5421

It’s clear they are different levels of “brokenness”. The AI needed a bit of help in bonus in videogame to compensate the better tactical plans of the players is normal, you could say that doens’t apply to the broken label. The AI needing bigger bonus to put a barely decent fight is higher in the brokenness scale. The AI not reacting well to flanking maneuvers in a game where is the no 1. tactical is still more broken. And the AI not attacking even if you are tearing up his army with your army composed of 90% of veteran archers is even more broken.

But hey, the game is just sooo pretty, so i give it a score of 95/100!

5422

I don’t know which AI was worse: Age of Wonders AI or Total War AI.

I have a gut feeling Shogun 2 might be better in this regard- given the more limited scope of Shogun as compared to Empire or Medieval.

Did modding help Total War AI much? I’ve kinda been boycotting the series since Medieval 2.

5423

Talking about AI, the AI in AI War: Fleet Command is generally celebrated. Check it out if you want a good AI time or maybe even an opponent so smart on highest difficulty levels it is yet to be beaten (with no cheating, I hear!)

5424

Well, considering the AI in AI war plays with entirely different rules than the player, having a highest difficulty that is yet to be beaten doesn’t mean anything. That is like saying a first person shooter has great AI because no-one can beat the highest difficulty which has you fighting against overwhelming odds. I’d say the AI in AI war is interesting and fun to play against, not necessarily smart.
Fighting the AI war AI is like fighting a force of nature, not an intelligent and cunning opponent.

5425

I was under the impression the AI still plays nicely within the established rules (even though they are purposely different from the player’s) rather than just overwhelming you with infinite, cheated resources. Isn’t there intelligence required for that? I haven’t delved very deep myself, that’s just the vibe I’ve picked up here and there.

5426

I’ll give it a go to try to explain the differences.

The problems with E:TW were/are serious, but you generally didn’t uncover them until you played it a while. Hence the reviews were often good, but any kind of in-depth playing revealed a fundamentally broken game. So that explains the difference between the review and the user scores on Metacritic. Also, some of the problems really didn’t happen until CA patched the game and broke it entirely for quite a few users (including me). Then instead of providing the promised hotfix, they waited over two months to fix the problem their patch created all the while releasing DLC instead of fixing the game. That’s when folks really got upset.

Elemental was undoubtedly in a worse state upon release and the problems far more easily apparent. Furthermore, even while CA had some gaffes on the PR side, they’re nothing like Stardock who has managed to put their foot in their mouth repeatedly since release. Finally, Brad’s often abrasive internet personality came back to bite him. He got into personal flame wars with people and those folks came out in droves to fan the flames, Matt G. just being the most obvious example. Call it the “Derek Smart” effect…

But that doesn’t mean that everyone has forgiven CA. I won’t ever buy one of their products again because they repeatedly fucked over their customers on E:TW. I still think it’s unforgivable to break the game with a patch and then drag their heels on fixing the problem while instead pushing out DLC.

5427

Some intelligence is necessary yes. But the rules change when the difficulty changes. The high difficulty is difficult not because the AI is smart, but because its rules give it ludicrous amounts of ships every wave, ludicrous amounts of free defenses etc. There are no ‘established rules’ for the AI. Different AI types are different because they have different rules. More difficult AI’s are more difficult because their rules are more favorable for them.
Because of the lack of ‘established rules’ for the AI, you cannot speak about how intelligent it plays.
For example:
If there is only one adjacent warp gate to your empire, the AI always assaults from that warpgate to your massively fortified planet. If you destroy that one warpgate, the AI can now use the non-adjacent warpgates (if AI difficulty is high enough), and therefore assault any border planet.
Now, is it a game rule that the AI ‘must’ warp directly to an opponent planet if able, even if that is suicide, or is it an AI flaw that it does?
The AI in AI war is like the wandering monsters and bandits in Elemental. They don’t produce, they spawn. Their actions are as much intelligence as game rules. They are not an opponent but the environment. Would you ever call the wandering monsters in Elemental a good AI opponent, even if they made smart decisions on whether to attack your cities or not?

5428

Ok, I stand corrected then. Thanks for the intel! It might save my fleet some space day.

5429

AI War was even part Tower Defense, reading the description of some people, so of course it’s not the type of game where you can judge the intelligence of the AI. Hell, i think to remember that some of the explicit rules of the game indicated a native unbalance between player and AI, it was designed that way. In fact, it was praised how the design was something “different from the usual”.

5430

my definition of AI seems much broad than you:

moving a character from tile A to tile B is AI. (pathfinding)
moving a tank from tile A to tile B avoiding areas where tanks die is AI. (pathfinding).
calculating the ration of turrets / tanks is AI.
searching for a empty spot to place a turret is AI. ( involve testing if a tile is unocciped, starting from the nearest tile from the buildcenter building )

all of the above create a simple RTS game.

5431

The real crux with E:TW is that unlikes most other total war titles you are playing a period in which 99% of all units and fights end up being 2-3 cannon units firing in the background while the rest are all identical units of lineman standing in front of one another, shooting.

Though historically fairly accurate, this is so mindnumbingly tedious and not varied even with the research tree(just unlocks different bayonets and deeper rank shooting, formations, etc, but almost no other unit types that make “sense”).

