How unpredictable are the algorithms that idenify the cheapest split ticket options?

[TUC]

I compared the split ticket options offered by Trainsplit and by Trainline for the same journeys.

All were for travel on Wed 12 June after 1000 (on the basis that it's a date that Advance tickets should be available, but not so close that they will be likely to have sold out.). The fares set out below are the cheapest journeys from the options offered.

1. Halifax to Plymouth. Trainsplit offered an £87.47 fare with seven splits and one change of train. Trainline offered a £96.20 fare with one split and one change of train.

2.Alexandria to Middlesbrough Trainsplit offered £27.97 with one split and three chsnges of train. Trainline offered £32.90 with a single ticket and three changes of train.

3. Oxford to Selby. Trainsplit offered £43.75 with two splits and three changes of train. Trainline offered £48.00 as a single ticket and two changes of train. (Both included tube transfers within these.)


What are the likely reasons for the differences? Is it:

a) Different Advance ticket allocations
b) Different number of ticket splits considered (although I was not asked how many I would accept(
c) Different numbers of changes of train acceptable
d) Different wider algorithms
e) Some other reason?

 

[MrJeeves]

a) Different Advance ticket allocations

Advance availability should theoretically be the same across all retailers.

Retailers with more sales naturally get more up-to-date availability information which they can cache (this is how Trainline can tell you what advances on other days start at, for example), which means that you're less likely to suddenly face a price increase when you try and book because the cache was out of date.

b) Different number of ticket splits considered (although I was not asked how many I would accept)
c) Different wider algorithms

Trainline typically finds fewer splits than TrainSplit's journey planner does, because it's simply not set up to know as many. I'm sure you've noticed that TrainSplit can sometimes feel slower than Trainline's journey planner (especially for longer or more complex journeys), and I'm sure you can guess why. I think it's safe to say that different retailers want to prioritise different aspects of their booking process, and speed is important to many customers and hence many to businesses too.

The journey planner TrainSplit uses (FastJP) has been fine-tuned over many years by very knowledgeable people to consider as many split-ticketing locations as possible, try various combinations of these locations together, and does so rather aggressively to help to reduce the cost.

If someone spots a split that saves money which TrainSplit/FastJP doesn't find, posting about it in the forum ticketing thread often means it'll be fixed for the next day which is service you don't get with any other retailer/JP as far as I'm aware!

There is no feature to specify "maximum split count" for either journey planner.

It's worth mentioning that Trainline can be cheaper, typically on routes where there is only one worthwhile split, as they do not charge anything except the base price of the tickets (* if you buy on the day, through the app), while TrainSplit will charge a portion of the saving compared to a through ticket.

I don't actually know anything about how Trainline's journey planner or split ticketing engine works, and I also don't really know much about how FastJP actually works behind the scenes, but I can usually see quite clearly how the results from the two differ.

 

[Watershed]

Trainline seems only to offer splits at a rather short, predefined list of stations, and in a rather suboptimal manner usually only involving singles. That can still deliver savings vs the cost of through tickets but retailers that have a wider range of split points and/or number of splits will inherently find a better choice of splits.

 

[Adam Williams]

Trainline seems only to offer splits at a rather short, predefined list of stations, and in a rather suboptimal manner usually only involving singles. That can still deliver savings vs the cost of through tickets but retailers that have a wider range of split points and/or number of splits will inherently find a better choice of splits.

This nuance is of course lost on most customers, who just see that their retailer "offers split tickets" due to prolific marketing - and unknowingly end up paying more than they need to.

 

[skyhigh]

Trainline seems only to offer splits at a rather short, predefined list of stations, and in a rather suboptimal manner usually only involving singles. That can still deliver savings vs the cost of through tickets but retailers that have a wider range of split points and/or number of splits will inherently find a better choice of splits.

Am I right in thinking Trainline will only ever offer a single split, whereas Trainsplit will use as many as possible?

 

[talldave]

To answer the OP's question, the algorithms are not unpredictable at all as they're hard coded and I'm assuming they aren't coded to randomly not give the cheapest option to some users! However, I suspect the question was really "how different"...

