I think you’re missing my point: we’re only able to predict large numbers of coin flips because we have an accurate model.

We don’t have an accurate model for weather, so we can’t predict it well.

I don’t see a reason to assume our model for climate is accurate, either.