AI in MOO3

[JamesH]

1stly gday everyone. im getting pumped for MOO3 and have been reading bits and pieces about the game.

I have never played any games in the series and was wondering what the AI has been like in previous versions. games like HOMM spring to mind where AI in 3 was good, but the AI in 4 was terrible.

Does anyone else think that the AI could be the achilles heel of this game, or has anyone any evidence to the contrary?

[kalbear]

It's hard to compare, since the games are being designed by totally different companies with no code being shared between the projects.

Moo1 had...adequate AI, I guess. The ship design was intelligent, the AI would build various types of fleets and would design ships based on what techs/tendencies you had, IIRC. As almost all empire building AIs go, it didn't prioritize highly for industry/research, so it'd often lag far behind. It did not take advantage strategically aside from the most basic chores, such as leaving an undefended colony near their border.

Moo2 was possibly worse in the AI department. The AI ships were horribly built and could easily be outclassed. The AI using the ships was, for the most part, vaguely decent but has exploitable tendencies. Planetary management AI is just plain dumb. Diplomacy was usually pretty bone-headed as well. Strategically, the MOO2 AI was an encephaletic, still born infant, who would often declare war when seeing one lone outpost it could blow up, even though you outclassed it in every single way.

Moo3 appears to be much better on a tactical and strategic level. There's indications that the AI will figure out your tendencies and build counter-mission ships to defeat you. It'll definitely use the ships correctly, and appears to do a good job managing weapons on those ships. It appears to do well on the tech tree. It might not be so great with industry buildup; hard to say, really. It does appear to be reasonably smart when declaring war and attempting to do realistic goals without totally irrational behavior. It appears to be pretty reasonable on prioritizing expansion over consolidation. Also appears to be quite decent at spying and diplomacy, especially in the senate.

It looks like it'll be better than something like, say, Civ 3. Honestly, only folks that really could tell you would be the beta testers.

[JamesH]

yes im worried that ive heard glowing reports about all sorts of aspects, but nothing saying OMG the AI is demonic.

[Jack Frost]

A) i'd just like to know if beta testers lose to the ai.

B) i'd like to know how much of a military advantage the AI needs in order to be competitive (in most strat games unless the computer has 3x as much stuff as you, you can win).

C) I'd like to know how moddable it is.

the thing will live and die on the AI.

[kalbear]

kebzero has stated that he won his first game on impossible just recently, and that was with the luckiest start he's ever had. That's a good sign for me.

I wish we knew how moddable it was. That's somewhat troubling. I do know that the graphics/sound are all in public domain formats, though the tools that were used to develop 'em aren't there yet. Supposedly, the useful information about the game is stored in excel files. What that information is, well, they've not said yet.

[Grumbold]

Originally posted by JamesH
yes im worried that ive heard glowing reports about all sorts of aspects, but nothing saying OMG the AI is demonic.

If the AI was truly demonic then a lot of the casual gamers would be put off. Most people want to have a bit of a fight on "normal" but be proved superior in the end. So the AI can't be too smart or QS risk losing sales (either of this game or the sequel.) The true TBS experts will be the ones jacking the difficulty up to the highest level and looking for a real struggle for survival but not one that is truly "impossible".

[BigBopper]

I would be in heaven if the normal level of difficulty the actually had to dumb down the AI. So that only on hard level would the AI would truly be unleashed. I hate it when you have to give the AI all kinds of advantages (cheats) just to keep up with the player. In civ3 for example the AI starts with full knowledge of the map in order to be able to function. That kind of crap drives me nuts. I want to have an AI be able to match me (or better me at first) given the same tools, and playing by the same rules that I have to.

[Grumbold]

Yes ideally the AI would be tuned for the top level then somehow made more stupid for its lower ratings, like Chess programs. Unfortunately I believe in Chess they do this by restricting the time and depth in which it can analyse its best move. In a TBS there are hundreds of decisions that go into making up a single turn to it can't work quite that way. Its simpler (and achievable!) to give them moderate AI coupled with progressive multipliers.

[Sarxis]

Does the AI cheat?

[Alex]

It has to cheat. Every AI cheats.

[BigBopper]

The advantage that the AI has, is that it can micromanage better then any human could hope to. Given the complexity of the game, however, the designers will almost certainly have to give the AI some help. However, I’ll be **ssed if they have to make up for dumb AI by giving it massive help. Too many designers have simply written brain dead AI’s and made them competitive by giving them massive cheats. We can only hope that the Moo3 team took their time to write a good AI. Of course we could also just play all our games as multiplayer

[kalbear]

The only AI that cheats is the New Orion AI. The other AI races do not start out with any more tech, resources, money, planets, ships, or anything else.

They are basically indistinguishable from human players at the start.

All AI does not cheat. It's a common practice because it does indeed do what you want - it makes the game harder and more challenging, which is the goal of any game.