 

[Watershed]

Am I right in thinking Trainline will only ever offer a single split, whereas Trainsplit will use as many as possible?

That's right, in my experience.

 

[Deafdoggie]

Am I right in thinking Trainline will only ever offer a single split, whereas Trainsplit will use as many as possible?

I don't think they use as many as possible! Just as many as required to get the cheapest fare! But I believe Trainline is limited to just 1

This nuance is of course lost on most customers, who just see that their retailer "offers split tickets" due to prolific marketing - and unknowingly end up paying more than they need to.

I was surprised with recent Trainline adverts that said the average customer saved £13 with split tickets. It didn't sound a lot to me. I know it's an average so some will save more, but it still didn't seem a massive saving. But I guess it depends what the average fare is.

 

[Paul Kelly]

To answer the OP's question, the algorithms are not unpredictable at all

Could sometimes be slightly unpredictable due to the need to avoid making too many reservation availability requests and the caching required to achieve that aim, which means an opportunity could be missed if the cache is out of date and lower-priced advances had been recently released for one segment of the journey, for example.

I don't think they use as many as possible! Just as many as required to get the cheapest fare!

It has been known for FastJP to offer more splits than really necessary, possibly even when a cheaper price would have been available with fewer splits. But generally nobody would really notice this - sometimes you get a really long set of splits and might think "wow, is it really necessary to have so many splits to get the best price" and the answer might be that it actually wasn't. But if the customer is happy with the price and it is better than any other journey planner can come up with, maybe it's not a big problem?

 

[TUC]

Could sometimes be slightly unpredictable due to the need to avoid making too many reservation availability requests and the caching required to achieve that aim, which means an opportunity could be missed if the cache is out of date and lower-priced advances had been recently released for one segment of the journey, for example.


It has been known for FastJP to offer more splits than really necessary, possibly even when a cheaper price would have been available with fewer splits. But generally nobody would really notice this - sometimes you get a really long set of splits and might think "wow, is it really necessary to have so many splits to get the best price" and the answer might be that it actually wasn't. But if the customer is happy with the price and it is better than any other journey planner can come up with, maybe it's not a big problem?

I don't think anyone seeking the cheapest price would complain about the number of splits.

To answer the OP's question, the algorithms are not unpredictable at all as they're hard coded and I'm assuming they aren't coded to randomly not give the cheapest option to some users! However, I suspect the question was really "how different"...

Perhaps 'unpredictable' wasn't the right word. What I meant, for example, is where a journey has several different route options, all with multiple differing splits possible, how well can the algorithms deal with that many variables? Also, if the algorithms can deal with them, how do different ticket split sites come out with different preferred options?

 

[Deafdoggie]

Could sometimes be slightly unpredictable due to the need to avoid making too many reservation availability requests and the caching required to achieve that aim, which means an opportunity could be missed if the cache is out of date and lower-priced advances had been recently released for one segment of the journey, for example.


It has been known for FastJP to offer more splits than really necessary, possibly even when a cheaper price would have been available with fewer splits. But generally nobody would really notice this - sometimes you get a really long set of splits and might think "wow, is it really necessary to have so many splits to get the best price" and the answer might be that it actually wasn't. But if the customer is happy with the price and it is better than any other journey planner can come up with, maybe it's not a big problem?

I'm not saying it's a problem. But to get the greatest number of splits would require a split at every single station the journey called at. Which clearly isn't what happens.

 

[yorkie]

I compared the split ticket options offered by Trainsplit and by Trainline for the same journeys.

All were for travel on Wed 12 June after 1000 (on the basis that it's a date that Advance tickets should be available, but not so close that they will be likely to have sold out.). The fares set out below are the cheapest journeys from the options offered.

1. Halifax to Plymouth. Trainsplit offered an £87.47 fare with seven splits and one change of train. Trainline offered a £96.20 fare with one split and one change of train.

2.Alexandria to Middlesbrough Trainsplit offered £27.97 with one split and three chsnges of train. Trainline offered £32.90 with a single ticket and three changes of train.

3. Oxford to Selby. Trainsplit offered £43.75 with two splits and three changes of train. Trainline offered £48.00 as a single ticket and two changes of train. (Both included tube transfers within these.)


