EDI in Context

October 25, 2022

Nearly six months ago now, I created the Eliot Deviation Index. While it certainly has gained quite a bit of popularity, I’ve always been wanting to write this post. I couldn’t write it immediately after the first post, because I needed experience with the Eliot Deviation Index (EDI from here on) first. Well how much experience? I didn’t know. Well, I think I have enough to write this now.

First, what has happened in the last six months? Six months ago, we had just about 300 stops in the database, just on the MBTA’s rapid transit network. We slowly, and then more rapidly, expanded to where we are now, at over 185,000 stops across many agencies around the entire United States (and Thunder Bay, Ontario, is currently the only non-US region with support). A number of friends (and others) have chipped in with suggestions for routes to calculate and even help calculate some routes in supported agencies. Some of the many agencies would not have been added if it weren’t for some suggestions. The history of this is a whole separate thing at this point.

Like a lot of statistics and measurements, the EDI is no exception in the fact that it should be looked at in context with other factors included. There is no scale in what is a good route, okay route, bad route, or just plain terrible route based on a route’s EDI. More important is where a route serves. Public transit is here to connect people with the places that they want to go. You see that quite differently in different environments. Cities tend to have bus routes along major corridors serving one particular street, and can sometimes supplement any other forms of transit, most commonly connecting to train lines. However, it is quite the opposite in suburban areas. In the suburbs, routes focus more on connecting important destinations together, to make the most out of the resources that they have. Suburban bus routes are known for taking more turns to connect various shopping centers, downtowns, and residential areas with as little transfers as possible. Both types of bus routes can be equally as good and well thought out, but the EDI of each type of route will be quite different.

Let’s look at some examples of well thought out bus routes in the city and in the suburbs. Chicago’s bus network is mostly in a grid pattern. Routes run east-west or north-south along one street and one street only. The 74 is on Fullerton, the 77 on Belmont, the 8 on Halsted, etc. Chicago’s streets are also laid out in a grid pattern so that no two east-west streets will ever intersect. Exceptions do exist in the case of the diagonal streets, but that’s not really related to this. Not every bus stays on one street either. Some routes have turns in case the road ends or has a gap, or to serve an important connection point. I’ll use the 74 as an example. The 74 along Fullerton runs between Grand and Nordica (the change in street name is irrelevant, slight diagonal on Grand) and ends in a small loop along Lincoln, Webster, and Halsted just east of the Red and Brown Lines at Fullerton. Overall, it is a very well thought out route. It compliments the north-south lines it intersects and also provides service along the Fullerton Ave corridor, linking many homes and businesses to the Red, Blue, and Brown Lines. As the route travels along a mostly straight path, the 74 has an EDI of 1.03. This low EDI shows that the 74 takes a very direct path between the western and eastern ends of the route, which it does.

For my suburban example, let’s take the A, operated by the WRTA surrounding the medium-sized city of Worcester, Massachusetts. The A is a very suburban route, operating wholly within the towns of Northbridge and Grafton before continuing non-stop to its northern end in Millbury. While the A is not the most direct path between the Walmart in Northbridge and the mall in Millbury, the A serves many important areas along the way. The A serves the downtown area of Northbridge, another shopping plaza in Northbridge, as well as the Grafton village of Fisherville, with many homes passed along the way. Riders can wave down the bus at any location in either town and the bus will stop if it is safe to do so. Even with the A being such a well thought out route, the A has an EDI of 2.07. It travels twice as far as the most direct route between the endpoints (that happens to be using a highway), but it serves everywhere else in between. Northbridge doesn’t have the demand and the WRTA doesn’t have the resources to support a system of more direct bus routes, so this works well for both.

So we know that the EDI is directly related to the path the route takes. High EDIs can be good, but did you know that a low EDI doesn’t necessarily mean that the route is also good? Take a visit to Newport, Rhode Island, a city of 25,000 and a popular summer tourism destination with a small subset of the RIPTA network centered around it. Some of these routes are local, some are longer-distance connecting Newport to the rest of Rhode Island. One route in particular will stand out - the 63. While me and my friends will call it a “drunk snake”, its EDI reflects that, with the route having a 2.18. This was considered high when it was originally calculated, it’s on par with the WRTA’s A as discussed earlier. This route is bad. While some connections can be made easier with this route, the 63 mainly serves areas that are already served by the other Newport routes. Maybe we cut the 63 and use the resources to increase service along the other routes in the area? Just an idea though. Some reroutes may have to be done on the other routes.

Good routes with high EDIs have purpose. On the complete opposite end is Davis, California, a college town centered around UC Davis with a student-operated bus system, Unitrans. The T is a once-a-day school trip, connecting together various bus routes to give local students a one-seat ride from school to their destinations. As a result, the T has an EDI of 7.8. SEVEN. NEARLY EIGHT. Here’s the thing - it has a purpose that warrants the 7.8, unlike RIPTA’s 63 mentioned earlier. Back in my home of Chicago, we have a network of night buses along some corridors that warrant overnight service. One of these corridors is one that doesn't serve as a one-seat ride during the day. Meet the N5. The N5 is an amalgamation of multiple bus routes along the South Shore of Chicago. The route provides night service to the neighborhood, and I assume that the individual bus routes in this area don’t get enough individual demand to see night service, so the CTA just created a new one. The N5 has an EDI of 3.6, but serves its purpose well. It connects the neighborhood to the Red Line at each end. A more extreme example can be found in San Francisco, Muni’s 91, with an EDI of 14.58. This route takes a much more circular routing than the N5, but has the same purpose.

I couldn’t write this post without bringing up an infamous route, though. New York City’s very own Q38. Like the N5 and the 91 mentioned in the last paragraph, the Q38 is also nearly a complete circle and without any reason for someone to ride from end to end. With an EDI of 24.34 - YES THAT SAYS TWENTY-FOUR - the Q38 is the most deviatory route we have in the database so far. This route operates all day, but not during late nights. This route is so pointless in its entirety that even the MTA thinks it's ridiculous, and plans to truncate half of it in the Queens Bus Network Redesign.

That’s only a small handful of all the routes with a calculated Eliot Deviation Index. As of the time I write this, there are 467 (and counting) routes in the database, so definitely go check that out for some more interesting cases that I don’t know the story behind. Don’t worry, the EDI will be here for many more months to come.

Posted: Oct 24, 2022 23:54


Want to submit a comment? Find out how here.