AirwaySim
Online Airline Management Simulation
Login
Username
Password
 
or login using:
 
My Account
Username:
E-mail:
Edit account
» Achievements
» Logout
Game Credits
Credit balance: 0 Cr
Buy credits
» Credit history
» Credits FAQ

Author Topic: [ok] Test World: Pax distribution  (Read 2684 times)

Offline JumboShrimp

  • Members
  • Posts: 6323
[ok] Test World: Pax distribution
« on: July 04, 2012, 11:57:01 PM »
First, this is not a bug, just analysis of a pax distribution on a trans atlantic route CDG - BOS that is ~3000nm (2986nm exactly).

All the data is in the attached spreadsheet snapshot.

There are 2 issues i can think of:

Looking at the data, especially Y pax, the ideal aircraft appears to be between 753 and 752.  This is either between 237 and 189 configured or between 279 and 238 Max HD Capacity.  The formulas discussed in feature request yield either either ~240 or ~250 pax.  Both are between 753 and 752, which means the system is using Max HD capacity to determine the ideal aircraft capacity.  This brings up

Issue 1:  If Max HD aircraft capacity is used to determine ideal aircraft, formula should change to increase the ideal aircraft by ~15 to 20%  something like 75 + distance / 12.

A better solution would be to for system to use configured capacity rather than Max HD capacity.

Leaving it as is would not change te passenger distribution model very much, and 757 will continue to reign supreme flying trans-atlantic.

Issue 2: Aircraft that is smaller than ideal capacity (A321) should never have higher LF than larger aircraft that is still smaller than ideal aircraft (752).

Example:
B752, 40 pax allocated, LF = 23.0%
A321, 37 pax allocated, LF = 24.3%

It is not a rounding error, because 36 pax for A321 would still have higher LF than 752.
« Last Edit: July 10, 2012, 09:22:15 PM by sami »

Offline alexgv1

  • Members
  • Posts: 2184
Re: Test World: Pax distribution
« Reply #1 on: July 05, 2012, 12:47:51 AM »
Re: issue 2

Isn't the more important thing to look at pax numbers and not LFs as Sami posted somewhere else that it is just a statistical value worked out after pax are allocated. Therefore pax allocation is the reward for selecting the right aircraft not a higher LF. The B752 gets more pax than A321 so surely that is right. I mean seats sold make the money not LFs 50/100 seats is LF 50% 50/250 is 20% all things the same then lower load factor is not a bad thing. 

Hope I explain my point clear I am on phone only ATM.

Ps. agree with issue one that actual config used should be what decides if you are using ideal size aircraft (if it doesn't already).
CEO of South Where Airlines (SWA|WH)

Offline JumboShrimp

  • Members
  • Posts: 6323
Re: Test World: Pax distribution
« Reply #2 on: July 05, 2012, 01:34:01 AM »
Re: issue 2

Isn't the more important thing to look at pax numbers and not LFs as Sami posted somewhere else that it is just a statistical value worked out after pax are allocated. Therefore pax allocation is the reward for selecting the right aircraft not a higher LF. The B752 gets more pax than A321 so surely that is right. I mean seats sold make the money not LFs 50/100 seats is LF 50% 50/250 is 20% all things the same then lower load factor is not a bad thing.  

Hope I explain my point clear I am on phone only ATM.

Couple of points:
752 vs. A321:  752 is a bigger plane, and in my example the A321 has 3 F seats while 752 has none..  So 752 does have 3 more pax allcated, but it should have > 3 extra pax allocated vs. A321.

Re: LF and frequency benefit

The key to frequency benefit is that a seat in small aircraft counts for more than a seat in large aircraft.  A seat in a small aircraft can easily be worth 2x or 3x vs. seat in large aircraft.  The way you see how much more the seat is worth is by looking at LF.

The result of seats of small aircraft being worth more than seats in larger aircraft are all the problems we are having with frequency benefit  (ab)use.

To combat it, the new pax allocation feature is supposed to remove the frequency benefit alltogether when flying smaller than ideal aircraft for the route.  The results show that the benefit is not removed completely.  Something is still there helping smaller aircraft (A321) gain disporportionate LF.

Quiz, 1 out of 4 of these flights is above breakeven.  Guess which one?

Online Sami

  • Administrator
  • Members
  • Posts: 15098
    • AirwaySim - Are you the next Richard Branson?
Re: Test World: Pax distribution
« Reply #3 on: July 05, 2012, 04:56:55 AM »
Issue 2: Aircraft that is smaller than ideal capacity (A321) should never have higher LF than larger aircraft that is still smaller than ideal aircraft (752).

load Factor is never a value that is calculated by the system, at any level, during the process.