What are the likely reasons for the differences? Is it:

a) Different Advance ticket allocations
b) Different number of ticket splits considered (although I was not asked how many I would accept(
c) Different numbers of changes of train acceptable
d) Different wider algorithms
e) Some other reason?

My understanding is that Trainline want to keep the number of tickets down and aren't fussed at being undercut by Trainsplit.

As long as Trainsplit (and, by extension, the forum's site) identifies the cheapest combination, then I am happy and nothing unexpected is happening.

However, if anyone sees that Trainline (or any other provider) identifies a cheaper combination for any given journey, please do let me know, and I will ensure that this is investigated and passed onto the relevant person; we can usually get it sorted within 24 hours!

It is very rare this happens; the last time was 5 months ago.

 

[redreni]

I don't think anyone seeking the cheapest price would complain about the number of splits.

That depends on the circumstances.

For example London to Weston-Super-Mare on flexible tickets tends to be much cheaper splitting at Didcot Parkway (especially for me as a Network Railcard holder as the ticket to Didcot attracts the 1/3 discount off-peak), but the direct trains tend not to stop there so it means you have to take a slightly slower train and change at Bristol.

It's worth it because the saving is significant but it wouldn't be worth it if the saving was small or non-existent.

On the other hand a split at a station at which all the trains you might want to catch have an open call is no issue at all.

 

[TUC]

My understanding is that Trainline want to keep the number of tickets down and aren't fussed at being undercut by Trainsplit.

As long as Trainsplit (and, by extension, the forum's site) identifies the cheapest combination, then I am happy and nothing unexpected is happening.

However, if anyone sees that Trainline (or any other provider) identifies a cheaper combination for any given journey, please do let me know, and I will ensure that this is investigated and passed onto the relevant person; we can usually get it sorted within 24 hours!

It is very rare this happens; the last time was 5 months ago.

In all the journey scenarios I tested Trainsplt was cheaper than Trainline. What concerns me is the usual issue of so many people defaulting to using Trainline. They might, not unreasonably, think that all ticketing sites claiming to offer splits will have similar products when in reality it looks like Trainline is distinctly second rate, not least in the limited number of splits it considers.

 

[yorkie]

In all the journey scenarios I tested Trainsplt was cheaper than Trainline.

Good; please let me know if you see any exceptions!

What concerns me is the usual issue of so many people defaulting to using Trainline. They might, not unreasonably, think that all ticketing sites claiming to offer splits will have similar products when in reality it looks like Trainline is distinctly second rate, not least in the limited number of splits it considers.

Trainline are going for a particular market and they do well at that.

Enabling one split point enables them to have enough of a selling point, and generate enough of a saving, to expand that market, without overcomplicating things.

People who are particularly price sensitive, and/or power users, who require multiple via/avoid points or seat selectors etc, are not in their target demographic.

 

[TUC]

Trainline are going for a particular market and they do well at that.

Enabling one split point enables them to have enough of a selling point, and generate enough of a saving, to expand that market, without overcomplicating things.

People who are particularly price sensitive, and/or power users, who require multiple via/avoid points or seat selectors etc, are not in their target demographic.

They should though make clear that they only offer one split point. I cannot see that anywhere on their app.

 

[Deafdoggie]

They should though make clear that they only offer one split point. I cannot see that anywhere on their app.

I don't see a real need to. They don't offer a lot of other things either, should they list everything they don't offer?

 

[Watershed]

They should though make clear that they only offer one split point. I cannot see that anywhere on their app.

It might be helpful for them to state this in some kind of FAQ, but I don't think they - or any other split ticket retailer - promise to offer the cheapest possible split. The number of theoretically possiblee split combinations is so unbelievably large it would be almost impossible to guarantee this in practical terms.

That being said, I could imagine an enterprising retailer coming up with a "split price guarantee" or similar as a marketing/customer service gesture, whereby if you can show that there was a cheaper split available within (say) 2 hours of booking, they'll refund you the difference.

 

[JB_B]

... That being said, I could imagine an enterprising retailer coming up with a "split price guarantee" or similar as a marketing/customer service gesture, whereby if you can show that there was a cheaper split available within (say) 2 hours of booking, they'll refund you the difference.

