Quick beta available

[Sami]

I have been working on a full rewrite of the passenger demand distribution system lately. The old code is based on early days of AWS and has been updated during the way with various bits here and there, and a bit of a cleanup was needed.

There are no major new features but this update will change some aspects of the demand distribution (more specs & details later!) and fix also a couple of bugs (for example a memory leak that has caused problems in MT#6 a few times). This update will be available on any new game worlds. It will still be called v.1.3 for practical reasons, instead of a new version number (to which it very well would be entitled for).

Since this update is a core-level change, testing is needed. Instructions are as follows (see this post later again for further info).

* Cost to join is 2 credits (flat fee). It will be refunded when the game is complete. Requirement for refund is that you are still a member of the game world when it closes down.

* The world has 10 minute days, $100 million start cash. Please get as many planes and routes as you can. Certain features like bases and alliances are disabled.

* Coding is still in progress, and some of the features are missing (see the end of this post).


* Please test especially the following:

   - passenger demand numbers (route planning). Are some routes that should have demand missing the pax demand figures (0 demand)?  Or otherwise any gross errors on the figures?

   - once the time gets started, please test the passenger distribution in a normal way first (to see that everything is logical), and after that please try to do something silly, like adding way too many flights (oversupply) or way too much frequency compared to what is reasonable, or play with the prices too.

   - test separately also the effect of ticket prices (up / down from default).

   - (Note; there is no randomization in the route sales figures; for easier testing - final version will have daily random fluctuation)

   - (Note2; aircraft purchasing limits for first turn and weekly limits are removed for testing purposes)


* Please report bugs on the bugs forum, and please notify clearly that it's about the test scenario. If it's about the pax distribution, don't touch any route data (like prices or flights) before it's investigated.

* Please post all other comments to this thread. Game does not have own forums (or alliance or press release functions).

* Disclaimer: This is a test. Game may be shut down at any point, and it will NOT run for the entire duration.



Feature / changes list of the airport-to-airport demand system:

- This feature determines how many passengers wish to travel from airport A to B.

- The system has been technically modified so that the demand values are calculated and stored differently than before. This is needed for the possible next new demand system (city demand). The actual calculation of the demand figures is exactly the same as in current games though.

- New feature for the system is that any demand changes are gradual. The demand values (airport to airport) are updated on certain intervals and if the values change, there is always a maximum change percentage per game month. (Please note that this may cause issues to airports where huge demand changes occur overnight, like fall of CCCP; untested so far)


Feature / changes list of the new demand distribution system:

- This feature allocates the demand to available flights, based on the data of the route and flights.

- The background process doing all this magic is some 30% faster than before, and a memory issue has been fixed that has caused issues in previous game worlds (time freeze).

- New system also features better debugging and testing abilities for me to make test runs of certain route combinations without the need to actually create the airlines and routes in game interface.

- Frequency limits for routes have been adjusted. The system includes now two separate checker mechanisms for this; another factors the plane size vs. route distance, and other factor is route frequency vs. route demand. In short, players are expected to use the appropriate sized aircraft equipment for routes (usually meaning a bit larger planes than in the past game worlds). Information to user interface about these will be built later.

- Minimum intervals for flights are also adjusted (= the time needed between two flights on same route by same airline). Again, if too close, they are counted (nearly) as a single route. 60 minutes is always recommended, but goes down linearily on busy routes.

- Technical stopovers carry about 20% penalty vs. non-techstop flights (each stopover adds another ~20% penalty). (increased from previous)

- Flight delay / cancellation and route oversupply calculations have also been technically modified, and made much simpler than in the past. (allows to develop the delay/cancellation module more easily in the future)

- RI plays a larger role now than in the past. It still grows like before from 0 to 100 and stays there, but effects of low RI have been increased. (RI will, perhaps, become a fluctuating number in the future instead of static once reaching 100). It is advisable to start new route with smaller equipment and less frequency, rather than blast the full demand right away.

