I’m sorry to burden you with another park factors article. My sense is this will be the last.
Having alluded to my prior articles, let me give you necessary background. First, I examined how ballparks influenced the barrel-to-home-run conversion rate in 2019. In other words, I determined, when hitters barrel the ball, in which stadiums those barrels were more or less likely to result in home runs. Then, I updated those park factors, expanding the sample back to 2017 with an interactive chart that also accounts for the handedness of the hitter.
I personally use those park factors for evaluating the home run hitting potential of hitters who have changed teams this offseason. But beyond home runs, they are lacking in applicability. And I wanted something useful for pitchers, too. Something that shows the net run-scoring envirnoment of each ballpark. Of course, pitchers are affected by more than just the park’s impact on home run hitting–if a park is more conducive to singles, double, or triples, that will negatively impact a pitcher, and vice versa. Isolating those park effects is the premise of this article.
wOBA and xwOBA
Maybe you’ve heard of wOBA and xwOBA, but aren’t sure how they operate. I’m going to give a short primer because it’s necessary for understanding the park factors I’ll derive.
Let’s begin with weighted on-base average (wOBA). Tom Tango’s wOBA attempts to assign a final numerical value to a hitter’s overall offense. Basically, calculating wOBA requires tallying up a hitter’s walks, hit by pitches, singles, doubles, triples, and home runs, and then weighting each by the value it has in proportion to its propensity to create runs (run expectancy). The weights change annually. Then, divide that total by the hitter’s at-bats, unintentional walks, hit by pitches, and sacrifice flies. As a measure of overall offensive value, wOBA is better than OPS, which undervalues getting on base relative to slugging.
Now that you understand wOBA, consider expected weighted on-base average (xwOBA) from Statcast. Don’t worry, we’ll be adding even more letters shortly. Unlike wOBA, the inputs for xwOBA include only exit velocity, launch angle and, on topped and weakly hit balls, sprint speed, as opposed to actual results like singles or walks. All the combinations of launch angle and exit velocity (with some including sprint speed) are assigned a probability of becoming a single, double, triple, or home run, and then plugged into the wOBA formula.
With that background, we can use wOBA and xwOBA to evaluate how various ballparks influence outcomes. For example, where xwOBA exceeds wOBA in a given venue, the park is curbing offensive output because expected run expectancy (sorry for the redundancy) is greater than the value of actual run expectancy in the park. That park would be a good place to pitch.
wOBAcon and xwOBAcon
As noted above, wOBA includes walks and hit by pitches. The denominator also includes strikeouts. To be sure, ballparks will have no impact on any of those outcomes. Therefore, we must remove them from the equation to isolate the effects different ballparks have on balls in play.
I mentioned above how there would be additional letters for our acronyms. Enter wOBAcon and xwOBAcon, which are simply short for wOBA and xwOBA on contact. These measures provide the run expectancy and the expected run expectancy, respectively, on those balls that are put into the field of play. They eliminate strikeouts, walks, and hit by pitches from wOBA and xwOBA. Comparing the two allows us to see how each park influences balls in play. We will ultimately look at the difference between wOBAcon and xwOBAcon in each park to derive park factors.
Notably, xwOBAcon does not factor in directionality (i.e., spray angle). Therefore, it undervalues pulled fly balls that are more likely to reach shorter fences and overvalues pulled ground balls that are more likely to die in the shift. Likewise, xwOBAcon overvalues fly balls hit straightaway–which are less likely to become home runs because ballparks are deeper to center–and it undervalues ground balls hit up the middle that are more likely to find the outfield than their pulled counterparts.
For our purposes, however, that’s quite alright. The reason is that xwOBAcon will undervalue and overvalue various balls in play equally in every ballpark, thus failing equally to capture directionality everywhere. Whereas wOBAcon does account for spray angle because it only considers outcomes (singles, doubles, triples, and home runs), so if a park’s short left-field fence is creating more pulled home runs for righties, for example, that will be reflected in the park’s wOBAcon mark. Accounting for that again in xwOBAcon and then subtracting xwOBAcon from the park’s wOBAcon (as we will do below) would essentially counterproductively cancel out the park’s effect in that regard.
