Okay, so you take 3000 patients with a particular blood disorder and they're a mix of people who recovered and people who didn't. Next you divide them randomly into two groups of equal size -- one is the control group, and one is the intervention group. Now: You say a prayer for everyone in the intervention group.
With me so far? So what happened? It turns out that the group that were prayed for were more likely to recover quicker and less likely to die than the control group. So far so power-of-prayer.
Get this: By the time the prayer had been said and the study was made, the patients had either recovered or died at least 6 years earlier.
It's in the British Medical Journal (the paranormal special issue, I admit): Effects of remote, retroactive intercessory prayer on outcomes in patients with bloodstream infection. Remote. Retroactive. It all violates my world view and universal model in pretty fundamental ways, but it's still fucking wicked.