by Outta Leftfield » Thu May 31, 2012 5:15 pm
My sense is that in order of accuracy the three measures are, from best to worst, NERP, then BR, then RC/27. Each has its uses, however.
The main advantage of RC/27 is that it's a quick and easy, do-it-yourself formula. It's pretty good accurate for such a simple stat, and its easy to understand, since RC/27 means the runs created using 27 outs, so in effect it puts a runs-per-game value on a player. But--as Bill James, it's inventor, has admitted--in its simple version it tends to exaggerate offensive values at the extreme high end of OBP and SLG, and that high end is where a lot of ATG6 happens, so it's not entirely reliable in this environment. Great as Babe Ruth is, he's not going to hit quite as well as RC/27 says he will. Plus it leaves out things like steals, DP, and other factors that might be important if they represent a real strength or weakness of the player.
Also, I think RC/27 tends to be overly kind to pitchers who walk a lot of people but don't give up many hits. Ryne Duren (.92M) is not really as good as he looks from the standpoint of RC/27.
So, I would lean toward the more sophisticated measures of BR or NERP. I think I understand the mechanics of BR a little better, but as Adrian says, NERP can be used with DeanTSC's defensive ratings to get a sense of the overall value of a player.
All 3 have their uses, but I would put the least trust in RC/27, especially when evaluating superstar players or players with special skills or weaknesses like SB or DP.