An uncomprehensive statistical analysis of some of the effects of split ticketing

[johnsmith147]

Hello all,

I apologise if this has been done before, but I can’t find any other threads that attempt to do what I’ve been looking into. My aim has been to analyse all possible ticket combinations between UK stations, and see the effects that split ticketing has on the overall statistics. It’s been an interesting challenge, and I think the results should probably be published somewhere properly, if I can get around to it. A byproduct of it is a database of all optimal splits for walk-up fares from any station to another. I can make this publicly available (if I’m allowed) if people would like to use it.

I’ve done the analysis using the available static data from the National Rail Data Feeds, using Off-Peak and Anytime fares only and valid as of the 5th July. For this post I’ll just look at the cheapest standard-class fares between stations. I’ve done 1st class as well, but there are already a lot of numbers here! Distances quoted are as the crow flies, not along the railway.

Ideally, imminent fares reform under GBR will make all of this obsolete. I’m a firm believer that if a cheaper walk-up split ticket for a journey exists, it is essential that this can be bought as a through ticket. The "fair" fare is easy enough to calculate (I’ve done it!) and so there’s no excuse for not doing so officially... One caveat is that I’ve not used the official routeing guide at all here – it's just too complicated. But cheaper splits that contravene the guide shouldn’t exist, in an ideal world. Anyway, some stats:

Minimum/Maximum Fares Per Mile:
Min. price per mile, return: Wye-Ashurst New Forest, £21.20, 108.9 miles,19.5ppm
Max. price per mile, return: Forest Gate-Wanstead Park, £16.10, 0.139 miles, 11597.8ppm
Max. price per mile, split return: Dalston Kingsland-Dalston Junction, £14.00, 0.136 miles, 10290.0ppm
Min. price per mile, single: Appley Bridge-Glossop, £4.20, 33.1 miles, 12.7ppm
Max. price per mile, single: Forest Gate-Wanstead Park, £9.80, 0.139 miles, 7059.5ppm
Max. price per mile, split single: Forest Gate-Wanstead Park, £7.70, 0.139 miles, 5546.8ppm

Maximum Absolute Savings Using Splits:
Return: Bicester North-Severn Tunnel Junction/Caldicot: £490.80, with splits £71.90
Single: Lockerbie-Newquay (/some others nearby): £320.80, with splits £103.50

Maximum Percentage Savings Using Splits:
These are tricky. Technically it’s Camarthen-Aberystwyth, which is £91.20 single, but it can be split by using a bus to be £9. I’m not sure if this counts. If not, then it’s the same as the absolute ones.

Overall Savings as a Result of Splits:

Returns:
Unsplit mean pence per mile (ppm): 80.6
Single-split mean ppm: 65.3, 17.8% cheaper on average
Multi-split mean ppm: 53.7, 32.7% cheaper on average
92.1% of fares can be made cheaper by splitting

Singles:
Unsplit mean ppm: 61.3
Single-split mean ppm: 47.1, 19.7% cheaper on average
Multi-split mean ppm: 42.5, 27.6% cheaper on average
89.1% of fares can be made cheaper by splitting

Least/Most Expensive Stations (mean price per mile for all destinations):
Min. price per mile, returns: Thurso, 60.2ppm
Min. price per mile, split returns: Garelochhead, 40.5ppm
Min. price per mile, singles: Thurso, 39.3ppm
Min. price per mile, split singles: Thurso, 30.9ppm
Max. price per mile, returns: Heathrow T4, 138.7ppm
Max. price per mile, split returns: Harrow on the Hill, 114.5ppm
Max. price per mile, singles: Heathrow T4, 95.2ppm
Max. price per mile, split singles: Farringdon, 69.4ppm

Most splits needed to achieve the cheapest price (obviously there are multiple contenders here):
Return: Lelant Saltings-Dyce. £328.20 vs £238.80. 14 splits:
BAS-TAU-ASD-BMV-BTG-STA-SNE-BLK-ADL-EGY-INK-DEE-STN-ABD
Single: Kingham-Dalmally. £165.80 vs £132.30. 12 splits:
HYB-BMV-ALV-STA-BRT-SOT-MAN-WGN-PRE-GRL-GCH-ART

Some graphs are attached as well, which might be of even more interest (especially the average price as a function of origin station latitude, I like that one!). Hopefully they're self-explanatory, except for the tartan-looking colour maps, which are the price from every station to every other station, ordered by latitude on both axes.

