I was reading my favorite Mariners blog and they had this tidbit:

As it happens, OPS + SLG (x 100) estimates runs x 162 ... when you use the constant 162, which is the number of games in a year.

It so happens that when you add OBP and SLG, and multiply by 100, then 162 is the constant you divide by, to get probable runs in nine innings.
When you sort every team in the MLB by runs (http://sports.espn.go.com/mlb/stats/...atting&group=9) you'll see that the top teams have far more runs scored than their OPS. Then when you go a little lower, starting with Atlanta, where it's almost dead on, they're very close. Then as you continue to move down, the OPS is significantly larger than their runs scored.

Blah, as a whole it seems like it'd be pretty accurate, but it seems unless you know in the neighborhood of how many runs you'll score that OPS doesn't really tell you anything. And that kind of defeats the purpose.