I know one thing.....I wouldn't use the GFS right now this far out, because it has had right-track bias of late. It has been better the last couple of days, but I just don't think Jeanne has been initialized well, and it may have that problem for a while.

As for worst model performance (sans NGM), the ETA was terrible with Ivan. It had it going into the Yucatan, west of N.O., and is still too far left. Even this close in, the only run that was close to verifying was the 6z. From here on out, it is a Nowcast anyway. In all fairness, this model wasn't designed to predict cat 4 storms. Something to remember though for future storms.