It's all our best guess; however our best guesses matches up very well with most of our predictions. This implies that our guess was pretty good.

For example, we predict that the sun and pretty much all others main sequence stars have an abundance of hydrogen fusing into helium. We've never grabbed a bit of the Sun to analyze directly (let alone any other star), but we still believe them to be made primarily of hydrogen. Our physics and cosmological theories make predictions like the spectral nulls, orbital rotations, etc.

That said their are several discrepancies, like differences in the calculated value of Hubble's constant depending on how it is generated. These failures of our predictions are actually some of the most exciting results for scientists since it indicates that there is likely some new physics there that we don't fully understand.