Better Than WHIP?

To preface this I must give the credit to Hank Sager for even suggesting such an idea, I'm simply the data and writer guy, he's the brain of the invention. Anyhow, recently he posed the idea of doing a statistic, similar to WHIP, only Total Bases allowed/innings pitched, giving you a number of bases allowed per inning, and seeing if it correlated better to ERA than WHIP did.

Above is the BPI/WHIP/ERA of every pitcher last season with at least 25 innings pitched. I have a few observations I'd like to note:

1. Note how WHIP doesn't include HBP.
   
2. BPI implements a certain power aspect, so if a pitcher is only giving up singles he's less likely to give up a run than a pitcher who gives up a two doubles. WHIP doesn't factor in that information.
   
3. In some cases the BPI doesn't seem to match with the ERA, the theory we came up with would be that some pitchers are either simply unlucky or, as Hank himself would say, the pitchers meltdown when runners get on, here's an from our initial research I'd like to provide:

Jay Witasick appears extremely unlucky while Juan Salas seems relatively lucky, to try and guess at what exactly happened here's a look at some important stats.

Stat        Salas    Witasick

BABIP        .259    .299

BABIPwRon    .219    .250

WHIPwRon*    0.83    1.21

BPIwRon    1.13    1.53    

Note that the innings for WHIP and BPI with runners on are estimations based on the amount of games listed with that scenario, essentially each G=1 IP. By this theory Witasick seemingly remains the same while Salas kicks it up a notch.

I'm going to begin using BPI and monitor it during the year, but what do you guys think?

Hat tip to Cork Gaines for helping out with the graphs and R2 numbers for different situations. Here's a look at the basic correlations between BPI and WHIP with ERA:

Okay and rather than upload the other four charts, although I am grateful, I'll just list the R2s here for each innings plateau:

INN...BPI R2...WHIP R2
50...0.727...0.702
100...0.698...0.638

As Gaines pointed out, it seems like BPI helps predict a starter's ERA better than a relievers, which is going to be something I'll release as a document soon, but I'd like to gather the past few years' data and see what type of consistency exists, if any.

Note: I've been informed that Ron Shandler has something very similar in name at least. Apologies to him, I had no idea.