The system looks at the flight data and determines the SEATS SOLD based on that data. LF is only a statistical value based on that data.

The config used here to determine the plane size factor is the true config of the plane, but I believe it has to be changed to for example "average size of the fleet group", making A318 and A321 the same in this sense, as they are really the same in passenger's point.

Offline JumboShrimp

  • Members
  • Posts: 6323
Re: Test World: Pax distribution
« Reply #4 on: July 05, 2012, 08:14:03 AM »
load Factor is never a value that is calculated by the system, at any level, during the process.

The system looks at the flight data and determines the SEATS SOLD based on that data. LF is only a statistical value based on that data.

The config used here to determine the plane size factor is the true config of the plane, but I believe it has to be changed to for example "average size of the fleet group", making A318 and A321 the same in this sense, as they are really the same in passenger's point.

I understand that LF is more a by-product of other calculations rather than something that is part of the calculations.  But LF is a proxy of the frequency benefit (which the new passenger allocation aims to to reduce)

In my opinion, there are 2 ways to check the allocation results:
1. Aircraft of ideal size and larger than ideal size will have the same number of pax allocated.  Frequency benefit applies, since constant passenger allocation on ever larger aircraft results in shrinking LFs.  In the example above, looking at constant Y pax of A332 and 753 shows that the allocation is working properly.

2. Aircraft of ideal size and smaller will have constant LF.  Meaning, there is no additional frequency benefit when flying below ideal aircraft size..  A seat is a seat.  Every Y seat on aircraft smaller than ideal should have the same LF.

As far as this case #2, it is not happeneing.  Seats of smaller and smaller aircraft are ending up with higher and higher LFs.  I can only speculate why this is happening.  It could be one of 2:
- ideal aircraft for this route is smaller than 752 (definitely should not be the case)
- something is wrong with the allocation, or something else is happening that I have no knowledge of.

The case #2 working correctly is the teeth behind the lowering the frequency benefit.  If it does not work, or it is not aggressive enougjh (the size of ideal aircraft is too small), all the extra work will yield only modest benefits.

The answer to my quiz, which of the 4 flights was making money (at RI of ~11):  It was A321.  The least appropriate aircraft was the only one making money...

Offline shaolin

  • Members
  • Posts: 77
Re: Test World: Pax distribution
« Reply #5 on: July 05, 2012, 10:43:49 AM »
Hi,

I'd like to just let you know that most of my 42 routes show a 100% LF, yet they are run with different ACs.

http://www.airwaysim.com/game/Info/Airline/88/#AirlineInfo

Offline JumboShrimp

  • Members
  • Posts: 6323
Re: Test World: Pax distribution
« Reply #6 on: July 05, 2012, 07:52:31 PM »
Update of CDG-BOS route.

RI now at 70.  I added tech stopped A321 to test the tech stop penalty.

More thoughts - more like a re-iteration of previous concerns:
- since 300 pax A332 and 237 pax 753 get the same number of pax allocated, it means that the "ideal aircraft" is less than equal to 237.  (should be really be higher in 2005)
- aircraft smaller than "ideal" aircraft continue to gain advantage on ideal aircraft.
- A321 has > 5% LF advantage over the "ideal" aircraft (753).   That means player flying A321s on 2986nm routes can drive player flying 753 out of business.
- Exceeding the "optimal" frequency" is not sufficient measure to stop it.  If there are 4 players flying the route, each supplying 50% (200% total).  3 players fly 753, 1 uses A321.  The A321 player will win
- if one of the objectives of this change was to reduce the advantage of 757 on typical transatlantic routes, it is not happening.  Not only that, 753 < 752 < A321 < 739ER.  I don't evem want to think 737-700ER.  So forget 767, A300, A330, 777, 747, A380 on transatlatic routes.
- while the frequency advantage is be lower in the 1.3x vs. current 1.3, it is still there, If the advantage of using smaller than appropriate aircraft is still there (even though smaller), the incentive remains the same: fly smaller aircraft.
« Last Edit: July 05, 2012, 10:54:27 PM by JumboShrimp »

Offline JumboShrimp

  • Members
  • Posts: 6323
Re: Test World: Pax distribution
« Reply #7 on: July 05, 2012, 09:41:21 PM »
Here is a spreadsheet of 5 different scenarios for ideal aircraft sizes.  My opinion is that Scenarios 3 and 4 would eliminate 75 to 90% of frequency abuse.  (yellow highlight)

Second highlighted area (green) is showing that when aircraft is larger than ideal, the same number of pax is allocated (to A332 and 753).

That means that 753 has frequency advantage (shown in orange).  Frequency advantage means higher LF.

4th highlighted area (blue) shows that flying smaller than ideal aircraft should not gain any frequency advantage.  No frequency advantge means constant LF, no matter how small the aircraft.