I not only chose not to account for directionality, I also opted not to break these park factors down by direction. Individual pitchers have little impact on the ball’s spray angle, meaning that batters will control the spray angle of most balls in play. Indeed, a quick analysis of qualified starting pitchers revealed a meager 0.26 R2 between their 2018 and 2019 pull rates.
For evaluating hitters then, it’s extremely useful to consider which parts of the park are most beneficial to offense because they will benefit certain types of hitters more than others (e.g., right vs. left-handers, pull vs. balanced hitters, etc.). However, for pitchers, it shouldn’t matter whether a park is more favorable to the left or right side. A right-handed pitcher, for instance, likely gets the same or similar benefit out of Oracle Park’s deep right-field wall as a lefty.
Park Factors For Pitchers
With that foundation, we can employ wOBAcon and xwOBAcon. To recap, as applied to a ballpark, wOBAcon essentially provides the actual run expectancy created by all singles, doubles, triples, and home runs at the park, and xwOBAcon illustrates the expected run expectancy based on the launch angle and exit velocity of those hits.
The value of using wOBAcon and xwOBAcon is simple. Comparing the two entirely removes hitter and pitcher quality and just says, what was, as compared to what should have been, the result of the balls in play in this park, regardless of how many of those hits occurred in the park, who hit them, or who gave them up.
|Ballpark||wOBAcon – xwOBAcon||Park Factor|
|Great American Ball Park||0.023||1.72|
|Minute Maid Park||0.020||1.51|
|Globe Life Park||0.010||0.82|
|Citizens Bank Park||0.008||0.68|
|Guaranteed Rate Field||0.001||0.20|
I have several notes before diving in:
- These park factors are slightly different from Alex Chamberlain’s. First, mine are limited to 2017-19, as the parks have largely remained the same since then (except for Chase Field, but more on that below). Moreover, those years presumably account equally for the state of the juiced ball in light of the extreme offensive outburst across MLB after the 2015 season. Second, he applied some modifications to the wOBAcon coefficients. Third, he broke his park factors down by directionality which, as explained above, is unnecessary for our relatively limited goal of determining the best and worst parks in which to pitch. Fourth, I wanted to create actual park factors, as opposed to organizing parks in descending wOBAcon-xwOBAcon order. Fifth, I wanted to have a discussion comparing my park factors to traditional park factors. Sixth, and finally, I wanted to explain the derivation and applicability of wOBA, xwOBA, wOBAcon, and xwOBAcon so that you know exactly what they are and why I’m employing them. Still, as with all of his work, Alex’s article and park factors are excellent and worthy of application beyond the scope of this article.
- The third column reflects the park factors, which are z-scores. A z-score shows the relationship to the mean of a group of values, measured in terms of standard deviations—degrees of spread—from the mean. Where a z-score is 0, the value is equivalent to the mean in the sample. Where a z-score is 1.0, the value is one standard deviation greater than the mean. Z-scores are useful because, in a vacuum, the wOBAcon-xwOBAcon values lack meaning. They are only telling in relation to one another. For instance, if you don’t know the mean for the parks, then does the Coors 0.046 mark indicate anything?
- I limited Chase Field to 2018-19 data because the humidor was installed just before the 2018 season. The humidor has dampened the amount of offense in the park and, accordingly, including 2017 data may skew the effect of the park as applied today.
- Finally, I’ll reiterate that these park factors are most useful for evaluating a pitching environment, especially for selecting pitchers (and streamers). Unlike my previous barrel park factors, which were properly used to evaluate home run hitting, these park factors consider singles, doubles, and triples, as well as home runs, and then isolate the additional value the park is creating on top of those actual outcomes. In that way, these park factors remove the quality of the offense and pitching in the park and just look at the effect of the park on all offense. The only caveat is that they do not eliminate the effect of defense which, hopefully, would be negligible anyway given the size of the sample.
