On: Surges of cosmic radiation from space directly lin...

[klipt]

It's fine to form a hypothesis from old data ... but it's not a theory yet until that hypothesis is challenged by (and isn't falsified by) future data :-)

[EA-3167 → klipt]

This is doubly true when we're talking about predicting earthquakes, a practice that has had no successes and a LOT of notable failures.

[TimTheTinker → klipt]

Does old data discovered in the future count?

[ithkuil → TimTheTinker]

Yes it does. Because you didn't use it to overfit your model.

Same goes for data that you had in the last but decided to ignore while building the model.

In both cases there is a legitimate risk of cheating.

The only data that you cannot cheat about is future data.