Taming the Wind
“This will be one of the most beautiful baseball parks of all time.”
—Vice President Richard Nixon, 1960, opening day at Candlestick Park
So many were fooled.
Players, sports writers, and even future presidents were in awe of the new home of the San Francisco Giants. Mesmerizing visitors with its lush green grass and the largest scoreboard in the league, Candlestick Park boasted an architecture that had never been seen before in major league baseball. “The Stick” was proclaimed as a man-made marvel, and would start a contagious boom of modern baseball stadiums in the 1960s that wasn’t unlike the way Oriole Park at Camden Yards inspired an explosion of retro ballparks throughout the 1990s and beyond.
Except Candlestick never had a shot. No one realized it at the time, but the Giants and their opponents would be sentenced to nearly four decades of ridicule and embarrassment for a park that was doomed before the first shovel hit the ground.
And it was all preventable…all it would have taken was a wind study.
Budding out onto the San Francisco Bay, with the Pacific Ocean on the opposite side of the peninsula, Candlestick’s gusting winds, cold temperatures, and chilling fog mocked the way the game was supposed to be played. The winds inside the stadium wrecked havoc like the place was possessed. Routine flies had outfielders doing the Foxtrot trying to guess where the ball would fall. Sudden gusts threw players off balance randomly like ghosts playing pranks. The park and its flaws were nationally exposed when Giant pitcher Stu Miller was famously blown off the mound and called for a balk during an All-Star game.
When the Giants expanded the seats of Candlestick in ’71 to fully enclose the joint and accommodate the San Francisco 49ers of the NFL, the team thought they had their wind issues licked. Architecturally, it was a huge makeover that even had the engineers optimistic that the field would be better shielded.
Negative. The ferocious winds remained and actually became even more schizophrenic than ever.
In 2000, the team finally did the right thing, and I’m not referring to their move five miles north into the new Pacific Bell Park (now AT&T Park). The Giants sought the advice of engineers skilled at studying air currents, and boy, did their advice come in handy. They wound up rotating the new park a quarter turn from its original plans, preventing what could have been another atmospheric debacle. With the help of other precautions such as adding windscreens, the Giants were finally able to temper the gales of the Pacific inside their new home, freeing themselves from the dizzying 60+ mph gusts that haunted Candlestick.
Baseball cities never intend to make their home field play like a freak show. Outside of complying with the standard 90-foot bases and 60’6″ mound distance, teams have the freedom to use a blend of custom fence curvatures and adorning facades—along with the jumbo scoreboards and picturesque skylines—to inject their own signature style on the park’s design. It gives fans a special “visit to the ballpark” feeling like no other sport. But as Candlestick proved, there’s plenty of opportunity for unplanned anomalies if the climate isn’t given its due respect.
In parts I and II of this post, we talked about how a ballpark’s dimensions and atmosphere can impact the performances on the field, and the need for a new rating system for measuring true ballpark impact (BPI). But considering what the Giants went through with Candlestick, do we even have a prayer in coming up with a way to account for the fickle force of wind?
Keep the faith. Science does a great job explaining what wind is, and how and when it occurs. Wind comes from the movement of air from high to low pressure areas in response to changes in temperature as different parts of the earth heat up and cool down. Think of an ocean breeze coming in from the shore. As the land temps rise, the warming air molecules want to “spread out,” rising into the sky; the cooler air pushes in from the ocean to replace it. That’s wind. In a nutshell, wind is simply nature’s attempt to satisfy the laws of thermodynamics by equalizing atmospheric conditions.
Weather patterns across the globe are heavily influenced by the warmer environments near the equator, the cooler ones at the poles, and the spinning of the earth on its axis, with the sun playing maestro using its radiant heat. But we’re primarily interested in the local picture here. Each major league ballpark is situated in a unique wind environment that’s affected by the presence of surrounding hills, flat land, valleys, mountains, bodies of water, and man-made structures.
The Reds’ Great American Ballpark, for instance, stands between the Ohio River and plenty of wind-blocking structures from downtown Cincinnati.
