Please spend additional resources on AI (An open letter to Pdox)

[Eddiit]

Dear Paradox,

If you're reading this I want to say firstly how excited I am that you're going into the world of Space 4x. Many of us are still waiting for a worthy successor to MOO2 and based on what I've read so far and the games you've made in the past I think you have shot at doing it. I'm pulling for you!

However, one of the biggest issues with the 4x genre in general (and some pdox games) is a lack of good AI. I know AI isn't a SEXY way to spend time and resources but it's necessary to really make this game great. I'm writing you to implore you, beg you to please bring in a specialized AI team and spend resources on making a top tier AI and then continue to add and adjust to it as time goes on via patches.

I understand that this is a tall order but believe me, if you make a game that is as open ended as you hope, with an AI that can surprise us at times you'll go down in history with MOO, MOO2 and others. I believe you can do this, but you'll have to be aggressive and do MORE than is standard.

In conclusion, thank you for making so many great games and thank you for going into space the 4X genre but please please please give AI the time, attention and resources it so badly needs. Even if it costs a bit more I'm sure the community would gladly support it and you'll go down a legend.

Thank you,

Ian Coyne
A very long time Pdox player and forum lurker

 

[Zupanicarr]

I really want it so that there'll be an option where the AI expands really quickly, and is really aggressive, without giving them too much of an artificial bonus. I understand it's needed for some things, and I also understand that, Wiz for example, tries to create a fine line for the AI between being tough, while also being natural and not too gamey, but I do hope it's aggressive enough to give challenge to even a veteran player.
I just want to have the potential of losing on even terms, something pretty much mostly Paradox games, and few others have ever done for me, but space games tend to have this issue a lot.
You usually either face off against foes that are miles behind you, or miles ahead of you, thankfully this not having tactical combat, or a rock-paper-scissors system should balance it out a bit.
I have faith Paradox can create some tough AI, but it might be a challenge in this new territory.

 

[Sindai]

In a symmetric strategy game the AI is not going to be a challenge for veteran players without a lot of cheating. Accept it now to avoid disappointment later.

On the bright side, Paradox is aware of this. The late game mega disasters exist to shake things up at the stage where the player is usually in no danger of losing to the starting empires.

 

[Eddiit]

I think the issue comes down to the fact that AI isn't a sexy sellable feature even though it's arguably one of the most important parts.

 

[Hans_Schnitzel]

I think the issue comes down to the fact that AI isn't a sexy sellable feature even though it's arguably one of the most important parts.

And a good AI is probably pretty hard to make.

 

[Axe99]

And a good AI is probably pretty hard to make.

Not just that, but that AI is pretty resource intensive (at least, as far as I understand, the way we understand and code AI at the moment).

That said, a big +1 for me for better AI. Better AI really builds on the longevity of games - if there's always a range of sensible things the AI can do, and it knows how to play the game well, then it makes the game far more enjoyable over the long run.

On that front, every Paradox game seems better than the next (EU4 was better than CK2 was better than Vicky 2 was better than HoI3 and so on - although due in no small part to ongoing development, I wouldn't say there's a big gap between EU4 and CK2) so fingers crossedStellaris will benefit from their ongoing development in this area.

 

[FirstCitizen80]

I kind of like the way Distant Worlds did it, although there is room for improvement. Open up the AI's decision making process, and allow the player to fine tune it....or better yet have it tied to racial characteristics and let the player choose those for his opponents.

 

[King Travis]

AI should be #1, then everything else. The problem is its hard to program things that are interesting and then have the AI use it correctly.

 

[Careful Plum]

I kind of like the way Distant Worlds did it, although there is room for improvement. Open up the AI's decision making process, and allow the player to fine tune it....or better yet have it tied to racial characteristics and let the player choose those for his opponents.

Non-moddable AI behavior is one of those things that plagque every Paradox game. It would be really great if we had more control over what the AI does! Right now, what we can do ourselves is pretty much limited on deciding which button on an event the AI presses.

So, make the AI better, but, somehow, also more moddable!

 

[stilgarpl]

The thing is... if you make the AI too good, the players will be tired of the game before they even start playing it. Even if the AI isn't cheating, if you lose 5 games in a row and you can't seem to find any mistakes you've made... it seems unfair and the game doesn't look fun. If the AI is too weak, the game isn't fun either. It's hard to balance, and even if you have cheating AI, you either can find a weak spot and exploit it or it's cheating so much that you can't do anything.

But, imagine some developer advertised the game, that the AI is so good, you have no chance of winning. And it's not a cheating AI. Would you buy it? Maybe, but most of the players would think of the game as "too hard". Players want a challenge, but they still want to win in the end.

 

[Careful Plum]

The thing is... if you make the AI too good, the players will be tired of the game before they even start playing it. Even if the AI isn't cheating, if you lose 5 games in a row and you can't seem to find any mistakes you've made... it seems unfair and the game doesn't look fun. If the AI is too weak, the game isn't fun either. It's hard to balance, and even if you have cheating AI, you either can find a weak spot and exploit it or it's cheating so much that you can't do anything.

But, imagine some developer advertised the game, that the AI is so good, you have no chance of winning. And it's not a cheating AI. Would you buy it? Maybe, but most of the players would think of the game as "too hard". Players want a challenge, but they still want to win in the end.

I literally can't think of a strategy game where this was true. Also, it is much, much easier to make an AI worse once it is good (for easy/medium difficulties) than actually making a good AI.

It's different for action games, where it's easy to make a perfect AI, but difficult to make a not-perfect-yet-believable AI.

 

[stilgarpl]

where this was true

Where what was true? That people don't buy games that are too hard? People are saying this about CK2, and we all know how "hard" CK2 is...

 

[Careful Plum]

Where what was true? That people don't buy games that are too hard? People are saying this about CK2, and we all know how "hard" CK2 is...

Where the AI is so good it's almost impossible to win without knowing the game very well.

CK2's AI is terrible (it's much worse than EU4 AI, IMO). What is perhaps difficult about the game is not its AI, but that fact that its mechanics are sometimes quite different from other games.

 

[stilgarpl]

Where the AI is so good it's almost impossible to win without knowing the game very well.

Do chess count?

 

[Careful Plum]

Do chess count?

Yes.



So, the obvious solution is for Paradox to create a giant database of every possible game of Stellaris and the optimal thing to do in every possible situation.

The game is, after all, technically turn based. And even with the addition of randomness, there will still always be that one move where the expected benefit is greatest .

 

[Thure]

Yes.



So, the obvious solution is for Paradox to create a giant database of every possible game of Stellaris and the optimal thing to do in every possible situation.

The game is, after all, technically turn based. And even with the addition of randomness, there will still always be that one move where the expected benefit is greatest .

Technically it's not turn based. ONly you call it a turn if all players act at the same turn and can change their moves

 

[Careful Plum]

Technically it's not turn based. ONly you call it a turn if all players act at the same turn and can change their moves

Well, ther are plenty of modern board games where players take their turns at the same time. Alsco CivV in MP. Players can't change their moves once the turn is finished, and before that I'd call it calculating what the consequences of a possible move would be, not actually taking it .

 

[stilgarpl]

Technically it's not turn based. ONly you call it a turn if all players act at the same turn and can change their moves

Please, let's not have this discussion again It leads to nowhere

 

[TheRedBat]

Doomdark said in the SpaceGameJunkies Q&A that the AI will be comparable to EU IV.

 

[MWSampson]

That means it will have plenty to be desired, but will be actually not terrible.

The EUIV AI is blocked by the fact it doesn't "play to win" and tends not to expand past its desired province list.