Very cool, thanks.

> I didn't expect to get so much data; need new disks drives

Anyone know how much data they'd get in 7 days, and what models they'll run on it to search for planet 9?

It isn't terribly complicated. They take images of a particular patch of sky on two different days, then xor them against each other to detect differences. Those differences are a moving object, something associated with our solar system rather than the (relatively) non-moving stars in the background. That gives you a detection and an angular speed, from which you can make assumptions about distance (stuff in orbit further away moves slower, really far away and it can seem to move in reverse).

Then you wait six months for earth to be on the other side of its orbit and take some more pics based on your previous guesses about the orbit. That lets you triangulate for a better distance estimate. Combine all the estimates and you can match it to a reasonable orbit. Do that over and over and you get an increasingly accurate estimate of the true orbit. And from distance+brightness+composition you get a size/mass.

Thanks! I was wondering if the devil is in the details--if the datasets are so large (petabytes+?) that there are millions of these "differences", most of which are false positives from factors like atmospheric noise or equipment issues, and what the computing challenges were sifting through it all...Or perhaps as you hint it is relatively straightforward after all.

With two images there are lots of false positives. But if you have three or more you can look for stuff moving in strait lines. That can be easily automated. Most near-earth object detection (bright, fast objects) is now totally automated. Hunting for planet X is about finding a little smudge of pixels moving a pixel or two to the right.