In Kansas City, there are very few buildings or bodies of water near Kauffman Stadium. But elevated mounds of earth nearby will factor into how the winds funnel into the Royals’ ballpark.
The first step into taking a park’s atmospheric environment into account for our BPI system would be to use historical weather data. The seasonal range of temperatures and wind behavior that each ballpark will get exposed to throughout the baseball season helps establish the best and worst weather conditions a baseball will face when it leaves the bat.
But we’re still missing a major component, and that’s figuring out how the air will flow around and through the ballpark’s structure and onto the playing field. A ballpark’s architecture has many properties that can alter the way the winds impact the play on the field. Unless we have a way to quantify this behavior, we might as well just toss a few blades of grass in the air and look at the overhead flags.
Taming the wind is so tough that most experts have punted on accuracy and elected to use average readings. The problem in doing this for measuring BPI is that the margin of error is huge. Wind is just too volatile to be throwing darts. In his book “The Physics of Baseball,” Robert Adair noted that every 1 mph of tailwind adds 3 feet to a 400-foot ball flight. What use is a BPI reading based on average wind unless that exact wind was blowing that day?
If we can’t be as precise as possible here, why bother? It would defeat the purpose of trying to quantify BPI in the first place, like building a new baseball field but pacing out the base paths. I realize we’re not exactly predicting Tsunami strikes here. But we’re living in a sporting world that’s increasing its analytical muscle by the season. The type of research being spewed from baseball junkies making valiant sabermetric contributions is all about carrying their analytical viewpoints to greater decimal places. Accuracy, not fudge factors, is the stamp of their integrity.
If you think I’m just referring to basement-dwelling stat geeks, check out the awareness factor at the MIT Sloan Sports Analytics Conference. Having sold out the Boston Convention & Exhibition Center early last March in just its fourth annual gathering, the one-day affair with the think-tank flavor has a 400-person waiting list. This past session found Dallas Mavericks owner Mark Cuban and Moneyball author Michael Lewis knee-deep in discussions on using sports metrics, all the while cementing the new principle in sports that in-depth statistical analysis makes business sense.
So I’m not ready to cut corners just yet. Besides, we do have some technological momentum going for us. Back in the 1960s, civil engineering genius Alan Davenport showed the world that not only can wind behavior be modeled, its study can prevent disaster. Revered for his understanding of how wind impacts large structures, Davenport used his Wind Tunnel Laboratory at the University of Western Ontario to analyze and test most of the tall buildings that were erected during the 1970s. He brilliantly figured out answers to catastrophically-relevant questions such as how fast a building should be allowed to sway within its meteorological environment. His fingerprints are all over the safety planning for some of the most famous buildings in the world, including the CN Tower in Toronto and the Normandy Bridge in France.
The Davenport-driven revolution in wind engineering came a little too late for the Giants and Candlestick Park. But by the turn of the 20th century, most major league teams were standing in line ready to take full advantage. Since the early 1990s, clubs have been shelling out $20K to $80K in wind analysis studies on their own new ballparks to understood how the atmosphere and their architectural designs will impact the comfort of their fans as well as the play on the field.
In the middle of this storm of investigations has been Rowan, Williams, Davies & Irwin Inc. (RWDI), a Canadian engineering firm that has lent their two cents into the construction of many of these new parks, including PNC Park in Pittsburgh, Chase Field in Arizona, and the Rogers Centre in Toronto, to name a few. It was RWDI who warned the Texas Rangers that unless they did something about blocking off the wide-open space in centerfield for their new ballpark (the original design called for the stands to end before center field) a baseball hit with any intentions of leaving the premises wouldn’t stand a chance against the blast of air that was expected to blow in constantly from the Texas plains. Up went an office building with huge advertising panels to close the gap, ensuring that muscle-bound Ranger sluggers Juan Gonzalez, Ivan Rodriguez, Rafael Palmeiro, and Jose Canseco would play key roles in those prolific home run hitting years of the 1990s.