First, let’s compare these park factors to ESPN’s 2019 ranks, with the top rank being the best place to hit.
|Ballpark||New Park Factors Rank||ESPN 2019 Rank||Difference|
|Great American Ball Park||2||11||-9|
|Minute Maid Park||3||7||-4|
|Globe Life Park||4||2||2|
|Citizens Bank Park||5||10||-5|
|Guaranteed Rate Field||11||20||-9|
ESPN’s formula for its park factors is: ((runs scored at home + runs allowed at home/home games) / (runs scored on the road + runs allowed on the road/road games)). While ESPN’s park factors are limited to single year samples, they attempt to answer the same question we are asking: in which ballparks are you more or less likely to score runs?
Sort the above table by the best place to hit according to my park factors, and then answer that question in your own mind. Initially, consider which parks you would have at the top. Would that include Comerica Park, Marlins Park, and Kauffman Stadium? My answer would be no, but ESPN says yes. And as for the bottom, the best places to pitch, would you include Miller Park, Yankee Stadium, and Guaranteed Rate Field? My answer would again be no, but ESPN says yes. ESPN further surprisingly considers Great American Ballpark to be a relatively neutral run-scoring environment.
There are some conclusions, however, that using ESPN’s park factors help us draw more decisively than we otherwise could just by looking at the new park factors. For example, Oracle Park, Oakland Coliseum, Busch Stadium, T-Mobile Park, Wrigley Field, and Tropicana Field are good places to pitch according to both measures. Likewise, Coors Field, Minute Maid Park, Globe Life Park (which will be a dome next season), Citizens Bank Park, Oriole Park, and Nationals Park are good places to hit.
Perhaps there are a few surprising outcomes in my park factors. Citi Field and Nationals Park are not traditionally considered hitting environments. And PNC Park and Target Field are, ordinarily, regarded as pitchers’ havens rather than neutral parks. Likewise, Fenway is often reasoned to be a bad place to stream a pitcher, though that may have more to do with the Boston lineup and, for Boston pitchers, the other lineups in that division, than anything else.
Otherwise, I (humbly) believe the new park factors are a better representation of the effect that the park, alone, has on run expectancy than ESPN’s. In my view, this is most evident at the extremes. Having all of Busch, the Coliseum, Oracle, Comerica, Kauffman, T-Mobile, and the Trop in the bottom ten seems sensible. Moreover, with Coors, GABP, Minute Maid, Globe Life Park, Citizens Bank Park, Miller Park, Oriole Park, and Yankee Stadium in the top ten, I am confident the wOBAcon-xwOBAcon analysis is a better method for isolating the impact of the various parks on run-scoring.
Before considering individual pitchers, just take a moment to appreciate the effect Coors Field has on run expectancy. It is over three standard deviations from the mean. In other words, it has more than ten times the magnifying effect on offensive outcomes as Yankee Stadium, and more than double the effect of Minute Maid Park. No wonder Nick Pollack and other detractors were correct about Germán Márquez last year. In fact, there is no analog on the other end of the spectrum–in terms of pitchers’ parks–that is even close to having such an extreme impact on run expectancy. Similarly, the relative ability of GABP and Minute Maid to augment offensive outcomes is unparalleled across MLB. On the flip side, Busch Stadium and the Coliseum have more of a suppressive effect on run expectancy than most probably realize, with Oracle unsurprisingly slotting in just after them.
With all of that said, let’s apply these new park factors to the various offseason pitcher moves thus far. Remember, this analysis does not consider the new quality of opponents the pitcher must face or the lineup or bullpen supporting him. For instance, moving simply from the AL to the NL is beneficial due to the lack of the DH in the NL, but, for simplicity’s sake, that is omitted from the following analysis:
- Dallas Keuchel finds himself in a poorer pitching environment moving from SunTrust Park to Guaranteed Rate Field.
- Madison Bumgarner‘s value, while landing in a good place to pitch, nevertheless takes a hit because Oracle Park has nearly twice the stifling effect on run expectancy as Chase Field. The market is probably undervaluing this impact.
- Cole Hamels‘s value likely remains unchanged considering how similar Wrigley and SunTrust are.
- Zack Wheeler will have a more difficult time suppressing runs at Citizens Bank Park, which was more than double the distance from the MLB average wOBAcon-xwOBAcon as Citi Field.