Obviously the prevalence of Advance fares makes most of this not particularly useful for long-distance trip planning. But the algorithm I've used here would work perfectly well for Advance fares on a single journey -- I can't wait for the day when they just do the maths such that A-B-C never costs less than A-C for advances on a single train! This could be taken a step further by reducing any tickets using the route of said advance train by the same amount, to ensure completely fair system. One can hope.

As I said, if there's keenness to make these splits into a searchable database, then I can do so. Similarly, any requests for other statistics are welcome.

Hope this is of interest!

Oliver

 

[Adam Williams]

How are you accounting for varying stopping patterns in determining which splits are valid for a particular journey?

One caveat is that I’ve not used the official routeing guide at all here – it's just too complicated. But cheaper splits that contravene the guide shouldn’t exist, in an ideal world

From an implementation perspective I can definitely understand taking the shortcut here, but it's a significant caveat that will mean some of the split possibilities in this analysis cannot actually be used in any practicality (e.g. banned by a negative easement, not permitted by a route restriction) on the segments you were hoping they would be used on.

 

[johnsmith147]

How are you accounting for varying stopping patterns in determining which splits are valid for a particular journey?

I'm not looking into such things (as it would be an extreme rabbit hole, and computationally far more difficult). This is literally just the cheapest combination of the cheapest walk-ups to get you between the destinations -- so you will be able to get to your destination that cheaply, but might take a very long time!

== Doublepost prevention - post automatically merged: 8 Jul 2026 ==

How are you accounting for varying stopping patterns in determining which splits are valid for a particular journey?


From an implementation perspective I can definitely understand taking the shortcut here, but it's a significant caveat that will mean some of the split possibilities in this analysis cannot actually be used in any practicality (e.g. banned by a negative easement, not permitted by a route restriction) on the segments you were hoping they would be used on.

Indeed, many of them are completely impractical. But I'd argue they would still represent a 'fair' minimum fare between the stations. I'm not particularly intending for this to be of any practical use -- splitting websites already do as much. I'm more trying to make a point that it's all a bit broken.

If someone can point me in the direction of some code which can calculate which stations are valid under permitted routes, then this could be improved upon, but I am yet to find such a thing.

 

[blimmo]

From an implementation perspective I can definitely understand taking the shortcut here, but it's a significant caveat that will mean some of the split possibilities in this analysis cannot actually be used in any practicality (e.g. banned by a negative easement, not permitted by a route restriction) on the segments you were hoping they would be used on.

I think OP is assuming that combining tickets for A-B with B-C will necessarily produce a ticket for a journey (not necessarily a direct train ofc) from A-C. I don't think that a negative easement would prevent that? I suppose it might mean there are no actual trains that would allow the journey to be completed but to verify that would require a full journey planner not just an implementation of the routeing guide

 

[Adam Williams]

I suppose it might mean there are no actual trains that would allow the journey to be completed

This is effectively my point. If the NRG effectively means that there are no trains that could be used, does it constitute a valid combination of tickets anymore?

 

[blimmo]

This is effectively my point. If the NRG effectively means that there are no trains that could be used, does it constitute a valid combination of tickets anymore?

Does that actually happen all that often?

 

[walkingsophie]

Hi, this is all amazing data, I love it. I am also working on trying to create something similar but with routeing data included to find the cheapest tickets and splits where you can start/end short. I know some of the routes you can draw have no way to fulfill them, but I am hoping that that wouldn't be much of a problem.

How are you accounting for varying stopping patterns in determining which splits are valid for a particular journey?