[Sarxis]

Does the AI have omniscience? And if it does, will it handle cloaks and such?

[LastFromVega]

About micro-managing, maybe I am naive but to me the Planetary AI and the Finance AI were the same as the one we have access to (as governors): the BTs all said they relied on those two heavily, certainly because it was well designed.

About the difficulty setting, a quote from LoreWeaver said:

The AI setting basically controls the AI's drive to win. The AI will basically be different diplomatically. If the AI thinks you are going to win (or on the path to winning) it will declare war on you.

the worst AI I have seen yet, was civ 3, not because it was lame, but because it cheated so much on knowing where the resources were.
I don't mind the AI cheats a bit on production, or on research points, but I do mind if key strategic info is known when it should not...

Maybe I am dreaming here, but maybe we will have a worthy AI for once...

time to go home, see you folks!

[Todd Hawks]

In fact, the Civ3 AI is the best I've ever seen. Sure, it cheated concerning resources but it didn't cheat on anything else. (I don't consider bonuses as cheats). I've never seen it doing strange stuff because of a resource known to it (well, apart from settling on small ice islands, but there's often oil on these islands I do the same).
And it is able to launch a successful attack with major forces, and strategically use alliances. And a lot of other stuff.

It's far from perfect but show me an AI that performs half as well as that of Civ3 (in a game as complex as Civ3 of course).

[Jack Frost]

mmm, i donno. The Civ3 ai was sort of a one trick pony IMO -> tech trade with all the other AIs, ignore the player. End of strategy.

But I agree. AI is always ****, its why I have trouble bringing myself to play a SP game. :/

[TheMaestro]

CivIII AI did cheat horribly...ie: you're pounding the hell out of them in war, you're taking city after city...yet you offer peace, your population is upset about war -- and the computer demands stuff from you for peace. LoL

Hopefully MOOIII AI isn't like that.

[Alexfrog]

Actually in civ3, if you pounded the AI and took a couple cities, they would be willing to give you tech or stuff for peace IF you had a good reputation (i.e. didnt sneak attack/break peace treaties before the set time.)
Go read the strat articles in the forum on this site....

moo3 should be a bit like that....not like moo2/1, where when you were killing them they would hate you forever and never talk to you again. no matter how overpowered they were...
in moo3 they should beg for mercy I think, which is good, and realistic.

Yes the AI "strat" of swapping all techs amongst itself, ignoring the human was dumb, and pissed me off. Especially at higher difficulties when it was impossible to gain a tech lead.

[Alexfrog]

Originally posted by Todd Hawks
In fact, the Civ3 AI is the best I've ever seen. ...
It's far from perfect but show me an AI that performs half as well as that of Civ3 (in a game as complex as Civ3 of course).

Civ3 was pretty good, IMO, except for being very unfair in deals with player, while swapping stuff like crazy amongst itself.

I expect moo3 to be at least as good, and cheat less. This is based on what the betatesters have said. I expect it to be fairly good at managing your colonies and thus reducing micromanagement in the mid and late game. Why? Because you can give it a lot of help in terms of what to build by creating priorities for different colony types, and then assigning each colony one of thsoe colony types. So you can specify a build priority for a "research world" and then tell it to make colonies x and y research worlds. And specify something else for a "shipyard world" and specify something else to be that....
it can therefore do good because you give it some general guidelines, and it is capable of following them.

in previous games, autobuild always sucked because you couldnt give it guidelines.

[Grumbold]

I did not like the way the Civ III AI cheated or conspired against the player to make up for its inability to control a mixed force army, but I've beaten that to death on the appropriate forum.

That Planetary AI and Finance AI can be pretty much left to get on with it is a good sign.

[Ozymandous]

Originally posted by Alexfrog


Civ3 was pretty good, IMO, except for being very unfair in deals with player, while swapping stuff like crazy amongst itself.

It only "swaps" if the other AI's can afford what it's offering to sell. In this aspect the AI treats every other player in the game, human and AI, the same. If your version of Civ3 AI doesn't work this way then you haven't played with the latest patch.

I expect moo3 to be at least as good, and cheat less. This is based on what the betatesters have said. I expect it to be fairly good at managing your colonies and thus reducing micromanagement in the mid and late game. Why? Because you can give it a lot of help in terms of what to build by creating priorities for different colony types, and then assigning each colony one of thsoe colony types. So you can specify a build priority for a "research world" and then tell it to make colonies x and y research worlds. And specify something else for a "shipyard world" and specify something else to be that....
it can therefore do good because you give it some general guidelines, and it is capable of following them.

in previous games, autobuild always sucked because you couldnt give it guidelines.

I am worried about the MOO3 AI because it will seem to have the SOSDD "strategy" of "if the human player is winning then EVERYONE dogpile on them ,regardless of past treaties, etc". I had thought this strategy was being discarded as companies finally put some real "I" in their AI, but from what some of the beta testers have said this is not true.

