I have a challenge for any AI maker

[xarick]

<Samu> I have a challenge for any AI maker! Make your AI profitable unbankrouptable with aircraft
- and infrastructure maintenance costs enabled
- running and construction costs set to high
- breakdowns reduced
- airplane crashes normal
- loan can be £500k
- inflation off, because inflation is just bad
- starting year 1950 and must build air routes in the starting year
- ending year 2050
- airports and vehicles expire
- no other transport type, only aircraft, any kind of aircraft
- no station spreading
- use default plane speed factor, 1/4
EDIT:
Settings with value different that default:

screenshot#1.png (41.7 KiB) Viewed 4567 times

World generation settings:

screenshot#2.png (79.02 KiB) Viewed 4567 times

EDIT 2:
Or just use this config:

openttd.cfg

(10.92 KiB) Downloaded 151 times

[odisseus]

I believe that's impossible for a human player, let alone an AI.

The three aircraft models that are available before year 1955 are just too small and too slow. It is very hard to run them at a profit even without maintenance costs.

The player could repay the loan and wait till FFP Dart becomes available, but the real problems are yet to begin.

In order for the city airport to stand a chance of being profitable, it must be surrounded by buildings that generate a lot of passengers. This can happen only when a town with at least 4000-5000 population has an airport-sized gap in the middle. Such towns get generated, but not too often.

Now, the player would need to find not one, but two such towns at exactly the right distance. The two airports must be as far apart as possible in order to maximize profit per trip, yet close enough so that the company doesn't go bankrupt before the first plane arrives. After the airports are placed, the player should immediately buy as many FFP Darts as possible, because that would be his last chance to buy anything.

If the player manages to achieve all this and buy at least 7-10 planes, then I'd say the company has a chance to survive, but that's still not certain.

In short, this is impossible except under the most favourable conditions, and even then probably impossible as well.

[perverted monkey]

Indeed, only Darts may have a chance.
1/4 speed factor is out of question. No way.

[LaChupacabra]

I believe that's impossible for a human player, let alone an AI.
...
In short, this is impossible except under the most favourable conditions, and even then probably impossible as well.

Have you checked?

You dramatize It realy doesn't take a lot here for a company to generate profits.
For the average player this is banal. Unfortunately, AI are currently hopeless in air transport, mainly because they do not take into account the cost of maintance and are building airports like crazy in every cities. Some AI are even trying to build airports on every... field. Typical of almost all AI is that they usually have more airports than planes.

In my opinion, this "challenge" is not pointless or faulty.

[jfs]

Have you checked?

You dramatize It realy doesn't take a lot here for a company to generate profits.
For the average player this is banal. Unfortunately, AI are currently hopeless in air transport, mainly because they do not take into account the cost of maintance and are building airports like crazy in every cities. Some AI are even trying to build airports on every... field. Typical of almost all AI is that they usually have more airports than planes.

In my opinion, this "challenge" is not pointless or faulty.

If you claim it's so easy then prove it. Write an AI that takes those additional factors into account and can reliably turn a profit under the conditions outlined in the first post.

[RailwAI]

I believe that's impossible for a human player, let alone an AI.

I saw the challenge and did two attempts myself (as human). First time I succeeded and gained money. The second time was worse, as I wanted to see what happened when I would make bad choices. Currently I have other topics higher on my list for AI development than supporting aircraft, but it's definitely worth a try.

[_dp_]

Tried it for a bit and for human it's not that hard to stay afloat. Can't be bothered to code it but shouldn't be that hard either IMO. May not be very useful for a general AI though.

[Yexo]

In my first 2 attemps doing this as human player I failed miserably. First game I played was on a 256x256 map and built 3 airports with 7 aircraft. 2nd game I built 4 airports to quickly and went bankrupt, could only afford 6 aircraft or so. Final game I made a very easy profit by building only 2 airports far apart from eachother (on a 512x512 map) in reasonably large cities and building 8 aircraft. I let the game fastforward a few years without doing anything else. In year 1953 I had 360k aircraft income vs 118k property maintenance.

I don't think any of the other settings apart from enabling infrastructure maintenance really matters for this. Both running and constructions costs are so low in comparison they barely register. Obviously NewGRFs can change this massively and swing it in another direction.
Any airport must have many aircraft, probably 4 or 5 minimum (that can take a full load away without having to wait too long). I haven't tested how much the distance between the airports matters.

With infrastructure maintenance disabled aircraft are almost always immensively profitable, so I'm not surprised that any AIs that don't take infrastructure maintenance into account fail on this by default. Then again, aircraft are fairly uninteresting from an AI writer perspective since there isn't very much to do / differentiate from other AIs, but it seems like it isn't too difficult to make a profit on this challenge, after updating an AI to take the maintenance cost into account and thus be aggressive with the amount of aircraft for each airport. (which in turn might mean to be way less aggressive with building airports in the first place)