Assuming that every ticket is possible to use, there is some combination of trains to be on that makes these splits useable, though you might need tickets over multiple days to account for the time it takes.

 

[johnsmith147]

Hi, this is all amazing data, I love it. I am also working on trying to create something similar but with routeing data included to find the cheapest tickets and splits where you can start/end short. I know some of the routes you can draw have no way to fulfill them, but I am hoping that that wouldn't be much of a problem.


Assuming that every ticket is possible to use, there is some combination of trains to be on that makes these splits useable, though you might need tickets over multiple days to account for the time it takes.

Thanks! And yes, I think the above points are moot as if the ticket exists, then a route should exist between them, even if it's silly. As I said, it's not meant to be practical...

Have you gotten anywhere with the routeing guide and automating such things? Of course to make this point properly routes should be accounted for, but I don't really have the time to take this any further for now.

 

[tutacre]

I think a searchable database would absolutely be useful and very helpful! With fares reform around the corner anyway, the risk of them being shut down or "fixed" is low as they'll disappear later anyway, if they really want to. In any event it's not apparent how they could fix all the splits just to prevent niche savings - without affecting "local" non-split journeys too.

Go for it!

Maybe ask BRFares.com if they'd be up for a collaboration?

 

[mickey]

Looks great! And yes I agree a searchable database would be incredibly useful. My thought also went to BRFares, I’d be using it alongside that in any case.

I’m not sure I understand the hope/optimism/expectation here that there’ll soon be fares reform that will remove splitting opportunities. If nothing else it would be an enormous shake-up that takes a huge amount of planning (and therefore time), but would likely also be a negative for most of us here in terms of cost and flexibility.

 

[Mainline421]

Min. price per mile, return: Wye-Ashurst New Forest, £21.20, 108.9 miles,19.5ppm

First immediate thought was this doesn’t make sense, Wye to Beaulieu Road is the same price

 

[johnsmith147]

Looks great! And yes I agree a searchable database would be incredibly useful. My thought also went to BRFares, I’d be using it alongside that in any case.

I’m not sure I understand the hope/optimism/expectation here that there’ll soon be fares reform that will remove splitting opportunities. If nothing else it would be an enormous shake-up that takes a huge amount of planning (and therefore time), but would likely also be a negative for most of us here in terms of cost and flexibility.

Thanks! It would take a fair bit of planning, indeed, but it needs to be done...

Although I'm one of them, I don't feel some people should get cheaper train tickets just because they know how the system works, it should just be easy and fair for all. I wouldn't propose that anything increases in price at all.

== Doublepost prevention - post automatically merged: 9 Jul 2026 ==

First immediate thought was this doesn’t make sense, Wye to Beaulieu Road is the same price

Ashurst is further from Wye than Beaulieu road in a direct line.

 

[AlterEgo]

Thanks! It would take a fair bit of planning, indeed, but it needs to be done...

Although I'm one of them, I don't feel some people should get cheaper train tickets just because they know how the system works, it should just be easy and fair for all. I wouldn't propose that anything increases in price at all.

You're proposing a significant loss of revenue for the railway instead. I have to say I'm not really sure what the point of this data aggregation is. What's the utility?

 

[styles]

You're proposing a significant loss of revenue for the railway instead. I have to say I'm not really sure what the point of this data aggregation is. What's the utility?

I don't think they're actually proposing a lot of these specific finds as such.

Doing this journey:

Single: Kingham-Dalmally. £165.80 vs £132.30. 12 splits:
HYB-BMV-ALV-STA-BRT-SOT-MAN-WGN-PRE-GRL-GCH-ART

would be ridiculous, particularly to save 20% on a £166 fare. I also imagine the number of non-split tickets for Kingham-Dalmally to be close to, if not exactly, zero in a given year

They've said they're not accounting for the NRG, stopping patterns, Advance fares etc so I think it's more theoretical than practical.

I think it's interesting just because sometimes data for data's sake is fun.

