Matrix Games Forums

ORIGINAL: Farrels

A different perspective:

Is the Disant Worlds AI not fit for purpose, or are you guys putting so many hours into the game that you have it beat?

IMO AI has been the bane of almost all 4X games since MOO. Space Empires, GalCiv they all have AI's that pose no signigicant challenge when you learn the mechanics of the game.

Its the same as Distant Worlds, the first dozen hours or so the AI can hammer you, but then you notice the deficiencies and then it becomes a cake walk. The ideas in this thread are good, but im sure after another 20 hours or so youd be back at square one as further deficincies in the Ai would become apparant.

Therefore

IMO id like to see Civ style difficulty settings. Anything over Prince and the AI gets significant bonus's, where you get penalties. I know Legends has this but it needs to be expanded upon, with bonus's to fleets, maintainence etc. Although this leads to the AI 'cheating' in a sense, Civ is the only AI that has troubled me over a long period as it levels the playing field. Even though the Ai can still make dumb decisions, it has massive armies, lots of resources etc. This coupled with improvements to ship design could see a more challenging AI IMHO.

Yeah, I totally agree. The AIs there had issues as well but, if you played Civ or Galciv at high difficulty levels, they'd pose a challenge.

Look on the bright side. You spent many hours finding out how you can beat the AI at more difficult settings. Time well spent I say. Isn't that the purpose of the game?

My ideal AI would not give the AI more money, resources, etc. But actually making smarter decisions. Yes, that means making less smart decisions at lower levels, while it can do better. A real challange for designers/developers I would say.

While your point is incredibly valid, Spaghetty, I just do not think a single-player game can be justified if the AI is not challenging for all skill levels--especially the people who have devoted the most time into learning how to play the game.

*And the civ example is pretty valid. I consider myself a rather average 4x gamer, and I had a great deal of trouble trying to beat a game of Civ 5 on anything higher than Prince. That said, DW and Civ5 have very little in common other than a vaguely similar 4x background and yet still this would be the best solution. If you do not have the resources or the time to create an AI that can compete with the player on a somewhat level playing field, then please by all means MAKE THE AI CHEAT. There is nothing wrong with that in my book. AI War: Fleet Command has a famously enjoyable AI that is based completely on the idea that it does not follow the same rules as the player. One of the biggest reasons for 4x Games not ever being all that challenging without really notching up the difficulty is that people focus on creating an AI that tries (and ultimately fails) to act like a human, instead of creating an AI that provides a fun challenge.

ORIGINAL: Spaghetty

Look on the bright side. You spent many hours finding out how you can beat the AI at more difficult settings. Time well spent I say. Isn't that the purpose of the game?

My ideal AI would not give the AI more money, resources, etc. But actually making smarter decisions. Yes, that means making less smart decisions at lower levels, while it can do better. A real challange for designers/developers I would say.

Absolutely, its the finest 4x game since MOO2, but like all other games of this genre is hampered by the AI. I have full faith in the devs improving the AI as well.

ORIGINAL: Farrels

A different perspective:

Is the Disant Worlds AI not fit for purpose, or are you guys putting so many hours into the game that you have it beat?

IMO AI has been the bane of almost all 4X games since MOO. Space Empires, GalCiv they all have AI's that pose no signigicant challenge when you learn the mechanics of the game.

Its the same as Distant Worlds, the first dozen hours or so the AI can hammer you, but then you notice the deficiencies and then it becomes a cake walk. The ideas in this thread are good, but im sure after another 20 hours or so youd be back at square one as further deficincies in the Ai would become apparant.

Therefore

IMO id like to see Civ style difficulty settings. Anything over Prince and the AI gets significant bonus's, where you get penalties. I know Legends has this but it needs to be expanded upon, with bonus's to fleets, maintainence etc. Although this leads to the AI 'cheating' in a sense, Civ is the only AI that has troubled me over a long period as it levels the playing field. Even though the Ai can still make dumb decisions, it has massive armies, lots of resources etc. This coupled with improvements to ship design could see a more challenging AI IMHO.

Well said.

Yes some very good points made. DW is such a beautiful game it would be great to see better mid to end game AI. I,d especially like to see improved invasion of home worlds.

Ah, necromancy! But when a forum is as dead as this one the dark arts are one's only recourse to breathe some life back into it.

But I digress.

One thing I would add after more reflection on the ai: ai is HARD!
I dont get the impression that the devs have any background in ai. And most games like this really need that. We can look at a map and recognize strategic positions, but try programming a computer to do the same. I have actually and utterly failed. If any of you guys know some applicable techniques for this, I am sure the devs would love to hear it. Just please post publicly because I would use the knowledge too!

Back in 1991-92 I started development of a game called "Galactic Zoo". It had pre-designed 255 ship types shared by all 16 races and no tech tree. The economy was fairly simple, with most cargo movement being the seven types of fuel. The code for the economy was 12000+ lines of c language code. I gave up on developing the game when I showed a demo to a friend and he said that with all the spreadsheets it looked too much like work. How times have changed! Anyway, the ship routing portion of the economy was 7000 lines and STILL not very good. As you said, AI is tough. In routing you have a multi-dimensional traveling salesman problem, shifting goods from producers to consumers. The more interdependent the components of the economy are, the more difficult it is to bring everything together. My estimate of combat code was another 10k+ lines and that was just to handle the combat functions without graphics... If I had sold 20k copies at $40 each for a gross of 800k, I would have gotten 80k for two years work, so it was just not worth the time to write the rest of the game. 20k copies would have been a HUGE strategy game sale back in the early 90s.

