Another look at this: from a really simple analysis I did for fun a year or two ago, here are the r-squared values for a variety of different team offensive statistics when regressed onto team runs scored.  Higher r-squared means that there's a tighter relationship between those variables and runs scored--in other words, they're probably a better measure of offensive production.

OPS 0.91
SLG 0.83
OBP 0.81
AVG 0.71
Hit# 0.68
HR# 0.52
BB# 0.35
K# 0.03
(SB-CS) 0.004
SB# 0.0005

These admittedly simplistic data indicate that OPS (on-base plus slugging) is the best single measure of offensive production among those I considered (if you include [1.8*OBP+SLG], you get an even better fit, though not by much--0.92).  SLG and OBP were roughly equal overall, though they tell you slightly different information, which is why OPS is better.  Batting average alone explains ~10% less variation in runs scored than either of OBP or SLG.  It's not worthless, just not as good at predicting runs scored as OBP or SLG.

Interestingly, K's and SB's have almost no effect on runs scored.  Some of the best hitters strike out a lot, but so do some of the worst hitters, which makes K's a poor predictor of production.  The issue with SB's, of course, is that teams that don't score a lot of runs tend to run a lot out of necessity.  But some teams that score a lot of runs and make good use of steals (Reds are an example), so again, they're not a good predictor of team runs scored.
-j