To determine the best pitchers we could look who had the lowest ERAs. But we might want to adjust those for the time period and park effects. In some years it might be easier to have a low ERA than other years. In 1968 the AL ERA was 2.98. In 1996 it was 5.00. The same can be said for parks. From 2002-04, 36% more runs were scored at Coors Field in Denver than the average NL park while 16% fewer were scored at Dodger Stadium. We might also take into account innings pitched since a pitcher can have a low ERA but if he pitches infrequently. One more factor is the role of fielders. Pitchers who have great fielders behind them will allow fewer hits and therefore have lower ERAs. I used all of these factors in this analysis.
Last week I wrote an article called "The Best Fielding Independent Pitching Seasons From 1920-2005." There I explain the methods I used to incorporate the factors that I mentioned above. The data I used includes all pitching seasons with 150+ IP (so any data from a pitcher's season with under 150 IP did not figure in-this means I'm pretty much only looking at starting pitchers). Pitchers were compared to the league averages in strikeouts, walks and HRs allowed using how many standard deviations above or below the mean each pitcher was. Once all of those factors were taken into account, each pitcher was placed in the context of a league with approximately a 3.70 ERA.
Here are the lowest 25 ERAs with a minimum of 1,500 IP:
For a minimum of 3,000 IP the lowest 25 are:
As I mentioned above, a low ERA is great but might not be valuable if a guy's IP are low. So I predicted how many wins each pitcher would get based on their predictd ERA assuming their team scored an average number of runs. For this I used what Bill James calls the "Pythagorean Formula." It says that winning percentage can be approximated by runs scored squared divided by runs scored squared plus runs allowed squared. Each pitcher's number of games was found by dividing their IP by 9. Then the pitcher's games won was predicted by multiplying the Pythagorean percentantage times the number of games. Subtracted from that was the number of games an average pitcher would have won (.5 times games) to get games won above average.
For example, Roger Clemens had 4,332.67 IP. That works out to 481.4 games. He had a predicted ERA of 2.45. If his team scored 3.696 runs per game the Pythagorean .695. That would give him 334.46 wins while the average pitcher would have had 240.7. So Clemens was 93.76 wins above average. Here are the top 25 in wins abvove average.
I also looked at wins above replacement level, assuming that a replacement level pitcher would have a .400 winning percentage (I think Dvd Avins may have been the first researcher to look at pitchers' value this way). Looking at Clemens again, in 481.4 games, the replacement level pitcher would win 192.56 games (481.4*.400). Clemens' 334.6 is about 142 wins above that. So here are the top 25:
I also tried a .333 pct for the replacement level pitcher. Here are the top 25:
The Sean Lahman Database