I figured out a relatively principled way for us to decide, when we try to solve an astronomical image
blind, whether or not our solution is the correct one. We want to be conservative (false positives lose us credibility) but we don't want to throw out the baby with the bath water. I figured out a criterion that will keep our false positive rate at the 10−6 level—to order-of-magnitude. Exact statistics are not possible here because the stars on the sky are not uniform (not a Poisson process), and the moments of the distribution are not known (they could be measured for the standards catalog, but they can't be known for arbitrary input images).
In other news, Lang got the blind system to solve this image today. Outrageous!