Ballpark Effects
Ballpark effects matter - they greatly affect a player's statistics. Ballpark adjustments attempt to normalize a player's statistics in order to remove the effect of a player's ballpark. For instance, since 1982, players at Fenway Park have batted about 20 points higher than when they bat at other ballparks; because Red Sox players play roughly half there games at Fenway, one might imagine that Fenway Park boosts the batting average of the typical Red Sox player by 10 points.
One measure is called the Park Index, which is an estimate of the amount by which a given statistic is inflated by the home ballpark. For example, take a stat like runs - the Park Run Index is simply the ratio of runs at home to runs on the road. Importantly, the Park Index looks at runs scored by both the home and visiting teams, since home teams usually hit better at home than they do on the road.
Example: Coors Field in 1996 registered the highest run index of all time. The Rockies scored 658 runs at home, while opponents scored 559 runs at Coors. On the road, the Rockies scored 303 runs and gave up 405. So the Run Index is:
Coors RI = (658+559)/(303+405) = 1.72
The ratio is usually stated as a percentage rather than as a decimal - thus, we say that Coors Field had a run index of 172.
The same process can be used to develop an index for runs scored, hits, extra-base hits, home runs, or any other stat.
The next step is to convert this index to a Park Factor. Say a ballpark boosts run production by 72% (i.e. has a Park Index of 172) - intuitively, one might think that a team's run production has been exaggerated by 36% throughout the year - thus, a Park Factor of 136 can be used to adjust a team's statistics, meaning that we should divide Colorado's runs by 1.36 to get a true measure of their ability to score runs.
But two adjustments have to be made. First of all, teams may play more innings on the road than at home - remember, the home team doesn't always have to hit in the bottom of the ninth. If the number of innings played at home and on the road differ, than the Park Factor shouldn't be the simple average of the Park Index and 100, but a weighted average.
Second, simply creating a Park Factor of 136 for all the players on the Colorado Rockies of 1996 will actually overcorrect the problem. That's because all the other players in the National League got to play some portion of their road games in Coors; the Rockies players did not get this advantage, so their road statistics will be depressed relative to what other teams produced.
Pretend that Colorado had played some portion - 1 in 15, to be precise - of their road games at Coors. Then their road statistics would be a little higher, boosted by the same element that helped all other NL teams. Ideally, to remove the park effects from all stats, we should have each of the 16 teams in the NL play 1/16th of their games in each park.
Statistically, we can get there by making the Park Factor a weighted average of the Park Index (adjusted for the innings differential) and 100 (the league average).
Example: The Park Factor for Coors Field in 1996 is calculated in the following manner. There are 16 teams in the NL, so each team plays 1 in 15 road games at Coors.
So if PI is the Park Index:
PF = ((15 x PI) + (16 - PI))/ (15 x 2) = 1.336
Again, the Park Factor is usually stated as a percentage rather than as a decimal - thus, we say that Coors Field had a Park Factor of 133.6
wil.