Using the same trajectory theory we talked about in Part II, RWDI combines theoretical ball flight information with simulated wind characteristics to determine how typical fly balls will react to the real-life atmospheric conditions of a ballpark. Computationally, it’s a bear. The simulation time for just one wind direction takes days even with a dedicated cluster of computers.
Not that it’s a completely virtual exercise. In the spirit of Alan Davenport and his Wind Tunnel Laboratory, RWDI builds a mock setup of a ballpark’s physical environment, applies sensors to whatever parts of the structure that need monitoring, and then exposes the entire area to a finely-orchestrated wind assault that’s tailored to emulate the actual conditions of the target site.
Once the data has been crunched and the aerodynamic equations solved, RWDI can project the wind impact on practically any type of structure—like they did for the Burj Khalifa, the tallest man-made structure in the world.
So why not re-use these wind analysis studies for our BPI system? The technology is obviously there and the type of insight these studies give is exactly what we’re looking for.
I’m just not sure how we’re going to get our hands on the data.
I spoke with Jon Lankin, a project manager at RWDI, about these ideas. Before his 22 years at RWDI, Lankin learned about wind engineering as a student of UWO from one of the best—Dr. Alan Davenport. Lankin uses an analogy to describe the nature of wind similar to the one I used in part II when describing air density as a “ball pit.” He says that wind is similarly made up of balls he calls gust cells, except these cells are all different shapes, averaging about 100 feet across. It might take three seconds for a 20-mph gust cell to pass you by.
Lankin’s experience at RWDI during the construction of all the new ballparks means that he “gets it” from a BPI perspective. He agrees that the basic premise of using wind engineering to help establish true measures of ballpark effects is the right way to go. “I like what you’re trying to do,” he told me.
So can RWDI help out?
“We’re fans of the game,” he admitted, telling me in so many words that he’d love to see a new BPI system happen, “but the wind studies we’ve performed belong to the clubs who paid for them. That information is proprietary…”
“…but with consent we would be pleased to provide access.”
Hmmm. That sounds like the crack of a door left open.
“Our main focus is on the construction of stadiums,” he went on, “but we’d be more than happy to calibrate every park [for BPI purposes].” He didn’t have to tell me that there’s a significant cost involved with what he just stated…perhaps one to two million dollars in consulting fees for evaluating all 30 ballparks.
Good luck with that, right?
Don’t turn out the lights just yet. If the challenge of getting our hands on ballpark wind information seems a bit on the insurmountable side, think about the synergies that have been struck between major league baseball and the technologies that are driving sabermetrics to the next level. The game’s revenue is mind-boggling, and the bucks are flowing in demand for what those high-tech ’f/x solutions have to say about performance. And it’s not just for cool TV and internet graphics.
If PITCHf/x can diagnose the failures of an ace pitcher to his release point or pitch selection, how much is that worth to a club owner or GM trying to keep their fiscal turnstiles spinning? HITf/x is making us re-think the way we evaluate hitters by exorcising all of the misleading gunk out of “event-driven” baseball stats like outs and hits and cutting right to the chase—what happens when bat meets ball. FIELDf/x, coming soon to a ballpark near you, will someday prove which centerfielder covers the most ground using precise measures of reaction and distance—and if we cross our fingers and say a prayer, it just might help coax baseball into redefining how Gold Gloves should be awarded.
The point is there’s a world of baseball analysis that has barely been tapped. A new and improved BPI rating system could be another catalyst to that revolution. Here are a few examples how:
- Managers will appreciate the intel they can get on how a ballpark will play on a certain day considering the weather and type of pitcher on the mound for the opposing team. Yes, it’s true that most of these field generals like to go with their instincts, but sometimes putting a number on a feeling is enough to convince someone that it’s real.
- What if that pre-game BPI report told a left-handed pull hitter that any ball hit in the air toward right—no matter how hard it was hit—will turn out to be a harmless pop-up or fly-out because of a persistent incoming 30-mph jet stream? I’m sure he’d re-think his approach at the plate.
- Matching the compatibility of a ballpark’s BPI signature with the hitting profile of a free agent candidate or trade prospect should have any GM’s attention before he commits to an $85 million contract for a player who will play in that ballpark 81 times a season for the next several years.