The AI in Distant Worlds will never be as smart as multiplayer human, because it depends upon strategies developed on the fly to exploit game features, which CHANGE as the game evolves. There is a lag between concept and implementation. The single player game versions are just a prequel to a multiplayer version, giving the developer some payback on getting the game fundamentals correct. Single player comes before multiplayer because once multiplayer is released the gaming community is going to be very resistant to changes in game concepts...

Just my thoughts, in no way do I represent the developers, but having attempted a game myself, I know their heartaches. LOL.

Lonnie Courtney Clay

+1 for cheating AI.

The game has never really challenged me, even when I was first playing it. You have so much money you just build huge fleets and then the AI either gets intimidated or else loses every battle.

It would be far too difficult and take too much time to try to build a very much smarter AI. Just let it cheat at higher difficulty levels, a la Civ, I have no problem with that. Are the developers resistant to this idea or haven't they had time yet to get around to it?

I think making things more difficult on yourself on purpose does not make fore a more enjoyable game

About ai, unfortunately the more features in the game (eg ship design, trade, colony selection) for a human to choose from, although it makes the game more interesting for humans to play, it makes it exponentially more difficult to make an ai that can simulate those decisions. Just in another thread someone was asking for wormholes and my first thought (unexpressed at the time) was how it would probably just make the ai even worse than it is now. I know that in MOO3 they made the decision to use "space lanes" specifically so they could implement an AI around it. That made it possible to develop an AI that could recognize pinch points and vulnerabilities. It was either that or hire a masters graduate from MIT I guess.

By the way thanks for the code quotes. I remember watching The Social Network and he was like "i did it with 20000 lines of code" and I thought "wtf is that supposed to impress me?". Real applications take 100000 lines of code. Enterprise apps are even more. Not say more lines of code equals better, but that the simplest apps still take a lot of work. Again I digress.

Tl;dr I am awesome.

I will stand on the side that higher difficulty AIs should have allowed to cheat.

They are supposed to be highly advanced space faring civilizations. They should be able to handle basic logistics pretty good. But coding resource is limited. Then there should be a way to simulate the nicely working logistic.

something like,
a. high difficulty AI freigheters never runs out of fuel. (still requires fuel to construct)
b. if not, they do runs out but speed reduction is not as significant.
c. Engine damage for them count as whole or none. No partial damage, thus no thrust reduction.
d. if neccessary, abc applies to military ships too.
e. OR, each higher diff. AI empires possess common and instant/everywhere on universe (but limited in upload/download speed and max reserve amount) fuel redistribution system. Like mineral and gas in Starcraft or other RTS.

ORIGINAL: Edward74

I think making things more difficult on yourself on purpose does not make fore a more enjoyable game

i disagree. but that's why there are difficulty sliders. im not saying that i want to be overrun every single game 15 mins in, i don't adore the defeat screen. i just want to feel threatened, at all. currently i don't feel that even slightly. when an empire declares war on me that's larger, i don't feel dread or danger, i see opportunity because that means i can start taking his worlds while he stumbles around with a fleet or two and does nothing particularly deadly. the most i have ever seen an ai do in a long war, is take two or three planets and destroy a couple of ships, after that, he just kinda shuffles around and waits to be defeated.

granted, i have not played the last few beta patches, so i don't know if any good changes happened to the ai, if there are please tell me, ill fire the game up and give it a try.

Well, here we are a year later and I must say I am unhappy with the Legends AI. And that's a shame because the system is rich and abundant with deep detail, so primed to be a classic. But I must play the game for the journey not the challenge. Perhaps Shadows will improve it. Time will tell.
Is anyone else struck with the similarity with GalCiv on many levels, like I am? Not that that is bad, just many eerie similarities.

[&:]

The best strategy AI opponents are the ones in real-time games. Because there the AI has the advantage of being able to calculate a lot of options and control every single unit in real-time while a human is limited by number of decisions by time. But even so, humans are able to beat AI in RTS games like Starcraft II with relatively ease. And that is because we can still beat it on the strategy and tactic level making better decisions despite the speed advantage.

In a complex strategy game like this, it's very hard for the AI to beat a human, just like jpwrunyan says above (some time ago though). Because it's very hard to create a smart AI with so many choices to make.

Distant worlds is a real-time game isn't it? Just on a larger scale than normal RTS. With obviously less emphasis in certain aspects in the genre, but it is played in real-time nonetheless.

The reason why humans can beat AI in an RTS game like Starcraft II is the difference in effort placed upon the AI for a multiplayer game. The AI simply isn't developed. Compare and contrast with Starcraft:Broodwar, where there has been some concerted effort to create AI capable of beating humans; eis.ucsc.edu/StarCraftAICompetition

In the end, creating an effective AI is a matter of recources, which is usually a waste in a multiplayer focused game.
The other side of creating an effective AI is understanding the best possible decision making, which if you look at how the AI automates, it is obvious that it simply hasn't been programmed to win.

The AI in Distant Worlds seems mainly be designed to create an interesting universe, as opposed to one that makes sense that the Races are trying to outcompete the opther races.

ORIGINAL: Astax

I think making things more difficult on yourself on purpose does not make fore a more enjoyable game.I wish the game had Multilayer so that real hardships can be experienced

I have learnt that facing other players is not more challenging, its just more predictable, since when I used to play RTS games back in the day when it was more about having fun, then just winning, once you learnt how a player plays the game, ultimately, it makes player vs player very predictable.

The AI is not predictable, they just suck at it. lol

If playing other humans are predictable, then clearly, the multiplayer RTS game you are playing is very bad, or you are playing against a very small select group of people who are also all very bad.