There has been discussion in the comments sections of a couple of recent articles (here and here) about what values should be used as a baseline for comparing players' seasonal wins above replacement (WAR). To address that issue, I incorporated work done by JBrew with several other suggestions and I came with a an average HOFer and replacement level HOFer (minimum value for induction) for pitchers and hitters.
To get the values I used the following procedure:
I collected all the HOFers who played their entire careers in the retrosheet era (which are the seasons presented in Sean Smith's WAR database). I made an exception for two pitchers, Hoyt Wilhelm and Whitey Ford, because they played only one season before 1953 and there aren't as many pitchers (19) as hitters (32).
For each player, I ranked their seasonal WAR values from highest to lowest.
All the highest single season values were grouped together, then the second highest seasons were grouped together, and so on down the line.
I took the value that was at the middle (median) of the pack for each grouping to use as the average HoF career path. I took the values that were at the 20th percentile of each grouping to use as the "replacement level" Hall of Famer. Yes, this implies some HoF players don't deserve to be enshrined.
Once the WAR values went negative or there were no more seasons at the median level, the final value was set to 0 and no more values were calculated. The series values were then used to create two reference lines.
Here are the values in graph and tabular form:
Data tables are after the jump:
||Replacement Level Hitter||Median Hitter||Replacement Level Pitcher||Median Pitcher|
The following is a chart of players whose career WAR totals closely match each of the four defined WAR trajectories.
|Replacement Level Hitter||Willie Stargell||58.2|
|Average Hitter||Tony Gwynn||70.6|
|Replacement Level Pitcher||Whitey Ford||48.8|
|Average Pitcher||Don Drysdale||65.6|
Feel free to use these when creating WAR graphs so we have a general comparison between graphs, and let me know if there's any way to improve the methodology.