In fact, if you test 20 hypothesis you don’t need to even bother collecting data to get a p-value of 0.05, the number of hypotheses virtually guarantees it.
I didn’t edit any of the data. I didn’t throw out any data points as “outliers”. I used the same technique (linear regression) with the same data set used by the researchers, and I reached the opposite conclusion with a p-value that’s 5 times more significant. It took me less than two hours.

That’s the power of multiplicity – it can “prove” any hypothesis you throw at it, and the opposite of that hypothesis as well.