I have a confession to make.
When I first fell headlong into baseball analysis, I kind of took K/BB as my pet statistic. I loved how it was essentially "luckless": the only things the pitcher couldn't control was the discipline of the batter and the strike zone of the umpire, the two things he can only understand (but never master). Not only was it good at explaining the two fundamental pitching statistics, it did it in a way that aped the GPA system in the American school system. If you had a 4.00 GPA, you were an excellent student: a 4.0 K/BB meant you were an excellent pitcher.
After a while, though, I began to have my doubts. Is a pitcher that strikes out 9 per 9 innings and walks 3 in the same span really only half as good as the guy that strikes out 6 and walks 1? If we assume a normal BABIP (lets say .300 exactly), these two pitchers profile thusly (over a 27 out span):
Pitcher A: 9 strikeouts, 18 regular outs, 7.7 hits, 3 walks, 10.7 baserunners
Pitcher B: 6 strikeouts, 21 regular outs. 9 hits, 1 walk, 10 baserunners
Pitcher B has a 6 K/BB, but allows nearly as many baserunners as Pitcher A does. Not only do they allow baserunners at close enough a rate as makes no difference, Pitcher B actually allows more active baserunners. I'd expect Pitcher B to allow more runs per 27 outs than Pitcher A does.
This sort of trickery is what allows Joe Blanton to accrue such gaudy K/BB totals while being a mediocre pitcher at best.
|2005||Athletics||5.19||3.00||1.73||1.03||13.9 %||8.0 %||.233||1.22||.248||75.3 %||83||105||4.43|
|2006||Athletics||4.96||2.69||1.84||0.79||12.5 %||6.8 %||.304||1.54||.335||68.9 %||108||94||4.16|
|2007||Athletics||5.48||1.57||3.50||0.63||14.7 %||4.2 %||.265||1.22||.299||68.0 %||90||81||3.50|
|2008||2 Teams||5.05||3.01||1.68||1.00||13.0 %||7.7 %||.269||1.40||.290||68.4 %||110||106||4.52|
|2009||Phillies||7.51||2.72||2.76||1.38||19.5 %||7.1 %||.257||1.32||.291||78.9 %||96||104||4.45|
|2010||Phillies (A)||9.00||0.00||2.00||0.00||33.3 %||0.0 %||.000||0.00||.000||100.0 %||1.20|
|2010||Phillies (AA)||5.63||2.25||2.50||2.25||14.3 %||5.7 %||.273||1.38||.269||73.2 %||5.95|
|2010||Phillies||6.87||2.20||3.12||1.38||17.5 %||5.6 %||.287||1.42||.321||69.1 %||120||108||4.34|
|2011||Phillies (A)||0.00||0.00||0.00||0.00||0.0 %||0.0 %||.000||0.00||.000||100.0 %||3.20|
|2011||Phillies||7.62||1.96||3.89||1.09||19.4 %||5.0 %||.306||1.48||.362||70.9 %||131||94||3.63|
|2012||2 Teams||7.82||1.60||4.88||1.37||20.6 %||4.2 %||.269||1.26||.310||67.9 %||121||101||3.91|
Clearly, K/BB isn't actually providing a meaningful statistic.
What matters in pitching is acrruing as many outs as you can in the fewest number of batters you can. All of the other variables are going to average themselves out, somehow.
If you take as a given DIPS theory, that a pitcher has no control over balls in play, it stands to reason that the same pitcher has little or no control over the distribution of the hits that take place 30% of the time. Therefore, the best way to reduce runs is to strikeout as many batters as possible while denying as many free passes as you can – the primary reason that strand rate is also essentially volatile. K/BB isn't a measure of this skill at all.
What is, however, is K-BB. The difference between strikeouts and walks is the difference between batsmen and batted balls. This is a much better statistic, because it describes an actual skill in reducing runs – coincidentally, the only skills he can actually control.
The only thing that's left is converting this count statistic into a rate statistic. The best way to do that is to take this difference in batsmen/batted balls and set it to actual batters faced. Thus, (K-BB)/PA is the statistic that best addresses this problem.
Let's go back to Pitcher A and Pitcher B.
Pitcher A: 6/28.7 = 20.9%
Pitcher B: 5/31 = 16.1%
This gives us a clearer view of the value of these two pitchers. Pitcher B has the far superior K/BB, but he opens himself up to more batted balls, and as such is worse (albeit only slightly worse).
The one concern with this statistic is that it doesn't take HR% into account. First, that's also a rebuke to the old stat (K/BB). More importantly, though, HR% is actually a fairly erroneous statistic in itself. It's open to all sort of disputes: is a HR in Coors Field the same as a HR in Petco Park? HR% is a "pitcher's skill" to some effect (home runs imply hard contact), but it's much, much less useful than K/BB.
If you have any questions about either stat, feel free to comment.