reduce neuromuscular control in soccer players – the same probably applies to baseball players. Here’s a money quote:

“Why is neuromuscular control important? Loss of neuromuscular control can result in poor balance and poor sport-specific technique. And, balance and movement technique are critical components in joint injury.”

You might think that teams used to playing in hot conditions might have an advantage – maybe secret hydration techniques, superior home team facilities, or better overall conditioning. That was always the line in youth sports – if you trained harder and in more averse conditions than your opponents, you’d beat them.

What we actually witness is that only two of the 10 teams performed better than .500. The best was Baltimore with just a .523 record. Huh? Altogether, the teams have a .454 winning percentage. Does frequent exposure to hot temperatures tear down athletes faster than they can adjust? Did we happen to catch a sample of sub-.500 teams? Maybe. We’ve definitely found something that deserves more research.

Okay, how about the cold weather teams? Here are all the games under 40 degrees. In my experience that’s about the temperature where the answer to staying warm and performing well is no longer “more layers.”

Games Played, Under 40 Degrees
Team Games Wins Losses
CHN 56 26 30
CHA 49 26 23
DET 38 20 18
CLE 34 13 21
MIL 27 14 13
TOR 25 14 11
KCA 25 13 12
MIN 24 14 10
COL 23 11 12
CIN 22 10 11

So, I mentioned Chicago. Let’s just throw in the entire Midwest, Canada and the Rocky Mountains. Over time, we’ll see Toronto fall off this list due to the Rogers Centre, while the Twins are likely to eventually find their way to the top.

Our sample sizes are much smaller than with the hot weather teams, which hampers even the descriptive analysis we’re conducting today. We could expand to include all games under 50 degrees, but the upper-40s isn’t so bad when you get to return to a heated dugout every 10-15 minutes. The relative ease of keeping an athlete warm might be why everyone but Cleveland is hovering right around .500. It doesn’t seem like there are enough cold games for teams to figure out any replicable advantage. Similarly, there is no popularly discussed pattern for injuries, although that doesn’t mean players are safe when the weather turns cold.

Maybe Minnesota should invest more effort into studying baseball performance in cold weather. Let’s see.

A Case Study of the Minnesota Twins—Is It Really That Cold?

Prior to the 2010 season, the Twins played all of their games indoors at the Hubert H. Humphrey Metrodome. As you may have heard, Minnesota can get quite cold. Sometimes there are snow storms in April. The nice thing about the Metrodome was that it kept the team out of some very unpleasant conditions. The not so nice thing about the Metrodome was just about everything else. The playing conditions were a step below the outdoor stadiums that most teams feature. The fan experience was very limited, and there was almost no opportunity for the large scale luxury box revenue that many teams enjoy.

Enter Target Field. Whereas the Arizona Diamondbacks opted for a retractable roof stadium to combat the extreme heat of Phoenix, the Twins counted on the mostly temperate summer of Minnesota and went with a purely outdoor (and cheaper) stadium. Since Opening Day of 2010, 326 games have been played in Minnesota with recorded weather. Here is how they break down by temperature.

Temperature of Twins Home Games
Degrees F Games Percent of Games
90-97 16 4.9%
80-89 80 24.5%
70-79 112 34.4%
60-69 59 18.1%
50-59 41 12.6%
40-49 13 4.0%
34-39 5 1.5%

Let’s see how that compares to league average.

Temperature of All Games Excluding Target Field
Degrees F Games Percent of Games
100-109 225 0.4%
90-99 3,218 6.0%
80-89 12,693 23.5%
70-79 19,203 35.6%
60-69 12,520 23.2%
50-59 4,582 8.5%
40-49 1,386 2.6%
30-39 175 0.3%
<30 12 0.0%

In the four years that they have played outdoors, it’s apparent that the Twins have experienced a higher than average rate of cold games. Since we already knew that about Minnesota in general, this finding isn’t a big surprise. No games were played over 97 degrees whereas about half a percent of all major league games fall in the 100 degrees or over bucket. Surprisingly, the Twins played fewer games than league average in the 60-79 degree range. Unsurprisingly, they played a higher percentage of games in the 30-59 degree range. They’ve yet to experience a sub-freezing game. Interestingly, the average temperature of all games played at Target Field is 71 degrees compared to 73 degrees at all other venues. That’s not a big gap.

