(Matt Petitt is a financial services consultant who grew up and currently resides in Chicago. Matt also contributes at the Baseball Prospectus flagship site for the Chicago Cubs: BP Wrigleyville. Feel free to reach out to him at mjpetitt@gmail.com for any questions or freelance opportunities; you can follow him on Twitter @petitt_matt)
While there are subtle differences between the individual arenas in the NBA, NFL, and NHL, the underlying dimensions for each playing surface in each of these sports remains the same. But anyone who has ever seen the Green Monster at Fenway Park, or watched highlights of Willie Mays racing back to the 483 ft. mark in centerfield at the Polo Grounds to rob Vic Wertz, knows that the dimensions of major league baseball stadiums can vary dramatically. This makes baseball unique in terms of North American team sports, and one could easily argue that part of the allure of baseball is tied to the individuality that each park (especially Wrigley Field and Fenway Park) can possess. However, while it may be aesthetically pleasing, having different dimensions for each baseball stadium can cloud our judgment (for better or for worse) in terms of evaluating an individual player, whether he be a position player or a pitcher. To combat this blind spot, statisticians and major league front office members use a variety of measures to try and ensure that each player is evaluated through their true talent alone, and not solely by their enhanced (or diminished) performance in their surrounding environment. You might know this measure as park adjustments or park factors.
Park factors have been around since Theo Epstein worked under Kevin Towers in the San Diego Padres front office in the late 1990’s. Chances are, if you frequent Baseball Prospectus, Fangraphs, or Baseball Reference you’ve encountered park adjusted stats even if you didn’t know it. Examples include: weighted runs created plus from Fangraphs (wRC+), true average from Baseball Prospectus (TAv), and on base plus slugging plus from Baseball Reference (OPS+). Each website calculates their park adjustments slightly differently, but the main framework remains the same: Add the total runs scored and allowed at a team’s home ball park, divide the number of games at home, and then divide this figure by the same equation for a team’s road games. While this is better than non-adjusted rate stats, park adjustments often act as a broad sword when a scalpel is needed. For example, since it’s opening in 2009 ,new Yankee Stadium has consistently ranked among the top stadiums in home runs allowed. Generally, park factors assume that this environment makes it advantageous to all hitters, but given Yankee Stadium’s short porch in right field vs. its left center field wall of nearly 400 feet, the stadium actually plays much differently for left handed power hitters vs. right handed power hitters. Fortunately, now that portions of Statcast data (including exit velocity and launch angle) are being released to the public, we can improve upon the initial calculation of park factors. In this post I will examine how the thin air of Coors Field can affect the exit velocity vs. distance traveled of balls in play as compared to the other twenty-nine ballparks.
First, let’s take a look at the relationship between exit velocity and hit distance.
The above graph includes all balls in play for 2015 and YTD 2016, as such it includes ground balls, which even when hit hard tend not to travel past infielders. As a result the correlation coefficient (r) between exit velocity and hit distance for all BIP is 0.458. However, if we were to remove ground balls and limit the sample to include BIP with a launch angle ≥ 11 degrees (i.e. line drives and above), we see a much stronger correlation (r=0.747) between exit velocity and hit distance, as evidenced by the plot below.
This is pretty straightforward to regular baseball viewers: the harder line drives are hit, the further they tend to go. But how much further do they go at Coors Field? To find out, I first added a “park” column to the Statcast data to allow for batted ball data to be sorted by individual major league parks. I then divided line drives and fly balls into four groups, depending on the launch angle:
- Launch Angle: 11 degrees ≤ 20 degrees
- Launch Angle: 21 degrees ≤ 30 degrees
- Launch Angle: 31 degrees ≤ 40 degrees
- Launch Angle: ≥ 41 degrees
For the first and fourth plot above (Launch Angle: 11 degrees ≤ 20 degrees & Launch Angle ≥ 41 degrees), the Coors Field BIP don’t seem to stick out. This is not the case for the middle two plots (Launch Angle: 21 degrees ≤ 30 degrees & Launch Angle: 31 degrees ≤ 40 degrees). In both of these plots when controlling for launch angle and exit velocity, batted balls generally travel a greater distance at Coors Field than they do at the 29 other major league ballparks. This is particularly important since this is the range of launch angles in which home runs are typically struck. Take a look at the histogram I put together below regarding the distribution of launch angles of home runs hit in 2015 – YTD 2016:
The mean of this launch angle distribution is 27.9 degrees. Thus, Coors Field is enhancing balls that have already been “barreled” up, allowing well struck (and potential extra base hits) balls to travel farther than they would in the other major league parks. The architects of Coors Field anticipated this issue and designed the park to have cavernous dimensions. However, in doing so, these architects also created a field that has by far the largest amount of fair territory required for a defense to patrol, with much of the burden falling on the three outfielders. I’ve plotted below exit velocity vs. launch angle for whether or not a BIP allowed a hitter to reach base (i.e. Error, Single, Double, Triple, Home Run, Fan Interference) using Statcast data from 2015 – YTD 2016 for Coors Field vs. the other major league parks.
In short, not only does the thin air in Denver allow hitters to “hit it high and watch it fly,” but the deep fences that were supposed to normalize offensive output instead have allowed for bloop hits to fall in at Coors Field at a greater rate, all to the eternal frustration of Rockies pitchers and front office members.
Once more Statcast data is accumulated it will be possible to adjust hitters rate stats with a more accurate park factor utilizing exit velocity and launch angle. This is particularly pertinent as the Rockies stellar third baseman Nolan Arenado looks like a perennial MVP candidate, and it would be useful for voters to know which of his hits were just typical fly balls that carried in the Rocky Mountain thin air, or softly hit balls that fell in because outfielders weren’t able to sufficiently cover the vast Coors Field outfield. Additionally, since park adjusted stats are included as the hitting component of many WAR calculations, utilizing Statcast data within park factors would lead to a more accurate WAR estimate – which is something all baseball fans can get behind.
Exploring Baseball Data with R 