Obviously, from a pitcher's standpoint giving up a hit is worse than giving up a walk is worse than getting an out. Thus, if it were possible to 'exchange' hits for walks, it would be a good thing. However, it is unlikely that pitching works that way; more likely, when a pitcher is on, his walk and hit rates both go down; when he's not on, they both go up. In other words, it is more likely an 'exchange' of an out for a walk.
I put 'exchange' in inverted commas because, of course, we can never know what would have happened (a hit or an out) to the batter had he not walked. So on some level, we can not know if such a proposed approach is possible in practice. Nonetheless, looking at some figures might lead to some insight.
First, let's look at ERA vs WHIP for all AL pitchers (except for DM) in 2008 with IP > 100 -- a sample we'll use as a control (data through 15/08/08):
The correlation is impressive: Pearson's r = 0.870 +/- 0.033, with a scatter (standard deviation) of 0.429 in ERA from the best linear fit. There are zero outliers of >3 standard deviations (stdv). However, DM (in the plot, the red star) falls 3.4 stdvs (a random probability of ~0.04%) below the fit. How can we understand this?
First, let's look at simply ERA versus hits per 9IP:
Also a very good correlation (r = 0.788 +/- 0.051, stdv = 0.535). Now look at ERA versus BB per 9 IP:
There is a detectable, but weaker, correlation (r = 0.311 +/- 0.122, stdv = 0.826), and DM is in an otherwise unpopulated locus. Thus, DM having a very atypical walk rate deviates him from the WHIP-ERA correlation, since walks have much less affect on ERA than do hits (obviously), and folding in BBs to the hits-data only improves the scatter by ~ 0.1 run.
So how do we interpret this? Is DM (uniquely?) able to 'exchange' hits for walks? What should we expect going forward:
1) DM can continue to give up hits and walks at his current rate, and the result will be:
(i) his ERA will remain low, as it is spot-on for his hit-rate and the BBs don't affect the ERA that much relative to hits allowed
(ii) his ERA going forward will be worse; his current ERA is a statistical fluke given his WHIP
2) his hit-rate is will increase; it is his low H/BB rate that is the fluke, and his ERA will rise accordingly
3) DM can reduce his walk rate without increasing his hit-rate, and he will then dominate the AL
* I have no idea whether this is DM's "approach" or just how it works out for him, but either way I think it's worth discussing. I will note, however, that he was not such an outlier in the 2007 whip-era data.
* As we all know, no matter the ERA, a high WHIP leads to shorter outings and increased use of/dependence on the bullpen, which is a bad thing. I mean to discuss these stats independent of this effect (although we of course shouldn't forget all about it!)
* There are better predictors of ERA (such as contextual ERA, etc.) than WHIP. I'm using it here as it is a simple, obvious stat; a comparable discussion using, eg., CERA, would be too dense.
Edited by dbn, 18 August 2008 - 01:01 PM.