- As before, the route passenger demand increases if there is a lot of seat supply, or low average prices offered by airlines on the route. And like before if there is no supply in let's say C class the pax demand will move to Y/F class if there are available seats there. (nothing new here, just adjusted from current systems). Passengers will also choose not to fly if the airlines offer too high prices or otherwise a poor product (= demand is not static)

- Price effect is more noticeable and makes greater impact to route sales. It is also more safe to increase prices over the suggested default level since there is not so steep "cliff" anymore on the pax travel desire with increased prices as before.

- Of the ~12 factors affecting passenger distribution the following are currently modelled in this test version: Prices, flight frequency, techstops, aircraft size.

- The following are not yet modelled, but will be added before the final release version of the system: dep/arrival times, route flight duration, aircraft type/condition bonus, company image effect, alliance effect, onboard service/seat quality, and few other minor factors.

[michael]

Perhaps open the gates and flood the used market so we can pick up some planes?

[LemonButt]

So I jumped right in and decided to jump right to the frequency oversupply/frequency problem.  The clock hasn't started yet, but once all my aircraft are delivered I will be flying 38x daily from O'Hare to LaGuardia with 737s (5100 total seats avail daily).  I thought I was being slick and overloading on frequency and capacity, which according to the game I should be.  However, I was curious as to the actual frequency in real life as the demand is ~4000 pax/day.  So I did some research and in the real world there are 35 daily flights!

Here is the breakdown of aircraft used:
14x MD83
11x E170
6x A320
2x B752
1x A319
1x B739

So in the real world, the largest aircraft is a ~190 seat 757-200 (or B739) and the smallest is a ~75 seat E170.  The average seats sold per flight, based on 4000 pax/daily, is 114 pax/plane.  My average seats per flight is 136 which means I should have a ~84% load factor assuming no competition.

You stated that we should test the pax distribution system in a normal way first and then try to do something silly with oversupply/frequency compared to what is reasonable.  At this point, I'm kind of confused as to whether what I did is "the normal way" or "the silly way" since it mirrors the real world.  Can you further elaborate as to how the new algorithm is supposed to work so we can attempt to "break" it?

[schro]

Nothing left in the used market.... so.. placed an order for a bunch of new jets...

[AAL558]

Yeah same here. I think it'd work out better if you flood the markets with a few decent domestic aircraft (A320 family /B737 family). Otherwise most of this test is going to be spent waiting for new aircraft deliveries.

[tacsniper]

Same here... nothing left in used market... just ordered 10 B737... but have to wait 3 months (game time) before first delivery... perhaps move the new jet delivery much faster for the first few days so we can get as many jets as we can with that $100M?

[libertyairlines]

Same here... nothing left in used market... just ordered 10 B737... but have to wait 3 months (game time) before first delivery... perhaps move the new jet delivery much faster for the first few days so we can get as many jets as we can with that $100M?

Agreed, just put 6 A320 series aircraft on order.

[markj23]

The test world will move quick enough with the 10 minute game days that i dont think there is any need to flood the market or speed up new deliveries

In the other quick beta a couple of months ago it didn't take me long to get a good fleet going.

[Sami]

flying 38x daily from O'Hare to LaGuardia with 737s (5100 total seats avail daily).  I thought I was being slick and overloading on frequency and capacity, which according to the game I should be.  However, I was curious as to the actual frequency in real life as the demand is ~4000 pax/day.  So I did some research and in the real world there are 35 daily flights!

So in the real world, the largest aircraft is a ~190 seat 757-200 (or B739) and the smallest is a ~75 seat E170.  The average seats sold per flight, based on 4000 pax/daily, is 114 pax/plane.  My average seats per flight is 136 which means I should have a ~84% load factor assuming no competition.

For this route, I bet the 35x daily flights are not supplied by a single airline. That's the main point here...

I pre-checked ORD-LGA, and with present settings the "minimum optimal" plane size for that route (demand 3300pax/day) is a 83-seater (below that you will get a penalty for passengers since they don't like to fly any smaller plane for that distance). And another factor; after 11 daily frequencies (which is considered the "maxmimum optimal frequency" for this route) you will see noticeable drop too.