It's a nice idea but from a customer service point of view I think that could end up being nightmare for the retailer.

Given the number of "apples and oranges" misunderstandings around fare comparisons you see in this forum simply explaining (and rejecting) misguided requests for refunds could take a lot of resource. You would also have to spend time/money detecting and rebutting malicious claims and you could also end up paying for any bugs/non-compliances/"interesting interpretations" in your competitors' systems that might allow them to offer cheaper fares.

 

[TUC]

It might be helpful for them to state this in some kind of FAQ, but I don't think they - or any other split ticket retailer - promise to offer the cheapest possible split. The number of theoretically possiblee split combinations is so unbelievably large it would be almost impossible to guarantee this in practical terms.

That being said, I could imagine an enterprising retailer coming up with a "split price guarantee" or similar as a marketing/customer service gesture, whereby if you can show that there was a cheaper split available within (say) 2 hours of booking, they'll refund you the difference.

I agree it would be impossible to give such a guarantee, but I also do not think many people use the Trainline site knowing that they only use one split. I think people, if they think about it at all, would assume ticket split sites will use as many as are reasonably practicable.

 

[miklcct]

It might be helpful for them to state this in some kind of FAQ, but I don't think they - or any other split ticket retailer - promise to offer the cheapest possible split. The number of theoretically possiblee split combinations is so unbelievably large it would be almost impossible to guarantee this in practical terms.

If we limit the promise to walk up tickets, isn't it just a simple "dynamic programming" algorithm on the fare database?

 

[AlbertBeale]

Do the algorithms of Trainsplit / the forum also look for savings which aren't a split, but - for example - the idea of buying a ticket from/to somewhere else that enables the journey concerned to be cheaper? An instance might be where different routes between two cities are different prices, but the more expensive route can be done at close to the cheaper price by buying a ticket with a different origin station and starting short? If you only look up fares between the places you're actually travelling from and to, and didn't know this "trick", would Trainsplit know it and offer it? I presume not. Similarly, re the current LNER abolition of off-peak tickets which is much discussed - would Trainsplit or any other site know to suggest buying to a different station instead (and stopping short) so as to save money?

 

[MrJeeves]

would Trainsplit or any other site know to suggest buying to a different station instead (and stopping short) so as to save money?

Doing so would both confuse most customers, and also be an accreditation failure, I believe.

 

[yorkie]

I don't think Trainline are doing anything wrong, except this is confusing me:

Split Tickets | Get Split Tickets with SplitSave from Trainline

Ticket splitting is a great way to save money on rail travel. Read our guide and learn more about how SplitSave works and why split tickets are cheaper.

www.thetrainline.com

Ready to book your split train tickets?

Found the perfect combination of tickets for you? Start your search for SplitSave tickets today.

Do the algorithms of Trainsplit / the forum also look for savings which aren't a split, but - for example - the idea of buying a ticket from/to somewhere else that enables the journey concerned to be cheaper? An instance might be where different routes between two cities are different prices, but the more expensive route can be done at close to the cheaper price by buying a ticket with a different origin station and starting short? If you only look up fares between the places you're actually travelling from and to, and didn't know this "trick", would Trainsplit know it and offer it? I presume not. Similarly, re the current LNER abolition of off-peak tickets which is much discussed - would Trainsplit or any other site know to suggest buying to a different station instead (and stopping short) so as to save money?

Fares to alternative destinations can't be considered.

 

[JB_B]

Do the algorithms of Trainsplit / the forum also look for savings which aren't a split, but - for example - the idea of buying a ticket from/to somewhere else that enables the journey concerned to be cheaper? An instance might be where different routes between two cities are different prices, but the more expensive route can be done at close to the cheaper price by buying a ticket with a different origin station and starting short? If you only look up fares between the places you're actually travelling from and to, and didn't know this "trick", would Trainsplit know it and offer it? I presume not. Similarly, re the current LNER abolition of off-peak tickets which is much discussed - would Trainsplit or any other site know to suggest buying to a different station instead (and stopping short) so as to save money?

No retail site does this.

A couple of earlier threads explore some of the reasons...

