Out-of-sample holdout testing as a remedy
To guard against spurious patterns, one should form a theory first, then test it out of sample, in different markets and environments, and tweak variables that should not matter; if the result survives changes to irrelevant variables, the finding is more likely real, but if it fails, the analyst is probably fitting noise.