One issue with this mini-case study of Target Field is that it contains only four years of data for the Twins. National weather patterns can vary by year and decade. There was the global cooling scare in the 1970s, global warming in the 1990s and 2000s, and climate change in more recent years. We’re comparing 2010-2013 in Minnesota to a much larger sample of general baseball data. That could be skewing our simple descriptive statistics in a number of ways.

That said, the early findings show that baseball might be much colder in Minnesota than the league average. It’s probably only a matter of time before they experience that first sub-freezing game.

How Much Colder are the Playoffs?

Or more specifically, how much colder is October compared to September? We have a sample of 1,345 October games, most of which are postseason contests. The data have their share of bias because teams are not evenly represented in the playoffs. There are a lot of October games in New York (155). Not so many in sunny San Diego (31). Let’s take a look at the distribution.

Brad-Daren-Chart-1

Atlanta, Los Angeles, New York, Boston, Chicago and St. Louis represent 25 percent of the teams and 43 percent of the games in the sample. This sample also includes a number of domed stadiums, which we’ll remove. Those include Minnesota, Seattle, Montreal, Tampa Bay and Houston. We can’t do anything about the retractable roof stadiums, most of which are probably closed for the postseason.

Brad-Daren-Chart-2
Removing domes pulls 94 observations out of our sample. We’re left with 1,251 games with known weather, some few of which may have actually been played indoors.

Temperatures range from seven to 100 degrees. The average temperature is 67 degrees. Not bad. Let’s break into cohorts.

Temperature of October Games
Degrees F Games Percent of Games
90-100 12 1.0%
80-89 131 10.5%
70-79 384 30.7%
60-69 416 33.3%
50-59 226 18.1%
40-49 77 6.2%
30-39 3 0.2%
<30 2 0.2%

Well, October is colder than Target Field as a whole. But how does October compare to September?

Temperature of September Games
Degrees F Games Percent of Games
101-106 17 0.2%
90-100 283 3.2%
80-89 1,730 19.8%
70-79 3,672 42.0%
60-69 2,518 28.8%
50-59 487 5.6%
40-49 31 0.4%
30-39 2 0.0%

For September, we have 8,740 observations with observed temperatures from 37 to 106 degrees. The average is a lovely 73 degrees. So just from an average temperature standpoint, September is a bit nicer.

When we focus on the extremes, we can see why October baseball seems so much colder than September. There are certainly a lot more games above 90 degrees in September—3.4 percent compared to just one percent. If we restrict ourselves to games under 60 degrees, we find that nearly a quarter of the October games fall into that bucket compared to just six percent of September games. Those cold playoff games really stick with us.

Is it Colder at the Beginning or End of the Season?

So we have our September and October data, but how does April compare?

Temperature of April Games
Degrees F Games Percent of Games
90-100 46 0.6%
80-89 499 6.5%
70-79 1,629 21.3%
60-69 2,394 31.2%
50-59 1,928 25.2%
40-49 996 13.0%
30-39 166 2.2%
20-29 6 0.1%

Now there’s something unexpected. April looks like it’s the coldest month of the season. We have 7,664 observations ranging from 23 to 99 degrees. The average temperature is just 62 degrees—a touch colder than October. In April, over 40 percent of the sample falls in the under 60 degrees bucket compared to 25 percent in October and six percent in September.

We always look at October as the coldest month of baseball, but it’s really April. I’m sure part of the explanation is the timing of the World Series. Those are the games that produce memorable moments of Mariano Rivera staring in to the catcher, steam billowing as he takes the sign and comes set. We get those same moments in April—more of them according to the data—they just mean less to us.