The railway should work for passengers though, so the general idea that passengers should be able to get the best value fares I think is a good one. It's a shame that split ticketing is quite so necessary in this country. I've certainly done it abroad as well, but I find in GB that pretty much any medium to long distance journey is cheaper, often a lot cheaper, by splitting.

 

[Swedenorer]

Anybody who thinks fares simplification is going to come quickly or at all will have a long wait I fear. The fact that you have to have a through fare - proabaly two or three between - Wye and Ashurst New Forest indicates the scale of the task. Most of the work would be similarly pointless. Then you'd have to deal with the pricing, peak hour restriction and easement anarchy that's been allowed to develop over recent years. The fact that the four through fares between Bicester North and Severn Tunnel Jct are set by XC rather hints at the issues irrespective of the fact that the number of tickets sold maybe zero a year. (Mind you in an age when we can have vanity by-elections anything is possible maybe.) As for the average savings using splits I suppose the really useful variable in this kind of analysis would be to come up with an index of inflexibility to go with the savings.

 

[tutacre]

Anybody who thinks fares simplification is going to come quickly or at all will have a long wait I fear. The fact that you have to have a through fare - proabaly two or three between - Wye and Ashurst New Forest indicates the scale of the task. Most of the work would be similarly pointless. Then you'd have to deal with the pricing, peak hour restriction and easement anarchy that's been allowed to develop over recent years. The fact that the four through fares between Bicester North and Severn Tunnel Jct are set by XC rather hints at the issues irrespective of the fact that the number of tickets sold maybe zero a year. As for the average savings using splits I suppose the really useful variable in this kind of analysis would be to come up with an index of inflexibility to go with the savings. Mind you in an age when we can have vanity by-elections anything is possible maybe.

I think you're making a dangerous assumption in the current climate - you don't have to have a through fare, certainly not two or three.

There probably will be a through fare in most cases, but they won't necessarily be obligated to. I doubt they will start completely from scratch here, but I do expect considerable overhaul to the fares system - not simplification - a complete change of approach from the buying process to the ticket check on the train.

Equally, Irish Rail bans split ticketing - nothing really stopping that from being added in to an existing byelaw or contractual condition of travel.

 

[johnsmith147]

You're proposing a significant loss of revenue for the railway instead. I have to say I'm not really sure what the point of this data aggregation is. What's the utility?

It gives an indication of what the true 'fair' cheapest price should be between any two stations. I'm not saying this is going to happen, but shows that calculating such a thing takes a few minutes' computation rather than the standard line of 'this is so complex it would take aeons to unravel' line which is usually stated. I don't think we all should receive a discount for knowing more about how the railway works.

would be ridiculous, particularly to save 20% on a £166 fare. I also imagine the number of non-split tickets for Kingham-Dalmally to be close to, if not exactly, zero in a given year

It is, but I suspect the saving would be close to that with far fewer splits. The best you can do with only one split is £158.30, though, so maybe not... I've been offered 8 splits by Trainsplit on XC journeys before (for a sensible route), which is the same order of magnitude of ridiculous.
The approach I've taken could use the NRG, but I'd need to find some open-source code which can deal with it. I think everything the retailers use is proprietary.

The fact that you have to have a through fare - probably two or three between - Wye and Ashurst New Forest indicates the scale of the task.

You don't have to go through all 8 million (ish) combinations manually though, so it would actually not be very difficult for a 'cheapest' (let's say off-peak) fare. It becomes even easier if returns are abolished (but this is only desirable if singles are reduced to precisely half of the previous return fare, but that didn't happen on LNER so...

Dealing with routes would be the same difficulty as what I've done already (order n^3 for n stations), you'd just need to combine the restrictions linearly.

Peak-train timings is more difficult. I think the only fair way would just be to define some journey legs (as in, a particular train between two particular stops) as a 'peak' service, for which you need an Anytime ticket. Then say you want a train YRK-KGX, but it's defined as peak from PBO-KGX, then your fare would be the Off-peak price for YRK-KGX, plus the difference between the Off-peak and Anytime price from PBO-KGX. I think that's the only way to do it without introducing more mess.

