Actually, they're not "predictions", they're "forecasts". The difference is crucial.
Economists aren't just putting a finger in the air and guessing some numbers. There are no crystal balls involved. Instead they're using formal economic models, putting appropriate numbers into the equations and publishing the results.
Putting slightly different numbers into those models produces slightly different results, which is how economists have been able to produce various forecasts for different types of Brexit.
The models are very well established, they're tried and tested. Which is why no economists have been able to publish a single peer-reviewed economic paper forecasting a positive economic outcome for Brexit.
Forming a consensus of economists is notoriously difficult, but Brexit is one of those rare cases where the evidence is undeniable.