Next Steps

The preceding is a descriptive analysis of weather in major league baseball. The obvious next steps are to examine the influence of weather on factors like attendance, performance and injuries. Analysis of attendance data with weather could tell us when teams should focus their promotions, especially if a team deals with predictable, seasonal weather.

Linking to performance data could be a treasure trove. We know that pitchers have a lower ERA in April than all other months. What share of that is the low temperature and other adverse conditions? Is it actually that certain temperatures influence run scoring?

Teams might be most interested in anything we can learn about injuries. Millions and millions of dollars are lost every season when players land on the disabled list. If training staffs can prepare certain preventative measures in anticipation of the weather, the sport will benefit as a whole.

And let’s not forget, why did the teams with the most hot weather games generally perform poorly in those games?

References & Resources


Brad is a former collegiate player who writes for FanGraphs, The Hardball Times, and Fake Baseball. Follow him on Twitter @BaseballATeam or email him here. Daren is a software developer and proprietor of the sites Baseball Savant, MLB Farm and NFL Savant. Follow him on @darenw or email him here.
13 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
aweb
9 years ago

Start temperature is the available measure, but especially later in the year with earlier sunsets, the temperature drop must be more pronounced during common game times (7-10pm). Are game-end temperatures available? If not more pronounced in absolute terms, then a 10 degree drop starting at 45 is a lot more unpleasant than if it starts at 65 or 85. Those are the really memorable cold moments in October to me – and since sunset is earlier in the day at that time of year (are playoff games longer too?), that might be an important factor in some ways.

Peter Jensen
9 years ago

Authors- Retrosheet tells you whether a dome is in use by the code 5 in the sky type in the game info table. the codes are all explained here: http://www.retrosheet.org/eventfile.htm

Even though you cite Retrosheet as a reference you should probably use the standard reference that they request here: http://www.retrosheet.org/game.htm#Notice

The effects of weather have previously been discussed many times over the years on THT Baseball Prospectus, the Book Blog and other sites. A search of archives of these sites could have avoided your rediscovery of the information in your article that is already well know from previous research.

MSpitz
9 years ago

This is really interesting stuff. One game in particular that stood out to me was August 11th, 1992 in Toronto. According to this graph, it was 20 degrees…that can’t possibly be right, can it???

Brad Johnson
9 years ago
Reply to  MSpitz

I noticed that one and cross referenced with historical weather data. It’s incorrect. I pulled it from the cold weather analysis section, but forgot to make a note (or maybe I decided it wasn’t worth noting).

I took a sample of other extreme weather points, but they all checked out. I’m sure there’s a couple other bad points of data scattered in there, but they’re probably in the common (less interesting) temperature ranges.

Billy
9 years ago

A couple of comments… First, the .454 winning percentage for the teams that play most frequently in 95+ degree temperatures is weird. Winning percentage in 95+ games has to be .500 since one team wins and one team loses. Is this suggesting that teams that rarely play in 95+ degree temperatures are significantly more likely to win games played in 95+ degree temperatures?! That’s what it sounds like, but I really have a hard time grasping why that would make any sense at all.

Second, this is only looking at the profile of weather based on game-day conditions. What about the larger trends? Is run scoring higher in warmer seasons? Are injuries more common? How about Spring Training? Are injuries more common when weather is colder during the spring? Or perhaps during the offseason? It was a cold winter, and it seems like there have been a lot of injuries so far this year. Is there a reall effect there, and if there is, is there anything that can be done about it?

Brad Johnson
9 years ago
Reply to  Billy

It’s so weird! Several of those teams were pretty bad throughout most of the sampled time period. My best guess is that’s why the winning percentage is so low. Even then, .454 is crazy.