Plenty of split ticketing sites...but is there a "long buying" site that identifies cheaper tickets to further along the line of route?

Once you get to Paddington, how do you get to South Wales without passing through Reading? That's the stupid part - you couldn't use this ticket to change trains at Reading, but you could take a train from Oxford to London stopping at Reading and then a train from London to South Wales stopping...

www.railforums.co.uk

National Rail equivalent of Skiplagged (hidden city ticketing price finder)?

Are there any ticket retailers which do hidden-city ticketing on the National Rail network, automatically finding tickets for the customer to start / stop short, preferably one which does multi-city as well (which is the equivalent of Skiplagged for airlines)? For example, if I want to go to...

www.railforums.co.uk

 

[TUC]

I can see

Do the algorithms of Trainsplit / the forum also look for savings which aren't a split, but - for example - the idea of buying a ticket from/to somewhere else that enables the journey concerned to be cheaper? An instance might be where different routes between two cities are different prices, but the more expensive route can be done at close to the cheaper price by buying a ticket with a different origin station and starting short? If you only look up fares between the places you're actually travelling from and to, and didn't know this "trick", would Trainsplit know it and offer it? I presume not. Similarly, re the current LNER abolition of off-peak tickets which is much discussed - would Trainsplit or any other site know to suggest buying to a different station instead (and stopping short) so as to save money?

One reason why that would be difficult is that some people will be intending to travel via a particular route, or breaking their journey at a particular point, and be unaware when the ticket to the alternative destination does not allow that route, or has different break of journey restrictions.

 

[AlbertBeale]

Do the algorithms of Trainsplit / the forum also look for savings which aren't a split, but - for example - the idea of buying a ticket from/to somewhere else that enables the journey concerned to be cheaper? An instance might be where different routes between two cities are different prices, but the more expensive route can be done at close to the cheaper price by buying a ticket with a different origin station and starting short? If you only look up fares between the places you're actually travelling from and to, and didn't know this "trick", would Trainsplit know it and offer it? I presume not. Similarly, re the current LNER abolition of off-peak tickets which is much discussed - would Trainsplit or any other site know to suggest buying to a different station instead (and stopping short) so as to save money?

Fares to alternative destinations can't be considered.

Thanks. I assumed that was the case...

Luckily though, thanks to tips on this site, I anyway know some instances of where I can do a journey more cheaply by buying to and/or from different end-points. "Regular travellers" I mention such things to are usually astounded that there are such possibilities, and of course there's no way that they would discover these options by using a ticket machine or an on-line ticket site, since you'd have to know the right question to ask in the first place.

 

[yorkie]

Thanks. I assumed that was the case...

Luckily though, thanks to tips on this site, I anyway know some instances of where I can do a journey more cheaply by buying to and/or from different end-points. "Regular travellers" I mention such things to are usually astounded that there are such possibilities, and of course there's no way that they would discover these options by using a ticket machine or an on-line ticket site, since you'd have to know the right question to ask in the first place.

It's rife in the aviation industry; it can also apply to the bus industry (but less commonly)

Away from transport, it happens regularly in supermarkets (though analogies are not helpful!), in particular relating to multipacks and multi-purchase deals, and occasionally in restaurants too.

It would be very difficult to stop it being the case; if anyone has any ideas regarding how to stop it, then I look forward to reading them in a new thread in Speculative Discussion!

 

[SickyNicky]

Just to add to this discussion, we absolutely have the ability to find fares that are cheaper by starting or ending journeys short (or both). I think we can actually find the most optimal ticket for this for any journey. However, tickets are offered that are clearly silly, even though technically valid, and which the passenger would have major problems actually using. In any case, I'm confident that our customers wouldn't be allowed to market them.

 

[yorkie]

@AlbertBeale next time you come to a forum meal remind me of this and I will tell you a couple of examples. But the crazy ones are mistakes and would be immediately rectified.

There are a lot of sensible routeings, let alone crazy ones, which have disappeared as greedy train companies, such as EMR for example, wanted to charge more than other companies that priced tickets valid on their routes, and didn't want to be undercut.

Edit: I will further edit this post regarding the EMT example later this evening