I think you're making a dangerous assumption in the current climate - you don't have to have a through fare, certainly not two or three.

Yes, you do. Or at least, when you search for fares between two stations on a machine or on an app, it needs to tell you the best price, and you need to have the same rights as any other ticket with regard to disruption etc.. People can't be expected to know which stations they'd need to stop/change/split at.

Equally, Irish Rail bans split ticketing - nothing really stopping that from being added in to an existing byelaw or contractual condition of travel.

This would be a massive step backwards, and must be quite impractical to enforce.

Advance fares are another matter, and one that could be improved upon very easily and immediately (no questions on routeing or peak time trains), but the time-dependent nature of the data make a large-scale analysis impractical. Perhaps I could obtain all advance ticket combinations for a single long XC journey at a particular time and see how 'unfair' they are. Ideally, long-distance advance fares should be constrained by combinations of the legs (between individual stops) which make them up. That would be computationally pretty much instant (and of course would need to be done every time a fare is sold), as even the longest advance journeys only have ~40 stops.

 

[signed]

This would be a massive step backwards, and must be quite impractical to enforce.

Devil advocate: If someone gives the conductor 2 tickets for the same train, and/or if the conductor comes back to check again and they give another ticket than the one previously scanned, then just fine them. That's basically how it is written for Irish Rail

https://assets-eu-01.kc-usercontent.com/d2e85689-fbf9-010a-a02e-587ed94b5032/26fdf067-aa8d-4125-ad94-2f7edd083c7b/d9eaa54d-308c-442c-a8f3-0eebe9cfc010.pdf

Except where specially authorised, passengers are not permitted to re-book at an intermediate station for the purpose of continuing their journey by the same train. Two or more tickets covering different portions of one journey are not available unless the fares paid for such tickets are equivalent in amount to the price of a single journey ticket between the same points. Any passenger using two or more tickets covering different portions of one journey will be liable to pay the full ordinary single fare for the journey made and he or she may be liable to prosecution

Of course in DOO land it's unenforcable

 

[Adam Williams]

Devil advocate: If someone gives the conductor 2 tickets for the same train, just fine them, or if the conductor comes back to check again and they give another ticket.

Until customers start changing their appearance between the two checks and moving around the train

 

[johnsmith147]

Devil advocate: If someone gives the conductor 2 tickets for the same train, just fine them, or if the conductor comes back to check again and they give another ticket.

What if you get off at the split station and get on different train a few minutes later? Worth doing such things if it saves you loads!

 

[30907]

A couple of observations: Wye to Ashurst (and similar) is route Barnham, which is pretty indirect. The equivalent via London is 3x as much!
At the other end, quoting paper fares door anywhere in the Oyster area distorts things hugely.

 

[signed]

What if you get off at the split station and get on different train a few minutes later? Worth doing such things if it saves you loads!

Even in places (XC land comes to mind) where it isn't rare for tickets to give you 5+ splits? You'd take the full day shuffling trains

Until customers start changing their appearance between the two checks and moving around the train

 

[walkingsophie]

Ashurst is further from Wye than Beaulieu road in a direct line.

You can use Dijkstra on the routeing guide data station links record to calculate the track distance between stations to get a price per track distance statistic which could be interesting.

Note that unlike the fares data this is comma separated so do be careful when parsing

 

[zwk500]

Similarly, any requests for other statistics are welcome.

Is it possible to see a comparison integrating demand as well as distance?

 

[greatkingrat]

What if you get off at the split station and get on different train a few minutes later? Worth doing such things if it saves you loads!

There is no way to stop that, however in most cases it won't be a few minutes, it will be half an hour or more until the next train.

 

[johnsmith147]

You can use Dijkstra on the routeing guide data station links record to calculate the track distance between stations to get a price per track distance statistic which could be interesting.

Indeed. Could you point me where this raw data could be obtained/do you have code I could use that will do it? Feel free to message if so.

