search

trynmaps / metrics-mvp

Prototype of public transit data visualization system
https://muni.opentransit.city
MIT License
30 stars 34 forks source link

Refine scoring of on-time percentage #541

Open exxonvaldez opened 4 years ago

exxonvaldez commented 4 years ago

@exxonvaldez wrote: We can create a new issue for refining how OTP is scored. The TCQSM (http://www.trb.org/Main/Blurbs/169437.aspx) page 5-30, has rough tiers for OTP:

Screen Shot 2020-02-05 at 4 21 56 PM

What's interesting is this metric is far more context sensitive than the other metrics (that is, the scale may need to be tuned to the agency, whereas the others may not). The SF City Charter calls for an OTP of 85%. In practice the system stat is in the 50s. The TCQSM says 70% may be the best that's possible where congestion is a factor, which would be true for SF. The TCQSM also talks about "headway adherence" for frequent lines. It's a pretty complicated stat, so I'll just show the table of tiers:

Screen Shot 2020-02-05 at 4 29 01 PM

Originally posted by @exxonvaldez in https://github.com/trynmaps/metrics-mvp/pull/528#issuecomment-582680394

For a first pass, we can set a per agency threshold that represents a "100 score" and a threshold for "0 score", and pick reasonable defaults. For defaults and muni, we could go with 0% to 70% or 20% to 70%.

For a second pass, we should look at factoring in bunching and gapping for higher frequency lines. This could get a little complicated, where the score is a blend of OTP score and bunching/gapping score, depending on frequency.

hathix commented 4 years ago

Also, I think Tal from SFMTA said that OTP doesn't matter to riders if the bus comes every <15min, since riders don't really check schedules for those. Perhaps we could elevate headways / waiting time in the UI for those routes.