Let's all hope the AI in Moo3 handles war declarations, honoring treaties, etc, as well as the Civ3 AI does. I like having repurcussions to back-stabbing the AI (invading without declaring war) and having relations with other races actually reflected in your general diplomacy.

[Ozymandous]

Originally posted by Grumbold
I did not like the way the Civ III AI cheated or conspired against the player to make up for its inability to control a mixed force army, but I've beaten that to death on the appropriate forum.

That Planetary AI and Finance AI can be pretty much left to get on with it is a good sign.

Ah, obviously you haven't played the newest version of Civ3 because the AI handles war better than any other 4X game on the market, IMHO.

The only way the AI "cheats" is to know where resources are before they appear, unless you count on the AI "bonuses" when you play at higher difficulty levels.

Compared to Moo1 & 2 and Civ 1 & 2, where the AI's main strategy was "dogpile on the human player" the AI in Civ3 is top notch.

[Grumbold]

Like I said, done and dusted imho. I demonstrated to my own satisfaction that the AI knew where my units were well outside of its legal view range and reacted to that knowledge. I'm basing my observations on the base game up to and including 1.29f and not including Pay More Money or wahtever it was called. You can search for the full exposition on Civ III General if you are remotely interested. Having uninstalled Civ III months ago and not looked back I'm done. MoO1 still holds my attention because the AI's can attack you in different ways (and I rarely have a problem with dogpiling either.)

[niteshade6]

Odd, I've never minded when the computer cheated by having some extra info that the player did not have. It bothered me more when he got major bonuses over the player and always had 2-3 times more units then you did (for reasons having nothing to do with fast expansions). Two of my favorite AIs were Civ 3 and HoMM I, both of them recieved extra info, but I never minded so much. Especialy since I don't believe in HOMM I they got any extra bonuses at all aside from the extra info.

[Kc7mxo]

Its always best if the player doesn't KNOW the AI is cheating. As long as the player isn't sure, as far as he knows, the computer could be just playing well. But when its cheating becomes overt, anger ensues. Flagrant rule breaking (like triremes in civ) will constantly piss off a human.

But if you don't know that the AI knows the whole map in starcraft, you might think he just scouted your position before he rushes you.

[niteshade6]

I definetely agree there. I think that's why it doesn't bother me when the computer just has more info then you do, it's hard to tell that the computer is cheating unless you really study his moves. It's only really bad when the extra info he gets makes certain strategies or options useless. Like if cloaking is useless because the enemy knows where you are anyway. I remember Master of Magic had a spell which disguised your unit as another unit, but they say in the very description of the spell that it didn't work against the computer. Given that there was no multiplayer, and the computer never used this spell, what was the point of it anyway? Of course a later patch fixed this by changing the spell into something else.

[Kc7mxo]

One bit of AI cheating I hope moo3 avoids is what happened in IG2. Playing that game on easy was a total breeze. The computer practically flipped over and played dead when you met him. When I attempted to up the difficulty, I got my rear handed to me. Repeatedly.

The final straw came when I was playing as the Kra'hen. The damn humans were busting me up with this single fleet of 16 massive battleships. This fleet wandered through my territory, shotting up planetary defenses, but not really invading. When I finnally smashed them with a fleet, most of which died in the battle, I was yelling in victory. it had been a terrible challenge, and a fantastic fight. But I was victorious. After the battle I looked at the map, and saw three fleets heading for my territory. Three fleets of 16 massive battleships. I watched as they smashed everything I had left, and invaded most of my planets. I then quit the game and unistalled it.

Its bad enough to get your butt handed to you, but its simply unacceptable to find out the computer was toying with you the whole time.

[kalbear]

kc7, that doesn't sound like cheating - that sounds like the AI finally getting back at you for taking advantage of various exploits throughout the ages.

That sounds kinda cool, actually. Why didn't I pick up that game?

[Kc7mxo]

That sounds kinda cool, actually. Why didn't I pick up that game?

The problem was that no matter how many times I tried, (and I reinsatlled by the way) I could never beat that damn AI on normal. I tried everything I could think of, I was even reloading to deal with enemy fleets. He always had more ships than me, better tech, more spies, and better planets.

I could hold them off for a while, especially if I researched the best defensive buildings and built them ALL over the place, but eventually a horde (and I do mean a horde) of extremely advanced tanks would drop out of the sky and annilate me.

Going back to easy did nothing for me. I could smash the AI there without even paying attention.

Its like if its the first time you play civ, and you have the option of playing on chietain or deity.

[Grumbold]

I dunno why Kc. I won the game on normal as all 3 races without huge difficulties both times I played through after an initial learning game. All I did was expand as fast as possible both in size and in technological development.

If it has the jump on you just keep it capturing the same planets over and over by recapturing them once the enemy megafleet has moved elsewhere. Once you can build 3 of the best fortresses on your main planets you're pretty safe even if his fleets are overwhelming.