There's nothing strange about them not knowing. It depends entirely on how certain, essentially chaotic, atmospheric circulations play out. There are long term trends based on
ENSO you can follow, but that does not account for
NAO,
AO, or others.
Even longer term is the
PDO and
AMO, which take ~60 years to cycle back and forth and also clearly dominate weather patterns but happen so infrequently that we don't fully understand their effects. Would take a life time of data for a single sample.
Fall, clearly October is cold but November should bounce back warmer. It's a good bet this winter will not be a torch in the US like last winter, and some computer models agree, but nothing is for certain. They do their best with what they have.