Baseball Judgments

Baseball Judgments |

Stats like 'range factor' biased by lurking variables? Perhaps not


By Roger Weber


Earned run average is the best pitching statistic. At least a lot of baseball fans and analysts think so. But ERA alone does not give the full picture of a pitcher because of a variable we often need to think about, sample size. Earned run average does not depend on the number of innings pitched or games appeared in. It is an average and like all averages the "n" or number of trials generally makes it closer to the actual value but does not change it significantly.


In addition to ERA, many analysts use a statistic like "innings pitched" to account for the sample size variability, a sort of "lurking" variable that affects this valuable average. Range factor, a measure of putouts and assists in a set unit of time, also comes under siege and criticism for being affected by "lurking" variables.


It is true that some players get more total chances fielding a position. This accounts for big differences between position range factors, but even between players of a single position some players may play on teams with more "ground ball pitchers" or a bigger playing field or weaker teammates, all of which can affect range factor. A selfish player who takes a lot of fly balls in situations where either fielder could take it has a higher range factor. Range factor also says little directly about the routes a player takes to the ball. But like ERA it tells a player's ability and performance indirectly and fairly accurately. The question is about how much these lurking variables affect the range factors.


A starting pitcher with a ground ball/fly ball ratio less or greater than normal can affect a fielder's range factor, but normally a starting pitcher doesn't pitch more than about 13-15% of his team's defensive innings. Still this can affect range factor a few decimal places, which can make a difference in a ranking of players although the affect is not terribly large.


But these claims that different factors affect range factor and thus make it a less valuable statistic do not really show up heavily in the stats. In 1985, Bill James calculated a sort of equivalent to victory-important offensive stats in his adjusted range factor figures. James' adjusted range factors only usually differ from the actual values by an average of about .3. And it rarely changes them more than half a point. And while this is, on average, nearly 10 percentile points, a large number considering a player said to be in the fiftieth percentile could be as low as the 4fortieth or as high as the sixtieth, the effect is felt by all players and is usually not as pronounced as it can appear because with rankings there is space between each value. The actual hierarchy of players measured defensively by range factor rarely changes much at all with adjusted numbers.


While James' adjustments are interesting and scientific they are also not readily available. Normal calculations of range factor are mostly accurate. Most teams have both ground ball and fly ball pitchers so the effect of these "lurking variables" is not nearly as great as it may sometimes seem. It is certainly not worth eliminating range factor from consideration of players, especially since it is a so much more accurate and valuable fielding statistic than most other basic defensive stats available.


Enter supporting content here