Once I mapped out the points, I went ahead and connected them using a simple heuristic - that one set of lines connect the lucky cities to each other, a second set connects the unlucky cities to each other, and the lines do not cross. There are few outliers - Providence, RI is lucky but just far enough west that you can't draw a line to it from New York City without passing through unlucky Bridgeport, CT. Unlucky Stockton, CA is surrounded by lucky cities on all sides. And while according to the heuristic Denver, CO connects into the West Coast grid, it is a considerable distance from the other cities in the grid and might be an outlier instead. Also, the cities in Alaska and Hawaii are really too far away to be included in either of the continental grids.
Looking at the map, a pattern does become clear. Most of the West Coast is lucky. In the east a lucky swath runs from the area around Baltimore and Philadelphia between the Ohio River and the Great Lakes. Once it reaches Kansas, the swath turns due south until it passes through Texas and Louisiana, reaching Houston and Baton Rouge. My original working hypothesis was that magical forces related to particular elements should follow the physical movement of the element in question. So Earth would move along fault lines, Water would move along rivers, and Air would move along with atmospheric currents. But with the way that the map looks, this analysis seems too simplistic. The lucky eastern swath sort of follows the Ohio and Mississippi rivers, but it runs north of the Ohio and west of the Mississippi, and cities on the river such as Memphis, TN and Little Rock, Ak, are distinctly unlucky.
Following the recent earthquake in Oklahoma I was looking up some information on the New Madrid Fault system and came across the following diagram showing the relative risk of earthquakes across the country. If you superimpose the two maps, you can see that there are some similarities.
If we adopt the operating assumption that the fault lines on the West Coast are somehow lucky, and the fault lines around the New Madrid formation and New England are unlucky, you can put together a model fitting much of the data in which luck follows major rivers but then is increased by lucky faults and decreased by unlucky ones.
In fact, there is a major geological difference between the West Coast fault lines lines like the San Andreas that represents the boundary between the Pacific and North American plates and those around the New Madrid fault which lie along the Midcontinent Rift System. On the West Coast, the fault lines are formed by tectonic plates coming together, whereas those on the rift are formed by the North American plate pulling apart.
This suggests a further rule that may help to explain the unlucky cities of the Upper Midwest. Minneapolis and Saint Paul are on the Mississippi, but Minneapolis is neutral and Saint Paul is slightly unlucky. The same is true over in Montana, where the Missouri River starts, and through the Dakotas. As the headwaters of the Mississippi and Missouri rivers, the Upper Midwest and Upper West are areas where the water is moving away rather than coming together. It seems like once enough water has come together via the various tributaries areas around the major rivers start becoming lucky except when interfered with by the geology around the New Madrid area.
Still, none of that really explains Florida, which is by far the unluckiest state. It has little in the way of major rivers and is not geologically active. I have yet to look at atmospheric patterns, so there could be a clue there. The eastern swath sort of tends to follow the Jet Stream, so perhaps that could be a factor in the luckiness of the cities along it as well, and Florida's weather is rather odd since it sits between "hurricane alley" in the Caribbean and the Gulf of Mexico. But that's something I'm still looking into. Atmospheric changes are harder to pin down because they vary a lot by season.
What I would really like to see is the full breakdown of data, not the combined scores. It seems fairly reasonable to assume that different classes of lucky or unlucky events might be related to different elements in terms of the magical forces involved and the general spheres of activity in which they occur. The magazine article lists its sources, but if I want to build maps broken out by type of luck I'm probably going to have to put them together myself. So that remains a question for another day, and if anyone would like to look into it before I get to it I would love to see the results.