A tale of two trains : The Indian Railways
Last week, I started collecting the running status of a few (<10) trains everyday. I wrote a blogpost last week about how I was collecting the data if you want to know more. Now, let's look at what I've collected so far.
(Open in the following images in a new page to take a better look at which stations are the most problematic and understand the general trend better)
Train 18029 - runs from Lokmanyatilak (Mumbai) to Shalimar. This train is mostly representative of what happens with the rest of the trains discussed below. There are stations enroute where the train makes up for lost time and then it loses any gains made. But, for the most part, I guess the delays are acceptable, given that they're within an hour of expected arrival time.
Train 12809 - runs from Mumbai CST to Howrah JN. This train was a little surprising because it's different compared to the rest of the lot. The train almost always makes up for delays in at the start of the route. There are a few places where there's a drastic reduction in delay but the gains are offset a few stations later (thrice)!
Train 12322 - runs from Mumbai CST to Howrah JN. This train displays two interesting trends. The first is that even though there are stations enroute where the train makes up for lost time (twice), it gets delayed again almost immediately. The second interesting trend is that beyond a certrain point enroute, the delay persists, and in 2/4 cases, the train can't make up for lost time.
Train 12622 - runs from New Delhi to Chennai Central. Can't complain about this train.
Train 12616 - runs from Delhi to Chennai Central. The interesting thing to note here is that there are points enroute where the train makes up for lost time - but, it gets delayed again almost immediately, negating any reduction in delay.
Train 12424 - runs from New Delhi to Dibrugarh Town via Guwahati. This train is just sad. At no point enroute does it show any prospect of making up lost time, if it's late.
Train 14056 - runs from Delhi to Dibrugarh via Guwahati. The running status of the train looks a little weird, doesn't it? After a certain point, the delays become very predictable instead of random. That is because I was asking for the running status of train at the wrong time - when the train was still in enroute. Of course, if I ask for the running status while a train in enroute, all I will get is estimated delay at future stations. Which is the reason behind long horizontal lines followed by dips.
Train 15910 - runs from Lalgarh JN to Dibrugarh via Guwahati. The running status of the above train shows the same behavior as the earlier one (14056) i.e asking for the running status while the train in still enroute WILL give me faulty estimates of delay beyond the current position of the train. And of course, the it's in the Indian Railways' best interests to estimate no delay instead of providing more accurate estimates.
That's all for now folks. I know, we didn't learn too much above why the delays are being caused or what routes lead to the most delay but we'll get there. I think. I'll try. I'll post the code I used to analyze the data and generate the plots tomorrow. If you can gleam anything more from the plots above or any other comments that you'd like to pass on to me, I'm all ears.
(Open in the following images in a new page to take a better look at which stations are the most problematic and understand the general trend better)
Train 18029 - runs from Lokmanyatilak (Mumbai) to Shalimar. This train is mostly representative of what happens with the rest of the trains discussed below. There are stations enroute where the train makes up for lost time and then it loses any gains made. But, for the most part, I guess the delays are acceptable, given that they're within an hour of expected arrival time.
Train 12809 - runs from Mumbai CST to Howrah JN. This train was a little surprising because it's different compared to the rest of the lot. The train almost always makes up for delays in at the start of the route. There are a few places where there's a drastic reduction in delay but the gains are offset a few stations later (thrice)!
Train 12322 - runs from Mumbai CST to Howrah JN. This train displays two interesting trends. The first is that even though there are stations enroute where the train makes up for lost time (twice), it gets delayed again almost immediately. The second interesting trend is that beyond a certrain point enroute, the delay persists, and in 2/4 cases, the train can't make up for lost time.
Train 12622 - runs from New Delhi to Chennai Central. Can't complain about this train.
Train 12616 - runs from Delhi to Chennai Central. The interesting thing to note here is that there are points enroute where the train makes up for lost time - but, it gets delayed again almost immediately, negating any reduction in delay.
Train 12424 - runs from New Delhi to Dibrugarh Town via Guwahati. This train is just sad. At no point enroute does it show any prospect of making up lost time, if it's late.
Train 14056 - runs from Delhi to Dibrugarh via Guwahati. The running status of the train looks a little weird, doesn't it? After a certain point, the delays become very predictable instead of random. That is because I was asking for the running status of train at the wrong time - when the train was still in enroute. Of course, if I ask for the running status while a train in enroute, all I will get is estimated delay at future stations. Which is the reason behind long horizontal lines followed by dips.
Train 15910 - runs from Lalgarh JN to Dibrugarh via Guwahati. The running status of the above train shows the same behavior as the earlier one (14056) i.e asking for the running status while the train in still enroute WILL give me faulty estimates of delay beyond the current position of the train. And of course, the it's in the Indian Railways' best interests to estimate no delay instead of providing more accurate estimates.
That's all for now folks. I know, we didn't learn too much above why the delays are being caused or what routes lead to the most delay but we'll get there. I think. I'll try. I'll post the code I used to analyze the data and generate the plots tomorrow. If you can gleam anything more from the plots above or any other comments that you'd like to pass on to me, I'm all ears.