As the point person for the ESPN fantasy baseball projections, deciding how to handle 2020 MLB park factors is one of the many challenges associated with generating the 2021 MLB projections. The efficacy of incorporating park indices from a 60-game schedule needs to be evaluated.
By means of a brief review, each major league venue influences everything happening in a different manner. An index for every recorded statistic can be generated. Some are more useful, such as runs, hit and homers. Others, such as walks and strikeouts are quite important but not discussed nearly as much.
The conventional means of computing park factors compares the outcomes of the specific event at home and on the road. For example, to determine the runs index for Wrigley Field, the runs scored for Cubs hitters and allowed by their pitchers at home are compared to the total of their runs scored and allowed on the road.
The formula is designed to flesh out as much bias with respect to quality of opponent as possible. However, a 162-game season isn't enough to completely normalize all factors, so park factors are expressed as a three-year average.
This begs the question, "If a three-year average is needed, how trustworthy is a factor based on a two-month 2020 season?"
Unfortunately, the answer is, "It's not."
The best way to demonstrate this is looking at park factors in two-month intervals. Data from 2017-2019 will be used, breaking the season into April-May, June-July, August-September. March games are rolled into the first segment while October contests are included with the last group. Factors for runs and home runs will be investigated using standard deviations.
Two different studies will be conducted. The first looks at each team's two-month factors per season from 2017-2019. The second evaluates each team's first third, middle third and final third factors from 2017-2019. As such, the standard deviation between three factors is calculated.
Here is the data from 2017-2019, looking at the standard deviations between the early, middle and final third of the season park factors.
For this analysis, let's use a venue calculated to be neutral (100) for homers in 2020. Based on an average standard deviation of 16, the actual factor could be between 84 and 116. This sets the range of a hitter projected for 30 homers in a neutral venue to be between 27.6 and 32.4. Granted, every projection should be considered a range. However, the variability in the park factor further widens said range. A five-home run difference, along with associated runs and RBIs separates players by at least one tier, maybe two.
Here, the average standard deviation is around 14. Applying the same process, a pitcher projected for a 3.60 ERA in a neutral venue has an ERA range between 3.35 and 3.85. That half a run of uncertainty on top of the already present variability associated with projecting ERA.
Keep in mind, 2020 factors would be averaged with the two prior seasons, so the above ranges would be lessened in a practical sense. But still, using 2020 park factors adds a lot of uncertainty to an already variance-laden process.
At this point, it's fair to wonder if the study should be conducted on just the final two-month segment, since the 2020 campaign was essentially played the last two months of a normal season. Here is the data looking at the standard deviation between each segment (early, middle and late) from 2017-2019.
The variance is the same, a standard deviation around 16. That is, using just the data from August and September does not reduce the uncertainty.
Again, the standard deviation matches the runs factor from the first study, resulting in a similar level of ERA noise if only the August and September data is employed.
There really isn't a specific standard deviation dictating whether 2020 park factors are sufficiently reliable to be used in the 2021 projection process. It's more of a subjective feel and, quite honestly, after evaluating the above data, my lean was to ignore 2020 factors and use the three-year average from 2017-2019.
Still, the decision was uncomfortable, until I realized there's another consideration. The geographical schedule completely skews the relative nature of the park factors. In essence, there were three different leagues: East, Central and West. An individual team's home park indices were only compared to nine other yards, those within their region.
This is an extreme example, but let's say the ten most homer-friendly venues were in the same region with the ten least homer-friendly in the same region. When determining the home run park factors, there will be a park in each region with a factor close to 100. However, the three parks are not the same for homers, it's just they came out in the middle of their subset.
Even if the above studies indicated park factors generated from two months of play are reliable, the specific results from last season could not be included in the three-year average. It's not an apples-to-apples incorporation.
As such, the 2021 projections utilize the same park factors used for 2020. They're simply more trustworthy than using 2020 park data.
Unfortunately, it's not that easy. Last season was the inaugural campaign for Globe Life Field as well as the first year following renovations to Marlins Park and Oracle Park. Further, after the season it was revealed humidors were installed at Fenway Park, T-Mobile Park and Citi Field. It's still unclear where Toronto will play home games. Finally, it was just announced five more unnamed venues will be equipped with a humidor and the 2021 baseball will be less bouncy. Good times.
Here is how each of the above will be handled in the 2021 MLB projections.