Is it possible to see a comparison integrating demand as well as distance?

I'd like to, if the station-station usage data are available (I'm not sure where, precisely?) then that would lead to much more meaningful statistics!

 

[The exile]

Devil advocate: If someone gives the conductor 2 tickets for the same train, and/or if the conductor comes back to check again and they give another ticket than the one previously scanned, then just fine them. That's basically how it is written for Irish Rail

So you hold an advance ticket from A-G and at the last minute you discover you need to go to H (ie 1 station further instead). The advance contingent you bought weeks ago has all gone so A-G now costs double what you paid. What are you supposed to do? Buy a new ticket from A-H and claim back the one you’re now not using for A-G or just buy a single from G-H - which means you’re effectively split ticketing.

 

[AlterEgo]

It gives an indication of what the true 'fair' cheapest price should be between any two stations.

It doesn't. It gives, quite simply, the cheapest combination of (flexible) splits - that is what it does. And, as you admit, it doesn't mean the splits are even usable in a practical sense, because a lot of them will tie you to 500 miles on stopping trains so you call at Bere Alston, Ashchurch for Tewkesbury, Stone, Blackrod etc etc all the way up to Aberdeen.

This is quite limited for start, as most people splitting are doing so because they are price sensitive and therefore would consider Advances, which are quota controlled and train-specific.

While I agree that in an ideal world people shouldn't be advantaged simply because they are power users of a public good, we have the system we have, and I do not think this tool would bring big-picture fairness in the way you think it does.

I'm not saying this is going to happen, but shows that calculating such a thing takes a few minutes' computation rather than the standard line of 'this is so complex it would take aeons to unravel' line which is usually stated. I don't think we all should receive a discount for knowing more about how the railway works.

What do you think the practical upshot of the industry knowing the cheapest combination of splits for every single station pair would be? That all the long distance fares would simply go down to combine all the splits into a through ticket, and the railway lose revenue? Or, as has been the case since the 90s, would many popular fare flows used for splitting simply rise to prevent creative use, making other people's short, unsplittable journeys more expensive?

 

[signed]

I'd like to, if the station-station usage data are available (I'm not sure where, precisely?) then that would lead to much more meaningful statistics!

Station-station real usage isn't possible, there is a O/D matrix based on ticketing (so it is skewed by things like PAYG/Split)

Rail Data Marketplace

https://dataportal.orr.gov.uk/media/j1mdc31j/passenger-rail-usage-jan-mar-2026.pdf

Page 18 of the above PDF explains the methodology to try to unskew Elizabeth Line data

 

[johnsmith147]

It doesn't. It gives, quite simply, the cheapest combination of (flexible) splits - that is what it does. And, as you admit, it doesn't mean the splits are even usable in a practical sense, because a lot of them will tie you to 500 miles on stopping trains so you call at Bere Alston, Ashchurch for Tewkesbury, Stone, Blackrod etc etc all the way up to Aberdeen.

Ideally it would be improved to be constrained to reasonable routes -- I've said as much.

I do not think this tool would bring big-picture fairness in the way you think it does.

Of course not, I know that's not how the world works... I'm just trying to make a point and some nice graphs.

That all the long distance fares would simply go down to combine all the splits into a through ticket, and the railway lose revenue?

Yes, precisely. If most people are using splits anyway (I'm actually forced to at work these days!), then there's nothing to be lost. I see no reason to be so pessimistic.

== Doublepost prevention - post automatically merged: 9 Jul 2026 ==

Station-station real usage isn't possible, there is a O/D matrix based on ticketing (so it is skewed by things like PAYG/Split)

Page 18 of the above PDF explains the methodology to try to unskew Elizabeth Line data

Thanks -- that link didn't come up in my search for some reason. I'll have a look at integrating this next week. The correction isn't really doable without data feeds I can't access.

Alas the stats are always going to be a bit messed up by London, as nobody is actually buying Off-peaks or Anytimes any more. Perhaps best to just exclude intra-London journeys for now...