article on Wrigley Field, in which I pointed out that Wrigley has not been a
great hitters’ park for the past two seasons, generated some e-mail questions
about park effects. I get a handful of these whenever I mention that this park or
that park is good or bad for run scoring, but I’ve never addressed them in a
Paraphrasing, the questions look something like this:
You say that Wrigley Field looks more like a pitchers’ park. Couldn’t that
be because the Cubs have had such lousy offenses the past few years, while
their pitching has improved? If you put the Indians in Wrigley Field,
wouldn’t it look more like a hitters’ park?
The short answer is, "no."
Before I go further, let me say that the methodology I’m about to explain is
how park effects have been calculated up until now. At the end of the
column, I’ll address some of the problems we’re running into this season.
Park effects are traditionally calculated by dividing the number of events
by both teams in all of a team’s home games by the number of those events in
a team’s road games. For example, in 1996, there were 813 runs scored,
total, in Pirates home games. There were 796 runs scored in Pirates road
games. Dividing the first number by the second yields 1.05, which we state
as a percentage. "In 1996, Three Rivers Stadium increased run scoring
Calculating park effects this way cancels out the impact of an individual
team’s characteristics, because their performance counts just as much as
that of the rest of the league. That the Braves have good pitching won’t
distort the numbers of Turner Field, because they carry that good pitching
with them on the road. If the Indians have a great offense and no pitching,
they’ll have that for 81 home games and 81 road games, and against all the
You can do this for any statistic. In 2000, 137 home runs were hit in County
Stadium by the Brewers and their National League opponents (see below for
more on that last clause). In Brewers road games in NL parks, 181 home runs
were hit. County Stadium’s final-season park effect on home runs was -28%.
I guess I should explain all of these caveats. Up until recently, each park
saw the same distribution of teams each season, making this method fairly
reliable. Park factors compared the same players playing in the same parks
throughout the league, which is why doing the factors this way worked. The
park was the variable, not the players.
Well, the introduction of interleague play changed that. The distribution of
teams playing in, say, Wrigley Field, was different from the distribution of
teams playing at Shea Stadium. Over the past few seasons, STATS, Inc.
calculated its park factors by ignoring interleague games, a fair solution.
This season, though, the unbalanced schedule has completely changed the
distribution of teams playing in various parks. Park factors can be
distorted, because the mix of teams playing in one park is nothing like the
mix of teams playing in another. Simply dividing runs scored in a team’s
home games by runs scored in a team’s road games is dangerous, because teams
aren’t playing home-and-home series, and are playing more games within their
division than outside of it. A different overall set of players is
responsible for the runs scored in every major-league park.
Given the importance of park factors in much of the work we do, it’s
essential that we account for this in evaluating parks, and by extension,
players. Clay Davenport’s park factors, for example, are actually weighted
by games played in each individual park, something he’s done for years. (The
unbalanced schedules used by many minor leagues caused him to develop his
The unbalanced schedule brings other issues into play: individuals will play
in an unequal distribution of parks. NL East pitchers, traveling to Shea
Stadium, Pro Player Park, and Turner Field, appear to have an easier time of
it than NL Central pitchers taking trips to Enron, Cinergy, and the new
parks in Milwaukee and Pittsburgh. How can we account for this when we
compare their performances?
Beyond parks, what about facing an unequal distribution of teams? It seems
clear that, say, your average NL Central pitcher will have faced a easier
slate of teams than your average NL West pitcher in 2001. Is this a
significant factor, and can we account for it in player and team evaluation?
One thing is certain, though, as people a hell of a lot smarter than me work
on optimizing park-effect calculations, the systems they devise will all
strive for the original goal: isolating the effect of a park on performance,
independent of the team that plays there.
Joe Sheehan is an author of Baseball Prospectus. You can contact him by