I like your questions. I asked some of the same questions. We decided to start at the surface and dive in deeper in the future.

tz
9 years ago
Reply to  Brad Johnson

Cool. That was going to be my first question on this (win % under extreme conditions vs. win% overall).

And to Billy’s point, I do wonder if there is such a thing as cumulative heat fatigue. If so, that may explain the Rangers’ under-performance in very hot weather.

Nathaniel Dawson
9 years ago

Am I mistaken, or is there no way to import this data to a spreadsheet? I would love to have this in spreadsheet form.

evo34
9 years ago

It’s going to be a tricky analysis, because you have so many correlated factors at work. Among them:

Non-temp. park effects vs. avg. temp. at given park (teams that play in hotter weather may have parks that affect runs scoring)
Temp. vs. daylight (day games are hotter, but light conditions may also have effect on run scoring)
Temp. vs. wind (many parks have wind direction and speeds that are correlated to temp.)
Sunset time vs. temp. (later sunsets generally mean higher temps., but they also mean more twilight conditions for night games)
Roster trends vs. temp. (Sept. games are cooler, and also have a lot more minor leaguers playing in them)

My advice would be to do all analysis on a per-park basis (to eliminate issue #1 above). We do know that a batted ball will travel further in hotter temps., all else being held constant. Also, pitch velocity rises in hotter temps. So, net, my guess would be that higher temps. cause more Ks and more extra base hits per BIP, with the effect the extra base hits overwhelming the added Ks, leading to a net effect of higher run scoring.

evo34
9 years ago

My previous comment was on run scoring effects. As far as competitive advantage effects, I’m not you’re going to be able to show anything remotely conclusive — simply because team talent is not randomly distributed (either for the team being measured or for the schedule of teams it faces). So if it so happens that mgmt. of a few hot-weather teams has been subpar on average over the years (which is pretty likely), there will be no way extract the true effect of temp., if any, on team performance. You already acknowledge this issue in the article, but just pointing out that, unlike the overcoming the challenges involved with measuring run-scoring effects (difficult, but doable), I don’t think there is any amount of analysis that can be done that will accurately measure the effect of temp. on net team performance.

Paul G.
9 years ago

Ah, college baseball in the cold. It brings back memories. There is nothing like trying to score a game while sitting on a metal bench wearing multiple layers of warm clothing including gloves in 20 degree weather. It makes you seriously reconsider changing allegiance to surfing.

As previously mentioned, just because the game starts at such and such of temperature does not mean it ends there. I was at a college game in Florida where the game started in the 80s without a cloud in the sky. After a few innings a storm came in, dumped a ridiculous amount of rain with high winds for a few minutes, then went away and it went back to hot without a cloud in the sky. Couple of innings later this happened again. They called it after the third time it happened; it was raining so hard with so much wind behind it that it actually was painful, like getting constantly pelted with pebbles.

More germane to the discussion, I’ve been to the Oakland Coliseum for a night game. It was very comfortable at the game start. I was shivering by the end of it. Shouldn’t have worn shorts.

Mr Baseball
9 years ago

Nice work.

I think you have to show the Rangers record for all games and then compare it to the “hot” games to determine if there is any difference.

Also, I once took a chart of annual US median temps and ran it next to annual median runs scored in MLB per game and the correlation was very high, 80% high. Look it up. It’s amazing. Highest scoring era is still the 30s, this is also the hottest era in recorded US history. Coldest? 60s-70s, also lowest scoring. Outside of deadball era, which was also relatively cold. 2nd warmest era? 90s-00s, also the second highest scoring era.

Satoshi Nakamoto
9 years ago

Authors: Are you planning on researching how weather affects Starting pitcher performance? Low temperatures and also rain wind and humidity?

The Blue Jays bullpen fell apart in the 8th inning vs the Twins in very cold weather. It seemed the pitchers couldn’t throw strikes.

It’s be useful for fantasy to see if I should bench a starter in cold weather or if rainy conditions hurt performance (delays disrupt their routine, mound is muddy, etc).