Comment by crazygringo | Hacker News Reader

Comment by crazygringo | Hacker News Reader

As a New Yorker...

...I don't understand the value crowdsourcing provides, for transit.

Official real-time subway data is already provided by the MTA in all sorts of forms (apps, API's, etc.). I don't know what there is to crowdsource -- you can tell just by analyzing the data directly whether trains are stuck or not, and for how long. And if you're stuck in a train car, it's very rare you have any knowledge of how long the delay is going to be anyways -- conductor announcements are almost always just "we hope to be moving soon". Same if you're on the platform -- just a generic "trains are running delayed". And for big delays (a train being shut down between stations for half an hour, etc.) the MTA posts real-time notices explaining. So what's to crowdsource?

(Not to mention that if your train is stuck between stations and you wanted to tell the app why, tough luck -- there's no cell or Wifi service between stations anyways when you're underground, so you can't.)

There's always room to improve on transit apps, and crowdsourcing is hugely valuable for car traffic, but crowdsourcing+transit makes no sense to me, at least in NYC.

The coverage and accuracy of the information provided by the MTA APIs is poor, in my experience. Additionally it doesn’t take into account crowding or any of the other quality-of-trip components (i.e. closed elevators, altered service, etc.).

What specifically is poor?

I've never used the API's -- only official MTA apps which presumably draw on the same data source -- but it covers all trains, and the train times have always seemed entirely accurate to me, if interpreted as "this is the minimum time the next train could take to get to you, from its current distance".

The apps certainly do take account of altered service (same as Google Maps does). Elevator closings are reported too, no crowdsourcing needed [1]. These are all official, non-crowdsourced sources.

Crowding is really the only interesting thing in my opinion -- but although I don't know of any apps that do it, it seems relatively easy to deduce from train history and known average hourly station usage per day, which is predictable enough to give decent results. But I'm not sure how often this provides value though -- how often is a station/train so crowded that you'll choose a viable alternate means of transportation in advance? Not often, in my experience, which is probably why apps don't offer it.

[1] https://advisory.mtanyct.info/EEoutage/EEOutageReport.aspx?S...

Funny. I thought crowd sourcing was the killer app for transit delays. And if deployed outside of NYC, time between is 30-60 minutes, vs 3-5,15-20 in metro NYC.

When a train slows down, the phones on board the train can know in tens of milliseconds. In aggregate this can lead to much better predictions.

Reply on news.ycombinator.com