Some Historical Park Factors
October 9, 2008 1 Comment
Ballparks come and go. Management moves the fences in and out. How much of an effect do these changes have on a player’s statistics?
Normally, half of a teams games are played at their home park, while the other half gets spread around the various other parks. A park factor will describe how much that home ballpark increases or decreases runs or any of the other batting statistics relative to the rest of the league, but each team has a different mix of road ballparks, and those can change even when the home park doesn’t.
There are many sets of park factors available, each with somewhat different methodologies. I wrote an article for Seamheads in June which described my approach. Here’s a summary of principles I used:
1. If there is a change in a park’s configuration, treat it as a new park by assigning a version number
2. For the period of time the park doesn’t change, it should have a single set of home factors
3. The team multiplier, which may change every season even if the home factor doesn’t, is a mean of all the home and road parks the team played in each season
To calculate my factors, I compared the home and road totals for the entire period of a ballpark version. The road stats were then normalized one park at a time with these preliminary factors, and the factors calculated again, and then a third time. Data processing was done on play by play data, courtesy of RetroSheet.
Weather and luck are examples of random factors that can effect the data being sampled. I ran a test of National League ballparks during the seven year period from 1985 to 1991, when there were no changes in park or schedule. Defining the seven year totals for each park as their “true” value, I calculated the total root mean square error of any consecutive one, two and three year periods to the seven year values. Except for triples and homeruns, which show the most variance, even one season was enough for one decimal accuracy. Three seasons was sufficient to get reliable results for all categories.
Links to spreadsheets:
Park factors by version, all batters and also split by batting right or left
Yearly team park multipliers
Yearly league park multipliers
Here are the high and low team factors for 2007, combining both home and road.
Historically, the AL’s HR factor has ranged from a low of 0.91 in 1966-1967 to a high of 1.05 from 1995-1999. The AL’s HR factor has been above 1.00 since 1981. The 1950’s saw homeruns fly out of the Polo Grounds, Ebbets Field and Crosley Field, giving the NL a league HR factor of 1.16 in 1956 and 1.14 in 1957. The NL’s lowest HR factors were 0.92 from 1965-1968 and 0.93 from 1979-1981. The NL’s HR factor has been above 1.00 since 1994, recently peaking at 1.05 or 1.06 from 2001-2004. Although individual parks vary plus or minus up to 10% in BABIP, the league factors have always been between 0.99 and 1.03.
The teams which had the highest HR factors were the 1956-57 Giants at 1.37 and the 1961 Angels at 1.36. The teams penalized the most with a low HR factor were the Royals of 1966-67 at 0.67 and the 1965 Astros at 0.68. After Kansas City as the lowest two spots, Houston owns the next 11 lowest seasons, playing their home games in the Astrodome. If you need a double, play for the Red Sox, which has 34 of the 36 highest seasons, rating between 1.12 and 1.14. For percentage of hits that go for doubles or triples, the Dodgers hold 22 of the 24 lowest factors, from 0.89 to 0.90.