Small sample size, but RRFS was pretty close in calling yesterday (when checking the day before). There also seemed to be pretty close agreement on all 3 of the more reliable models (NAM, HRRR, and RRFS).

Would be very interested to check out your historical forecast data. In compiling that, were you always recording the forecast data a specific amount of time before the “actual” (for example, did you record the forecast the night before a flying day)? That would definitely factor in.

Purely anecdotal evidence, but it seems like certain models are “better” for certain sites. I’m sure the weather experts here will light me up for this, but my personal opinion for Marshall seems like that when observing in the evening the following day’s conditions, NAM seems to have been close-ish most of the time on top of lift, and tends to underestimate on wind speed (in a few weeks, a moot point since it’s going away); HRRR tends to read both high and low on top of lift (inconsistent) and slightly overestimates wind speed. GFS seems to only be good at ballpark guesstimations many days out on general wind direction (general “will it be flyable” kinda stuff), but not very useful. ICON and RAP I don’t even bother with, as they don’t seem very useful for the site either.

I’ve seen others say this about other sites as well, and I tend to agree, and that’s to take NAM and HRRR and take the worse reading of both (so higher wind speed, and lower TOL), plus hope for agreement with them all (agreement being a better signal of accuracy). I heard that RRFS is trending towards “shooting for the middle” between the 2 so far, but time will tell on that.

Also seems that anytime a significant weather event is nearby (like hurricanes, etc), that messes them all up and you have to take that into account.

Ok – now everyone can tell me how wrong I am on that. :)