Used market is also refreshing normally now, that setting was off...

[LostInBKK]

Happy to try the Beta but struggling with the flu at the moment 

[NorgeFly]

I've joined also and loaded leased eight 737-300/400 aircraft. Initially I've loaded 16 daily flights from Oslo to Bergen with 737-400s which meets demand quite well in game. I've also overloaded the route between Oslo and Kristiansand by around 50% and underspplied Oslo to Bodo to see how the routes perform in comparison.

Similar to another player above, I have compared the AWS demand figures with the real world schedule (this is particularly easy for Norway using the avinor website).

In September 2012 (I ignored peak summer season as frequency reduces greatly over summer as its a business route mainly) there are 28 daily flights on a 737 (Norwegian and SAS) between Oslo and Bergen, plus 3 Dash 8 flights (these are multi-stop flights however). Even though this game is 2004 vs today, it seems like true demand is much greater than the AWS model suggests. The same can be said for other domestic routes from Oslo:

September 3rd 2012 (Mon)
Stavanger 28x daily 737
Trondheim 30x daily 737
Bodo 11x daily 737
Tromso 14x daily 737
Kristiansand 9x daily 737

The list goes on... The domestic market is Norway is huge.

I'd be happy to put together a full list of up to date stats for the whole of Norway if it would be helpful/used.

[LemonButt]

For this route, I bet the 35x daily flights are not supplied by a single airline. That's the main point here...

I pre-checked ORD-LGA, and with present settings the "minimum optimal" plane size for that route (demand 3300pax/day) is a 83-seater (below that you will get a penalty for passengers since they don't like to fly any smaller plane for that distance). And another factor; after 11 daily frequencies (which is considered the "maxmimum optimal frequency" for this route) you will see noticeable drop too.


Used market is also refreshing normally now, that setting was off...

This opens up a huge can of worms:

1. 35x daily are supplied by several airlines and not just one, but the only reason this is true is because of real hubs in ORD and LGA creating this need.  Using AWS's current static demand model, this is not taken into consideration.  This means if you have no competition, you would be forced to fly 11x daily with 300+ seat aircraft to meet demand.  Once the competition shows up, you'll be forced to switch to 150 seaters flying 11x daily to avoid going broke.  This doesn't make any sense.  The obvious way to counter this in my mind would be revenue management where you could fly 11x daily with 150 seaters and overwhelming demand where you can make just as much money as a 300 seater.

2. Is the minimum optimal plane size handled as discrete or continuous data?  I went back to verify and that E170 is actually an E175.  An E175 can have 78 seats in 2-class config or 88 seats in a 1-class config.  Does configuring for less seats mean you get a penalty (minimum is 83 seats) or how is this data handled?

3.  What is the minimum optimal plane size calculated using?  Just flight distance or does demand factor in?  A CRJ200 only has 50 seats, but there are many flights IRL that are considerably longer than the ORD-LGA flight (635nm): http://flightaware.com/live/aircrafttype/CRJ2?;offset=0;sort=DESC;order=filed_ete  The longest flight with a CRJ200 in the air as I type this is from Des Moines to Phoenix, which is a 1200nm flight with a 50 seater!  You said pax don't want to fly a smaller plane for "that distance".  With a jet, whether it be a CRJ or E-jet, the flight is only 1:45.  I guess my big concern here is that it's already very difficult to run a regional airline, so penalizing them for flying CRJs etc long distances is counterproductive, especially since the number of potential flights for a regional is very limited outside of the US/Europe where things are close together.