As of now, most experienced AWS players are conditioned to gain advantage through frequency.  The only way to stop it is when there is absolutely nothing to be gained by flying smaller aircraft.  In the current test world, there is something (not much but still something) to be gained by flying smaller aircraft.  (A321 advantage over 753)
« Last Edit: July 05, 2012, 10:13:58 PM by JumboShrimp »

Offline swiftus27

  • Members
  • Posts: 4395
Re: Test World: Pax distribution
« Reply #8 on: July 05, 2012, 11:29:24 PM »
So, let's find another way to prevent planes from not doing things that they are... I.E...  there's a A320 flying from LAX to HNL...  ETOPS much?

Online Sami

  • Administrator
  • Members
  • Posts: 15098
    • AirwaySim - Are you the next Richard Branson?
Re: Test World: Pax distribution
« Reply #9 on: July 06, 2012, 04:10:03 AM »
- aircraft smaller than "ideal" aircraft continue to gain advantage on ideal aircraft.
- A321 has > 5% LF advantage over the "ideal" aircraft (753).   That means player flying A321s on 2986nm routes can drive player flying 753 out of business.

Again, LF is not something that is considered by the calculation. And will not be considered, ever...

You can fly a 50 seater there and have 100% LF :P   (in theory.. but practically that is of course something that should not work)

If you look at the number of seats sold (figure #1), the system works just as it has been intended. Techstop gets least seats, then the smallest plane (321) and the biggest planes sell most seats.


I will also make a few changes to the calculations later today, by adding a factor if the route is dom/intl/longhaul which should improve this (and allowing "longhaul domestics" with narrowbodies) , and also adding a fleet group factor so 752 and 753 will be then equal (which they in pax'es eyes are). And it also makes the A321 worse since it's "comparable size" goes down since it will be similar to 320-319.
« Last Edit: July 06, 2012, 04:21:25 AM by sami »

Offline JumboShrimp

  • Members
  • Posts: 6323
Re: Test World: Pax distribution
« Reply #10 on: July 06, 2012, 06:21:23 AM »
You can fly a 50 seater there and have 100% LF :P   (in theory.. but practically that is of course something that should not work)

I was hoping this would be avoided, that aircraft that is just too small would completely lose frequency boost for its seats.  If a 50 seater could theoretically have 100% LF, each seat of the 50 seater is 2x more valuable than a seat in a 757, that can't even get to 50% LF.

If you look at the number of seats sold (figure #1), the system works just as it has been intended. Techstop gets least seats, then the smallest plane (321) and the biggest planes sell most seats.

Yes, it is improvement over current algorithm, which, on an oversupplied route would gravitate to allocating each aircraft equal pax...

I will also make a few changes to the calculations later today, by adding a factor if the route is dom/intl/longhaul which should improve this (and allowing "longhaul domestics" with narrowbodies) , and also adding a fleet group factor so 752 and 753 will be then equal (which they in pax'es eyes are). And it also makes the A321 worse since it's "comparable size" goes down since it will be similar to 320-319.

I am not sure I understand how pax allocation tied to fleet type > actual aircraft within the fleet.  Conceptually I understand that to pax, the aircraft within fleet look the same.  But how will A321 be worse of by looking smaller to the allocation formula if there is no problem in being smaller? (50 seater might get 100% LF)

Glob-Al

  • Former member
Re: Test World: Pax distribution
« Reply #11 on: July 06, 2012, 06:54:52 AM »
I understand that LF is not part of the calculations but it's still illustrative because it's quite a good proxy of which flights are more profitable (it's also a criteria for scoring alliance points, so people do care about it).

JumboShrimp - it would be interesting to see the data on the running costs of these various aircraft on the CDG > BOS route. If a plane has a higher % of the passengers on a route than it contributes to the total  costs of running the route, that will be a sign of its competitive advantage.

Offline JumboShrimp

  • Members
  • Posts: 6323
Re: Test World: Pax distribution
« Reply #12 on: July 06, 2012, 01:33:25 PM »
I understand that LF is not part of the calculations but it's still illustrative because it's quite a good proxy of which flights are more profitable (it's also a criteria for scoring alliance points, so people do care about it).

JumboShrimp - it would be interesting to see the data on the running costs of these various aircraft on the CDG > BOS route. If a plane has a higher % of the passengers on a route than it contributes to the total  costs of running the route, that will be a sign of its competitive advantage.

Here is profitability of the 4 aircraft on CDG->BOS leg:
A332: 40 142 USD
B753: 39 808 USD
B752: 28 835 USD
A321: 40 251 USD

A321 is the most profitable, and was also the first one to turn profit, when RI was low.  Route is 150% oversupplied.