In the end, E:TW is standing rows of men shooting at one another until one side runs out, first, whereas almost every other era and TW game was a LOT more rock/paper/scissors with melee/archers/horses/optionally cannons and intra-melee division rockpaperscissor mechanics as well.

So that game(again, due to the ERA) has effectively crippled the core variety and combat fnu while still having interface/research bugs after patch 10++.

I would say just wait for Shogun 2 :P
Fingers crossed!

5432

Ugh. Artificial Intelligence is one of the biggest naming blunders in computer science, because the field is a technical one and both “artificial” and “intelligence” are extremely loaded words for mostly non-technical reasons. Gamers and developers love to throw “AI” around even more, even though the technical aspects of game AI rarely overlap with the technical aspects of AI as it exists in computer science, and even though the words are arguably even more loaded in the context of games than they are generally. As soon as you say “AI,” players bring in a huge number of different, contradictory, and almost always unrealistic expectations.

Look at poor Chris Park. He writes long blog posts about all the code he wrote for AI War so his computer opponent would be interesting and different from other computer opponents, and even make decisions emergently so as to be less exploitable. What happens? It’s too confusing for people because it’s different: people thinks it cheats because it starts with more resources, among other things. They pick apart every flaw they can find, real or imagined, and fault the AI for making decisions differently at different difficulty levels. Meanwhile Brad Wardell writes an AI (for GalCiv2) that doesn’t understand how to run its own economy and is therefore a complete pushover until you crank the difficulty up to where it actually does “cheat” by getting an arbitrary resource multiplier relative to the player, and everyone hails him as an AI genius because the computer will occasionally tell you it sees the transports you’re building, or whatever. Everyone reads all sorts of diabolical personality and genius into what the AI opponents do (call it the Dwarf Fortress Effect), and no one knows or cares that the AI makes decisions with different algorithms (rules) at different difficulty levels.

Developers already know that the computer’s job is to lose without forcing the player to think too hard, but IMHO they still tend to underestimate the extent to which players prefer an opponent that will present some basic appearance of having human personality characteristics over one that gives them interesting tactical or strategic choices to make, all other things being equal. And that holds regardless of whether or not the opponent is supposed to be at all human-like, thematically.

If I were a developer, I would not talk about the AI at all. Who cares if the “AI” is “smart,” or plays “fair.” If I did talk about the computer opponent, I would emphasize its fun personality rather than its artificiality or its intelligence. The computer opponent can have extremely limited behavior, and no one will care as long as it has personality: think Plants vs Zombies. Doesn’t apply to 4x strategy games? Consider this: do you want to play against the best and most clever algorithm in the world on a level playing field, or do you want to play against the ridiculously powerful evil sorcerer who’s running amok with undead hordes and insulting your sister, and who also happens to be an arrogant moron? Of course you want to play against the best and most clever algorithm in the world, right after you crush the evil sorcerer underneath the heel of your righteous fury, several times over. Who does he think he is, anyway? What a dumb jerk.

5433

So … um … games should be fun? (despite my flippant response - which was entirely just for the heck of it - I get the drift of what you’re saying and agree with it even though I don’t hold the same opinions of the examples you used)

5434

No fair teasing opinions you aren’t going to post! What do you disagree with?

Yeah, I think games should be fun. I agree with Tom that “fun” is vague to the point of almost being useless on its own (though not as loaded as “AI”), but I’m at least narrowing it down to having characters or agents with human personality traits.

5435

Cool story bro.

5436

This is a complete mischaracterization of the discussion.

AI War is not a symmetric game. The AIs are not playing the same game that the humans are. They don’t follow the same rules in any fashion. If you put a human in control of the AIs’ resources, the game would end within minutes, because the AIs start with an absolutely overwhelming advantage. Managing the AIs’ aggression is a critical part of the game, because they’ll stomp you flat if you goad them enough in the early game.

That’s not a flaw, and this isn’t a criticism. It’s the kind of game AI War is.

5437

As people have said already:
Game AI is about losing gracefully

That’s why GalCiv II’s AI for example got good reviews, it “talks” with you with more character and gives you more memorable moments.

5438

I don’t care if it cheats, I do care when it can’t play its own game, or is stupid as hell.

5439

I want an AI that will play by my rules, and give me a challenge while doing so. Personality can be done with in-game stuff like diplomatic dialogue.

5440

Well, I tend to pull my punches around these parts due to the likelihood that I’d eventually step on someone’s toes because they had a hand in making the game in question. So without specifics, I think the key that you touch upon is that opponents should have engaging personalities if you want widespread appeal. Something like AI Wars presents opponents without that trait by design, and that can lead to dissappointing sales if you were hoping for the unwashed masses to fork over $30.

Likewise, you and others have noted that the ideal game is one that everyone who plays it can also find some measure of success in doing so. Frustration too often leads to bad word of mouth and games collecting dust. AI Wars certainly does allow someone to succeed, but it’s unusual enough that I suspect the lack of an easily identifiable path to success puts off many would-be customers. Of course, on the other hand it provides such differentiation so as to entice the target niche into purchasing the game. Still, it’s a choice between targeting narrow or wide distribution and sales, and they again chose narrow.

There’s absolutely nothing wrong with creating a niche game. There’s only a problem when you expect it to outgrow the niche without laying the groundwork for it.