- BPI could become the go-to metric for performance numbers that are compensated for the environment, replacing Park Factors in that department.
These reasons, and the fact that even some of our Grade-A baseball researchers have been making off-handed remarks about the shortcomings of Park Factors for years, tells us that the motivations are there to make BPI happen someday really soon.
Let’s take a look at what this new BPI system might look like.
First of all, we’ve got to get away from the idea that each ballpark is going to be stamped with a single BPI number that’s supposed to tell us everything we need to know about how it affects performance. There’s just too much relevant information hidden inside the box just screaming to get out. Let’s start off with the two components we’ve been talking about since Part II. Consider the following two types of BPI ratings for measuring home runs:
Physical Environment Score (PHYS)
Performance Feedback Score (PERF)
The PHYS score measures the difficulty of hitting a home run in a ballpark considering only the physical environment. It’s based on a scale of 0 to 100, with 100 being your backyard, and 0 being the Grand Canyon. I’m partially kidding. I can envision these limits defined by some theoretical mock-up of two ballparks that represent the hardest and easiest sites to hit a home run, with parameters that would fall just beyond the realm of how a real ballpark would ever get built. For instance, a HR-PHYS score of 0 could represent a ballpark located at sea-level with greatly-stretched fence dimensions, say 400-425-450-425-400 (LF-LC-C-RC-R), combining the heaviest air with the longest fence distances. On the flip side, a HR-PHYS score of 100 might correspond to a park located at the highest elevation that a ballpark would ever be built, with short fence distances measuring something like 325-350-375-350-325.
I did think about using a neutral score approach where a score of 100 represents a middle-of-the-road ballpark, similar to Park Factors. But what would be that middle-of-the-road here? Even if we collected several years of HITf/x data, figured out the average distributions for the speed and angles of a batted ball by a major leaguer, and came up with a set of fence distances that fell dead-center of those distribution plots, we still have some issues because that data may shift from generation to generation. Think about what those numbers might have been in 1979 compared to 1999. Very different. We’d be introducing performance feedback—bias—into our fundamental makeup of BPI; we want to leave that for our PERF scores. We want to keep our PHYS rating free of anything other than the physical environment of the ballpark being rated.
I can see the PHYS score being split into a raw score and a weather-impacted score. The raw score captures the vulnerability of a ballpark considering only the fence dimensions, elevation, the range of possible launch angles and spin, and the minimum speeds of the ball off the bat needed to hit one over the fence. We’d set the wind to zero and use a nominal temperature of 72 degrees Fahrenheit here. It’s a thumbnail measure of a park’s BPI in neutral weather.
Our weather-impacted score would factor in the fluctuating range of temperatures and wind for a site. Let’s face it—parks like Wrigley Field need this type of number. Like many fields in baseball, Wrigley can play long or short, and most of the time this depends on the moods of the wind. We’ll need our wind study information here, however way we get it. Jon Lankin makes a good point in that the outfield should be divided into five sectors since the air flow in the three-dimensional airspace above each could be very different.
We’re also going to need some feedback scores for our BPI system. Where PHYS scores give us a theoretical flavor of how a park should play, PERF scores tells us how it actually did play. One excellent reason for doing this is the observance that players and teams tend to play according to a ballpark’s vulnerabilities. It’s why you can’t use performance feedback exclusively as an indicator of ballpark impact, a la Park Factors, because that performance gets biased and skews the BPI rating from what its true measure should be.
Go back 50-60 years. Shortstop Alvin Dark was a singles hitter when he started his career with the Boston Braves, hitting just three home runs in each of his first two full seasons. Then he was traded to the New York Giants, who played at the Polo Grounds. Dark took advantage of the dimensions of the Giants home park, learning to hook the ball to the only reachable part of the fences for him—right down the left-field line, 279 feet away. Over his six seasons there, he hit 73 home runs at home, but managed just 23 everywhere else. Powerful third baseman Sid Gordon made similar adjustments as a New York Giant around the same time frame. After being frustrated by having his long drives frequently run down by outfielders in the deep gaps of the Polo Grounds, Gordon changed his grip and learned to pull like crazy, turning his long outs into four-baggers. By playing to the ballpark but not necessarily to his strengths, Gordon transformed his career at age 30. In both cases, Dark and Gordon biased their own performances relative to what the park would nominally be willing to give.