4.  My understanding of the frequency issue are the transatlantic 757 flights and other "long range narrow body aircraft", which is why people want ETOPs etc. to combat abuse.  I think flights under 500nm in length and over 2000nm should be the only ones subject to ideal frequency limits (or some other arbitrary limit).  The frequency problem is not really in issue in the 500-2000nm range IMO.  The under 500nm routes with heavy demand, such as Taipei to Hong Kong are subject to abuse as well as the trans-atlantic routes.  I initially thought the upper limit should be 2500nm, but then I remembered I used to fly 24x daily with MD95 between SFO and JFK, which is part of the abuse we're trying to end   Other restrictions could be based on domestic/international flights (for US at least) and jet/prop aircraft.  I can fly a Dash-8 from Chicago to Los Angeles since the range is 1510nm, but it would take 5.5 hours since it is a prop, but making that flight with a CRJ200LR will get there in 4.5 hours.  However, the Dash-8 would be preferred because it is a "bigger" plane based on number of seats and even though the CRJ would be preferred by passengers IRL, the Dash-8 will receive preference in AWS.

5. Once city-based demand is implemented, routes like ORD-LGA will likely get bigger--not smaller, since all the demand from LGA/EWR/JFK would be/should be lumped in together as will the demand from ORD/MDW.  This is probably going to be around 10,000 pax/day I'm guessing and 11x daily isn't going to cut it.

6. Can you provide a sample matrix on minimum plane size, maximum frequency, and how demand plays into this to give us an idea of how it works?

[Sami]

- The first message of the thread has been updated; contains information about what's changed.

- Time starts to tick now.

- If you are wondering why some route has certain sales, please post a question here, with the appropriate route details. Do note that RI effect is quite high at the beginning - 0 route image combined with low demand and multiple daily flights is not profitable, but once RI reaches 100 you will attain full sales.

- Note (from the first message) that parts of the values affecting demand distribution are still missing (like effect of route dep / arr times). These will be added later on.

[Frost33]

Hi sami, I am based out of Baku (UBBB) and as the first day changed the first l/f results came through. They seem far too low in my opinion, I operate one flight a day to each of four destinations from Baku (one of which detailed in the attachment) and even though the aircraft nearly/only just covers the advertised requirement, I am only getting around 5% l/f.
Regards, Gianni

[ARASKA]

Hi sami, I am based out of Baku (UBBB) and as the first day changed the first l/f results came through. They seem far too low in my opinion, I operate one flight a day to each of four destinations from Baku (one of which detailed in the attachment) and even though the aircraft nearly/only just covers the advertised requirement, I am only getting around 5% l/f.
Regards, Gianni

Having the same problem. My Miami-Aruba route has gotten the same LF for 4 days straight, (10.8%). Also, my routes that I fly once a day are getting LFs of around 15%.

[libertyairlines]

Having the same problem. My Miami-Aruba route has gotten the same LF for 4 days straight, (10.8%). Also, my routes that I fly once a day are getting LFs of around 15%.

Agreed, every flight is hardly pulling any load factor at all and also isn't showing up in the income statement as it should. Two of my flight are showing prose at 70% LF but only after advertising from the beginning.

[brique]

Having the same problem. My Miami-Aruba route has gotten the same LF for 4 days straight, (10.8%). Also, my routes that I fly once a day are getting LFs of around 15%.

In sami's updated first post today:

' - RI plays a larger role now than in the past. It still grows like before from 0 to 100 and stays there, but effects of low RI have been increased. (RI will, perhaps, become a fluctuating number in the future instead of static once reaching 100). It is advisable to start new route with smaller equipment and less frequency, rather than blast the full demand right away.'

I would think that this more pronounced RI-related effect is being rather severe in the early stage?

[ARASKA]

In sami's updated first post today:

' - RI plays a larger role now than in the past. It still grows like before from 0 to 100 and stays there, but effects of low RI have been increased. (RI will, perhaps, become a fluctuating number in the future instead of static once reaching 100). It is advisable to start new route with smaller equipment and less frequency, rather than blast the full demand right away.'

I would think that this more pronounced RI-related effect is being rather severe in the early stage?

Then how come two airlines have 100% LFs since the first day?

[Frost33]

exactly what I was going to ask.

[01miki10]

   - (Note2; aircraft purchasing limits for first turn and weekly limits are removed for testing purposes)

If the purchasing limits are removed, how can I get this message: