Take ten smart kids and stick them under an oak tree and you'll have a better school than a 100 million dollar facility full of idiots. That's the real problem, and there's no solving it.
I've seen great teachers in horrible schools make a huge difference. The tough part though is that follow-up with kids is tough. You see a class make incredible improvements one year and then over the next four years have less than stellar teachers again. They slowly give back what thy've gained.
This is why the Harlem Children Zone is an example of the right step.
Here's a nice graphic from the NYTimes: http://www.nytimes.com/imagepages/2010/08/01/education/01sch...
The real problem is that we have an artificial line for proficiency. If you look at the graphics you can see that they can move that line once again, and again largely affect minority schools downward, even if they make substantial gains over the next few years.
It would be far better to have each school release mean/median/stdev or deciles of their scores. This will really give you a much better idea of the improvement in a school.
But anyways, back to HCZ. They've still shown great improvement, even if proficiency has taken a step back from 100% in some cases due to a change in the proficiency line.
One of them recently suggested that supplementing their school's non-curriculum with Sal Khan's online exercises (explained here: http://www.youtube.com/watch?v=wRf6XiEZ_Y8) might make for something close to an ideal education.
Of course, since I'm so humble, I may be wrong and ignored. :-)