Figuring PERF scores for home runs would involve the techniques used by Alan Nathan and Greg Rybarczyk that I described in parts I and II. The initial launch parameters of fly balls hit in major league parks that are home run candidates would have to be collected and analyzed so that we can determine how a park’s atmosphere and dimensions influenced the ball’s flight and ultimately, its outcome (home run, fly out, etc.).
The card below shows a sample of what the BPI ratings for the physical environment at Citi Field might look like. The numbers are just guesstimates on my part; they are not based on any real data except for my casual observance of how those parks play and their Hit Tracker data. The raw PHYS score of 23 on a scale from 0 to 100 shows how difficult it is to hit a home run there. The “30″ in parenthesis represents how this score ranks with the league—dead last.
Below is a BPI card for Citizens Bank Park. Again, I’m just taking liberties here, but I imagine that the Phillies’ home ballpark, rated a 65 here, is above-average for hitting home runs under normal conditions because of some of the park’s friendlier dimensions. But when the wind starts kicking, its BPI impact would theoretically jump as high as a score of 83 out of 100, which is what the weather PHYS min-max range of “56/83″ represents.
The positive number (+24%) for the overall impact of weather on Citizens Bank park shows that weather makes the park more vulnerable to home runs, meaning the wind is more apt to blow out, blow stronger, and/or the temperatures are more likely to be higher than average. Citi Field’s -8% indicates just the opposite, that weather tends to make Citi Field an even more difficult place to hammer one out.
This BPI system doesn’t recognize “handedness,” a classic knock on Park Factors. A ballpark’s impact on right-handed and left-handed batters can be very different. No doubt about it. But there’s a broader scope to this that makes me think that we can’t get away with doing “splits” with BPI. It has to do with a hitter’s hitting profile, or his tendencies for spreading the ball around the field. Today’s ballplayer arguably sprays the long ball more than previous generations (i.e. before 1990). Using left-right splits would put too much spin on the analysis. It would be flat-out wrong for hitters like Ryan Howard and Manny Ramirez, who, as you can see from their spray charts, like to scatter their blasts without directional preference.
One cool way to counteract the lack of explicit handedness in BPI would be to define a player’s hitting pattern in a way that can be matched up to a ballpark’s BPI signature—call it ballpark compatibility. It would be a great way of speculating how well a player would perform in a certain park.
Yes, we are very interested in using BPI ratings for more than just home runs. It’s an important part of the sabermetric fabric to figure out how the ballpark influences run scoring and defensive ability—performance measures that really matter to winning, building a franchise, compensating players, and so on. To do this, we’ll need to incorporate the precise geometric shape of the entire field, including foul territory, just as a start. There’s so much more to it; I won’t be covering it here.
The technology for giving major league baseball a new BPI rating system is 90% there. There’s nothing holding us back from coming up with a raw physical environment rating for every ballpark today. With trajectory information already being logged at every ballpark, we should be able to come up with the performance feedback numbers as well.
That last 10%, our last mile, requires the use of sophisticated wind studies to gauge the impact of weather. Judging from what’s been accomplished with baseball’s ’f/x technologies, you gotta believe that sooner or later, there’s going to be an app for that too.
I hope you enjoyed this three-part series on Park Factors and the introduction of new ways to measure ballpark impact. I have to admit that it grew in size as I was writing it. I hope you were able to tag along.
Thank you, Jon Lankin, for our insightful conversations on these topics and taking an interest in how wind studies could help make for better baseball analysis.
“They tickled their ears and fed them muffins.” —Dimitri Cappello, 11, on what a swarm of butterflies did to distract an army of crocodiles in his imaginary tale about a conflict between the crocodiles and the unicorns.