This morning's GFS forecast drops 700-mb temperatures to about -21C by 0000 UTC Wednesday.
My 20/20 rule is not really a rule, but derives from my impression there are only a small number of days where the 700-mb temperature at KSLC is lower than -20C or greater than 20C. Thus, anytime I see -20C or +20C air, it piques my interest and I know we're out there on the ends of the bell curve for temperature. Plus, 20/20 is easy to remember. On the otherhand, one should always verify such rules of thumb. Perhaps someone can run some sounding stats and we'll see if my intuition is any good.
Also at issue is if this GFS forecast will verify. Beyond just the usual issues related to the need to think probabalistically at this time scale, the source of air this cold is interior western Canada and in the past the GFS sometimes had a cold bias in these scenarios because the real-world topography produces more blocking and airmass transformation. The GFS has been upgraded recently to an effect grid spacing of ~27km, so perhaps it will do better in these scenarios.