I agree about LF.  LF is the end result for the player.  It determines player's profitability, and it depends what aircraft player chooses.  So in this case, A321 is a hands down winner on a trans-atlantic flight.  I don't really think this is the outcome we want...

Online Sami

  • Administrator
  • Members
  • Posts: 15098
    • AirwaySim - Are you the next Richard Branson?
Re: Test World: Pax distribution
« Reply #13 on: July 06, 2012, 03:19:06 PM »
http://www.airwaysim.com/forum/index.php/topic,41408.msg224056.html#msg224056


Quick test at local server on the route:

type - sold seats
A321 - 83/10/3
A330 - 133/15/4
752 - 113/13
753 - 113/13


C/F class distribution seems to be on the high side still, may adjust that. But with these sales, the profits are so that A330 wins by 10-15k margin per flight over other 3 which share rather equal profit levels.

Offline schro

  • Members
  • Posts: 3440
Re: Test World: Pax distribution
« Reply #14 on: July 06, 2012, 04:09:55 PM »
http://www.airwaysim.com/forum/index.php/topic,41408.msg224056.html#msg224056


Quick test at local server on the route:

type - sold seats
A321 - 83/10/3
A330 - 133/15/4
752 - 113/13
753 - 113/13


C/F class distribution seems to be on the high side still, may adjust that. But with these sales, the profits are so that A330 wins by 10-15k margin per flight over other 3 which share rather equal profit levels.


But when you factor in the cost of ownership/leasing and staffing, the A321 is slapping the other three with a rather large rainbow trout....

Offline JumboShrimp

  • Members
  • Posts: 6323
Re: Test World: Pax distribution
« Reply #15 on: July 06, 2012, 05:06:32 PM »
http://www.airwaysim.com/forum/index.php/topic,41408.msg224056.html#msg224056

Quick test at local server on the route:

type - sold seats
A321 - 83/10/3
A330 - 133/15/4
752 - 113/13
753 - 113/13

C/F class distribution seems to be on the high side still, may adjust that. But with these sales, the profits are so that A330 wins by 10-15k margin per flight over other 3 which share rather equal profit levels.


I see these results now as well (for one day before one of the 4 aircraft went to C check).  It is definitely better than before.  I am still trying to understand the allocation by fleet concept, if it improves things or not...

Online Sami

  • Administrator
  • Members
  • Posts: 15098
    • AirwaySim - Are you the next Richard Branson?
Re: Test World: Pax distribution
« Reply #16 on: July 06, 2012, 05:14:24 PM »
From pax'es point of view it's the "same aircraft" and then also individual a/c config or variant size is eliminated in calc.

Offline JumboShrimp

  • Members
  • Posts: 6323
Re: Test World: Pax distribution
« Reply #17 on: July 06, 2012, 05:19:36 PM »
But when you factor in the cost of ownership/leasing and staffing, the A321 is slapping the other three with a rather large rainbow trout....

Here is the rest of the figures (including LF and profitability):

A330 - 133/15/4 - Y LF=51% - Profit = 47 207 USD
753 - 113/13 - Y LF=51% - Profit = In C Check, lower than 752
752 - 113/13 - Y LF=64% - Profit = 36 143 USD
A321 - 83/10/3 - Y LF=54% - Profit = 34 523 USD

When staffing and aircraft acquisition is taken into account, you are right, A321 is still winning...

And with the new fleet level allocation, 753 is really sufferning the most, while size-wise, it may be closest to ideal for the distance...

Offline JumboShrimp

  • Members
  • Posts: 6323
Re: Test World: Pax distribution
« Reply #18 on: July 06, 2012, 05:39:15 PM »
From pax'es point of view it's the "same aircraft" and then also individual a/c config or variant size is eliminated in calc.

I am just tring to think about this.  Looking at the figures, if I can get an A319 or 737ER on the route, they would end up with fantastic load factors...

The same way 752 ends up with much better LFs than 753.

I am just trying to think of the implications of allocation by fleet type....
« Last Edit: July 06, 2012, 06:17:22 PM by JumboShrimp »

Online Sami

  • Administrator
  • Members
  • Posts: 15098
    • AirwaySim - Are you the next Richard Branson?
Re: Test World: Pax distribution
« Reply #19 on: July 06, 2012, 06:14:39 PM »
Adjusted a bit to find the proper settings. But does the 757 get too penalized now?

330 - 154
757 - 101
320 - 74


(though I understood that one issue has been too widely spread 757 usage on longhaulers.. another issue is that larger 320 variants and 757 are just too close to each others to program any meaningful formula there.)

 

WARNING! This website is not compatible with the old version of Internet Explorer you are using.

If you are using the latest version please turn OFF the compatibility mode.