- Believe it or not, Gerrit Cole gets a significant boost–in terms of park only, of course, as he may face more difficult lineups in the AL East–moving out of extremely hitter-friendly Minute Maid to Yankee Stadium. While Cole remains a first-round pick in most drafts, many are probably failing to recognize the significance of this move.
- Like Hamels, Hyun-Jin Ryu‘s move to Toronto and Kenta Maeda‘s move to Minnesota have little effect on their respective values.
- Alex Wood lands in a much more favorable environment moving from GABP to Dodger Stadium. He could be a sneaky add.
- Homer Bailey loses the extreme benefit of pitching in the Coliseum, instead having to pitch at neutral Target Field.
- Julio Teherán gets a much better home ballpark in Anaheim than he had in Atlanta.
- Wade Miley somehow found a worse place to pitch than Minute Maid.
- Tanner Roark‘s value improves quite substantially moving from GABP to the Rogers Centre.
- Dylan Bundy moves from a relatively unfavorable home ballpark to one of the best in baseball for pitchers. He, too, would be a sneaky late-round draft pick.
- Chase Anderson gets out of dangerous Miller Park and fortunately lands in Toronto, a neutral run-scoring environment.
As I’ve said in my previous ballpark articles, I will update these park factors after every season. Next season, for example, I’ll remove Globe Life Park and only include data from Globe Life Field, just as I limited the sample for Chase Field to 2018-19.
And, as always, please draw your own conclusions about the data. The first table is the most important aspect of this article, far more so than my discussion that follows about the individual players.
Featured image by Justin Paradis (@freshmeatcomm on Twitter)
It would be nice if it had the Team names rather than the stadium names which don’t mean too much to me
Ah sorry about that. When I update these next season I will certainly have the team names, if not both. Thanks for reading!
Isn’t wOBAcon impacted by the defense on the field? A bad ranging defense increases wOBACon because there would be more singles, doubles, and triples. If the home team fields a very good or very bad defense for multiple seasons between 2017-2019 wouldn’t that impact the wOBAcon stadium’s since half of the at bats at that park would come with that home defense on the field? I’m curious if the results would be different if you only evaluated the wOBAcon of the home hitters while the opposing defense is on the field so you don’t have the huge weight of the home defense impacting the wOBAcon results.
Also there’s a major issue with using wOBAcon—foul ball outs. The parks with big foul territories get a major boost because the foul pop out that’s caught in Oakland but not at Fenway only impacts Oakland’s wOBAcon. At Fenway it’s a foul ball and not considered “contact” but the pitcher loses an out but it’s not captured in wOBAcon.
I think that’s why Fenway/Wrigley are lower than what feels right and Busch/Oakland are higher than Oracle and Comerica. Factoring the benefit hitters get at the smaller foul territory parks should help make the HR happy parks like Coors, Minute Maid, and GABP less extreme outliers.
You’re correct foul outs are not included in the formula. According to FanGraphs, they have little impact on a hitter’s wOBA when included. The reason is clear to me: there are few pop-ups to begin with (about 7% is league average) and even fewer that would land in the space where they would result in an out in one park but not another.
I disagree with some of your park-specific analysis, which is why this is tricky when done ad hoc. With limited foul territory, there’s limited space to create foul outs (i.e., a benefit for the hitter). I suppose Fenway and Wrigley may move up slightly for that reason. However, with more space to create outs in places like Busch/Oakland, pitchers would get a benefit, yet those parks are already the two lowest on the list. I’m not sure accounting for foul outs would be helpful because it would just move the latter two parks down and the former up.
As for defense, I think it’s ultimately noise that will be swallowed up by the size of the data set. Here, we have three years of a sample. That means not only three years of away defenses hopefully purging any effect by the home defense, but also three different home defenses. Even if the players haven’t changed (which they likely have for each team in some capacity), players age and their defense changes accordingly.
Maybe you should take the average for the home team so it will equal 1/30 instead of 15/30? I honestly don’t know if that makes sense as I’m not knowledgeable on this topic.