Game needs opponents

[mothyposh]

The game would be better if there were rival companies to compete against IMHO.

[MarkShot9]

Yes, but coding a credible AI (even one that cheats is a major undertaking), I would say it would double the code base of the existing game.

Further, this game was obviously not conceived as more than a sandbox.

To go back and add such at this late stage is really horrific from a coding point of view.

Assuming the developers/players really saw this as the next logical step, it really should be the start of a fresh development cycle.

I played RT2/RT3, and the challenge was more the scenario than the AI competitors. I haven't seen a decent rendition of Tycoon AI, since Sid's RT1. (and it clearly cheated, but was fun)

[Brulikrax]

We do not need AI competition. Go play Rail Empire instead.

[genemead]

(It's"Railway Empire").

The problem with AI is it's either too easy, or too hard. AI's don't learn, and they usually cheat.
The only AI that's a "worthy" opponent is another human, and the best AI to play against is.... yourself!

[theSeb]

Nope. It does not.

[Starbucaneer]

This is not designed to be that kind of game. There are other games you can play if you need an opponent of some kind ;-)

[SoftwareSimian]

It really doesn't.

[cheese-water]

I agree. I came to this game from Chris Sawyer's Locomotion from 2004. The other companies had... interesting ideas about how to build railroads, but they could still be a formidable opponent.

This game has a lot of benefits over CSL, but I'm getting pretty tired of "free" games because they lack stakes (don't get me wrong, I do like a low-stakes game mode where I can 'try stuff' without getting steamrolled) and the Campaign doesn't offer much room in the way of creative freedom when you're always being told step-by-step what to do, and those steps usually involve hemorrhaging money, so it's more like "Simon Says" than a tycoon game. If there's some game mode I don't know about, let me know.

What I really want is... what Locomotion does. Scenarios with a goal, which can be achieved in whatever way the player sees fit, and which may or may not also involve competing companies. I've been finding that "play until you get bored" kinds of games haven't been hitting the spot for me lately, so having some kind of continuing opposition would really help keep the interest up. As it is right now though, once you've paid off your starting loan and are making a net profit, all the stakes are gone, and there's nothing that can be done about that other than deliberately sabotage your own company or exit the game and start up Locomotion.

[Lord Bludgeon]

This AI cheats. Even with no competitors. I can reference the Swiss campaign scenario... Most of the stations are combination cargo/pax terminals... and you aren't allowed to fix this, so you only have 1 platform to run your cargo. If you have multiple lines running through that station (and you need to) the Farm (for instance) will assign the lion's share of the grain produced to the least productive line at the platform and none to the most productive. Basically, you have a train with 12 cars (96 ton capacity) waiting at a station, loading 1 ton at a time as the farm produces it, while there are 300 tons of grain sitting at the same platform waiting for a train with 5 cars that can't get to the station until the 12 cars are loaded. SO, I cancelled the 5-car line AND 300 TONS OF GRAIN JUST VANISHES!

I'm done.

[phobos2077]

For examples of what AI can be look at OpenTTD. There are people who did actual scientific papers on making AI for that game. But TTD has this simplistic voxel-based world. Now consider the challenge involved in making a formidable AI opponent in a sandbox game with floating-point coordinate system.. I dunno, It is definitely possible. But it's not only a programming challenge but also incredible game design challenge. And based on the state of the current game we know that good game design is not the strongest skill of this particular developer.

I only wish they would expand on the modding API to allow for AI opponents.. I think I saw a function to create new player. The game already supports multi-player of sorts (not in a traditional sense, but in terms of player as a in-game entity - one that has it's own bank account and assets). Also there are functions to build stuff, buy stuff, create lines etc. And the full read-only access to ECS world. So maybe AI IS already doable. Just that nobody has yet stepped up to do it.

My suggestions for developers would be to make a sample dummy AI mod to showcase that it's possible and advertise it somehow. If they are lucky, the game might attract some great minds. A lot of people are now learning AI and research, and games are great medium for them. I can easily imagine someone making a deep-learning based AI that will maximize profits or town growth...

[Tsubame ⭐]

The problem is OpenTTD is orders of magnitude simpler than Transport Fever 2. You can only build in four directions, and even tunnels and bridges are very very simplified, among other things.

Also, OpenTTD is open source which can be freely modified. This is not the case for Transport Fever 2. The developers will not work on something that they think is not going to generate enough return of investment to them.

[phobos2077]

Originally posted by Tsubame:

The problem is OpenTTD is orders of magnitude simpler than Transport Fever 2. You can only build in four directions, and even tunnels and bridges are very very simplified, among other things.

This adds exactly nothing to the arguments provided in my previous message.

Originally posted by Tsubame:

Also, OpenTTD is open source which can be freely modified. This is not the case for Transport Fever 2. The developers will not work on something that they think is not going to generate enough return of investment to them.

The developers already invested in the modding functionality and the game engine should already allow to make AI by third party. I don't see what was the point you were trying to make.

[Tsubame ⭐]

Originally posted by phobos2077:

Originally posted by Tsubame:

The problem is OpenTTD is orders of magnitude simpler than Transport Fever 2. You can only build in four directions, and even tunnels and bridges are very very simplified, among other things.

This adds exactly nothing to the arguments provided in my previous message.

Originally posted by Tsubame:

Also, OpenTTD is open source which can be freely modified. This is not the case for Transport Fever 2. The developers will not work on something that they think is not going to generate enough return of investment to them.

The developers already invested in the modding functionality and the game engine should already allow to make AI by third party. I don't see what was the point you were trying to make.

TpF 2 game is moddable, but to a limited degree only. AI, and multiplayer as an extension, would have to be added by the developers, or at least, they would have to allow us to fiddle with the game code to do so. This would not be a small task.

The point is that there is likely not enough financial demand - new customers - to justify their time and money to add this feature to the game. The developing team is small, and they think that their limited resources are better spent doing something else.

This has been discussed ad nauseum in the forums here and in TpF 1 as well.

[Miguel Baptista]

Originally posted by phobos2077:

For examples of what AI can be look at OpenTTD. There are people who did actual scientific papers on making AI for that game. But TTD has this simplistic voxel-based world. Now consider the challenge involved in making a formidable AI opponent in a sandbox game with floating-point coordinate system.. I dunno, It is definitely possible. But it's not only a programming challenge but also incredible game design challenge. And based on the state of the current game we know that good game design is not the strongest skill of this particular developer.

It's a challenge, but as you said, perfectly doable.
I mean, the route choice, even with floating-point coordinates, has two very important variables that are hard-choiced by AI strattegy, which is starting and end point.

Given that, and the already available information to the player about route costs, you just have to make the code compute which route is cheapest between two set points; and given a percentage of AI toughness, reduce the error-margin to that value the harder the AI is.
This is just one very rough logical and mathematically sound solution to the problem. But you could also add in other factors, like elevation changes/mountain delimitations/tunneling/bridging etc.

As for the rest, I agree 100%.

[Zapp]

Originally posted by mothyposh:

The game would be better if there were rival companies to compete against IMHO.

Counterpoint:

The game would be worse if there were rival companies to compete against IMHO.