An idea for how to calculate the load factors in a better way

When building up my network for the planes, I was thinking about the load factors. Somehow, they are a bit annoying, e.g. it can make a big difference if you change the ticket price by just one dollar. Furthermore, it seems to be useless to build an airline with working with hubs, as the connecting passengers are not calculated realistically (e.g. if I have a hub in FRA, having flights to LHR and to some small German airport, with me being the only one flying to this small airport, I should get all the passengers that want to fly between LHR and this small airport, at least when neglecting trains and busses). So I had some thoughts about how this could be made in a better way, and as it would be a real improvement, I wanted to share these thoughts; maybe they will help the developers in making AE nicer, and if not, it's just one small piece of text in the anyhow flooded internet

.

As far as I'm concerned, the load factors only take the considered route into account; furthermore, connecting passengers are created from nowhere. However, in a realistic world, connecting passengers would naturally be drawn from another route. E.g. if I get a connecting passenger on LHR to FRA, he must fly to somewhere from FRA; let's say it is CDG. However, then this connecting passengers has to be drawn away from the route LHR-CDG. To get a realistic simulation, such effects would have to be included.

As it is almost impossible to include all such effects if the load factors are calculated immediately (e.g., if I draw this passenger from LHR-CDG to connect with me in FRA, maybe this free seat on LHR-CDG will then be used by a passenger flying from JFK-CDG connecting in LHR), this can only be done by calculating the load factors maybe once a day.

My idea of how this could be done is the following: Use a Monte-Carlo-Method. In particular, do not simulate the routes (as done now), but the passengers. Therefore, every day you do the following procedure 200,000 times (i.e. for 200,000 passengers):
1. Choose a starting point. The starting point is chosen randomly, with the probability corresponding to the size of an airport.
2. Choose the end of the journey. Also chosen randomly, but the probability now depends on the demand for each route to another airport.
3. Choose the luxury standards the passenger wants, set a minimum. For the luxury, find a formula how to put IFE, flight class, leg room, airline image and length of flight into one number - the class should probably have the biggest impact.
4. Depending on the distance of the flight, the luxury standard and some randomness, set a maximum ticket price the passenger would like to pay.
5. Find the possible routes with up to two connections (connections can be done only at Hubs and only between flights within an alliance) which are below 1.5 times the maximum the passenger would like to pay; to account for good or bad connections: add a time for the connection to the flight time; it should depend on the number of daily flights from the starting point to the hub and from the hub to the next point. Only flights which are not booked out can be chosen (see below).
6. Give each flight a probability depending on luxury and price. In general, the probability function should lower with the price and increase with the luxury; there should be a strong drop at the maximum money the passenger likes to pay and a strong increase at the minimum luxury the passenger would like.
7. According to the probability, the passenger now chooses one flight and is added as a daily passenger to this flight.
8. If there was no suitable flight at all, the passenger doesn't fly at all.

According to the IATA, there are something like 2.5 or 3 billion flight passengers a year (worldwide), which makes roughly 5-10 million passengers a day. However, as calculating 10 million pasengers a day (i.e. every 30 minutes) may take some time, I suggest only to calculate every day 200,000 passengers (or, at the early 80s probably less), and do the following: For every route, the load consists of 24 parts (standing for the 24 days of an AE-month), each part is calculated once a month. The parts are divided almost equally, but so that they contain an integer number of seats (otherwise it would get complicated). When the load factor for the ith part is calculated, a flight is only taken into account, if this ith part is not fully loaded yet.

So, these were my thoughts. Although it might be a big change, I think it's worth considering, as it would introduce a lot more realism. Also, I guess it is feasible, and if the 200,000 passengers are too much to calculate every half an hour, one can also get to a smaller number by adding another variable which is the number of passengers travelling together. Thereby the number can be reduced further. Also, this method adds some fluctuation in the passenger numbers, and also it overcomes some other problem we have right now: Opening a route at a time of high demand leads to much better load factors than at a time of low demand, and if no one recalculates the route by changing something, such things can be kept.

Also, there are some other nice things about this scheme: It takes some time until a route is fully used (one moth due to the 24 parts), it is not clever to open and close routes very often. Also, one might can show the players, which routes the connecting passengers actually take. There is the possibility to pay extra for a good hub, which would then lower the flight time added for the connection by 50%. Also, one could allow the players not only to set the ticket price for the direct flight, but also for connecting passengers, thereby making sure that on a route where the ticket price is just accounting for the cost without making profit the connecting passengers can not make this route loosing money. However, one could even set the price for connecting passengers higher, as the competition on that routes might be less. On the other hand, it might be even useful to have a route making loss because it brings passengers to a high-profit transatlantic flight.

Yes, there are also some problems: How can the players find out whether a route is good or not if they install it? Well, they can't directly. But they can look at the demand of the route, and maybe they can also be shown the demands from this route to airports in reach by their hubs. Of course, one would have to think about how to display this. And also I have no idea how long it would take to calculate 200,000 passengers. If it is one minute, it's probably okay, if it is 10 minutes, it would be annoying. However, there would be ways to shorten this time, e.g. by including groups into the passengers, by dividing the load factor into 48 parts etc.

EDIT: I have seen now that there is something like this planned for version 4; so I hope it my thoughts will help a bit.
Ok, as I already mentioned, these are just some thoughts. But they would bring a lot of great things to AE, making it far more realistic, allowing to compete on a route also by connecting flights, making alliances more useful (and making it less clever to join an alliance with players having almost the same network), forcing the players to build up more intelligent networks, and allowing to bring the demands to more realistic values, so I think there is some use in my idea.

Edited by joflo, 27 August 2010 - 05:57 PM.