That's the only way you can create models of something as complex as society. It's way more complex and chaotic than weather and still economists think they can predict months and years into the future.
There are also very very granular models.
Do you seriously mean that smaller economies can be modelled several months into the future with accuracy and be correct more than 50% of the time?
("50%" might not always be the right way to look at things, but I assume you mean something better than random or previous.)
"To assess the quality of the forecasts, they are compared with the actual GDP figures. Since GDP is normally revised several times, it is necessary to decide which figure to take as the outcome. Following the literature, we use the first available estimate for real GDP growth. In our case, this is the annual average calculated by the State Secretariat for Economic Affairs (seco) in March of each year on the basis of its quarterly estimates.2"
Physics isn't "complete nonsense" just because we don't know the exact mass of a proton but rather have some small uncertainty there.
Same with inflation. It's hard to measure correctly in the first place and becomes even more useless as it is defined differently at every occasion or when it does not match whatever one tries to achieve.