Globe Life Field
We are essentially where we were heading into 2020, having to make an educated guess. Based on dimensions and the assumption the roof would be closed, Globe Life Field was assumed to be far less of a hitter's park compared to Globe Life Park, checking in around neutral. However, and this is admittedly anecdotal based on batting practice, players indicate Globe Life Field plays much smaller with the roof open. In last season's abbreviated schedule, the roof was open only once -- and it was a slug fest.
Obviously, one game means nothing, but there's a chance the roof is open more early in the season, and science tells us it should play more hitter-friendly in the open-air conditions. The problem is the number of games contested with the roof open is unknown, as is the level of the corresponding adjustment. As such, the same speculative factors used for 2020 will be carried over to 2021. Just be aware the range of outcomes surrounding Texas Rangers projections is larger than those with more reliable park factors.
Marlins Park
A small subjective adjustment was made heading into 2020, accounting for moving the right field fences in a few feet. There weren't ample games played to gauge the actual effect (and won't be for three full seasons), so the empirical adjustment was carried over to this season.
Oracle Park
Now things get hairy. Like Marlins Park, the fences in right field were moved in prior to the 2020 season. Therefore, the same logic should apply with respect to simply using the 2020 indices. However, there were numerous reports how the wind currents were different after the club boarded up the open-air viewing area in right field. Throughout summer camp and into the season, San Francisco Giants hitters and pitchers marveled at how well the ball was carrying and seemingly noticed different wind patterns. There are other factors in play, such as an unusually warm summer, but the extent of the player's narrative was alarming.
Projections are predominantly a data-driven process. That said, observation is integral to the scientific method. There isn't ample data to validate the assumption, but there is strong circumstantial evidence Oracle Park will play less pitcher-friendly if the holes in right field remain covered. Ergo, subjective alterations were made to Oracle Park's indices. I still assumed it to be pitcher-friendly, just not to the same extreme.
From a drafter's perspective, the way to account for the uncertainty is taking chances on some middle- and lower-tier batters, such as Mike Yastrzemski, hoping the park indeed is friendlier to batters. On the flip side, be careful chasing the likes of Kevin Gausman, who stands to benefit from working in a pitcher-friendly venue after spending most of his career in launching pads.
Fenway Park, T-Mobile Park and Citi Field
Adjusting for a humidor isn't as simple as determining what happened in Colorado and Arizona and then applying it to the other venues. The reason for the humidor in Coors Field and Chase Field is the baseballs were drying out. The causes were slightly different, but the effect was the same: harder, bouncier balls. Storing in a humidor added moisture, reducing the elasticity, hence the velocity it came off the bat.
The conditions in Boston, Seattle and New York City aren't anywhere close to that in Denver and Phoenix. The non-humidor balls weren't nearly as dry, etc. This is extreme, but depending on the location and how the balls are stored, putting them in a humidor could in fact increase the bounciness if the original ball trapped more water. Again, this isn't likely, but it's plausible. The bottom line is unless we can gauge the difference in the baseball with and without the humidor in those venues, it's not wise to make any adjustments.
Things are already wild enough; hopefully, the Blue Jays can play in Rogers Centre, but that is out of MLB's hands. Last season, Toronto played many of its home games in Sahlen Field, home of their Triple-A affiliate, the Buffalo Bisons. With no minor league season last year, the park was free, there were no conflicts.
The same will not be the case this summer, hinting TD Ballpark, the Blue Jays spring training home in Dunedin may be the choice. The venue's dimensions aren't extreme in either direction, though it's unclear how the weather may affect things. The initial speculation is the yard will favor pitchers. However, for now, it is assumed Toronto will play half its games in Rogers Centre.
2021 baseballs
It has been reported the 2021 ball will be a little looser than usual, rendering it less bouncy. The effect will be a reduction in exit velocity, which is much more pertinent for power than hits. However, the ball is also supposed to be smaller, which could counteract the exit velocity as a smaller ball will incur less air resistance, hence fly balls could travel further. The effects can be modeled, but until the new balls are used in games, it's still a guess, albeit educated. Barring further information, no changes will be made resulting from this announcement.
That said, it's safe to assume the sluggers with high-end exit velocity on fly balls will be least affected. There could be a significant group of batters for whom losing five to ten feet of flight will be detrimental. This group encompasses those most benefiting from the 2019 ball. While regression is already coming for those hitters, an even greater correction is in store if the ball indeed loses five feet of distance.