Throw your darts.
There isn’t enough initialization data to make such predictions. And where we have good data coverage, it’s accuracy (inherent error) out to month long timescales that bites.
We’d all be much better off killing this aspect of the science.
Do it for fun; but let’s stop there.