Of the 3 CA cities 'most similar', two (S. Jose and Glendale==So. Cal) are where I spent (nearby) most-all formative years..
Considering SF/San Mateo/ -vs-'San Jose' and 'Glendale'/Pasadena -vs-San Diego,
that's zeroing-in, with radii of:

1) 20 miles and 40 miles and
2) 20 miles and maybe 45 miles
==On average: "30" and "33" Mi. Radius [!!] on a 3000 mi wide 'place' of R. ~ 1500 Mi. 1-dimensionally

Isn't that ~ 0.2% accuracy AND precision?
(Using merely 'width'/horiz. above, whereas most errors noted were map-'height'/vertical, thus less-precise in just that dimension.
Max-error could then be called 0.4% or averaged to 0.3% as: silly to do map vectors too! unless we knew the granularity of their db, I wot.)

See? Scott: Metadata CAN Finger-->You ... to a fare-thee-Well (or-Ill.)
Convinced yet? Even unto q.e.d.??

;-0

Heh: least-similar were Newark/Paterson NJ, Philly and Toledo; could I understand er, crazy?
(Hope someone will give us a clue re. the ongoing in NJ.) My wish: Not-typical-Murican 2013.
Please, Not THAT.. 'justice'.



Ed:
PS: were they cheating re IP addy: they would have picked SF not San Jose--I'm 50 mi N. of SF (and maybe Santa Rosa isn't 'much' on their data-set?)
Just to be punctilious, that is. And wtf is math for, if ya can't Use it to crush all opponents?