The answer is it's a tie! ECMWF gives me a general idea of what to expect the furthest in advance. On the day of activity, I used to rely on NAM for a more precise, localized prediction, but lately HRRR often seems to be the winner of the two. I often compare all three to look for consensus, much as your article described. GFS is typically very inaccurate for me and I haven't figured out how to incorporate ICON yet. I live in Florida, USA. Windy is the best, thank you!