Well, complaining about the AI is a little harsh - that said, given how much has been talked about the AI, I'm certain that some folks had very high, even unrealistic, expectations for it. What I would hope that Brad would be open to is suggestions from said "best human players" on how the AI could be improved.

Oh, and Skynet is a sore loser.

what is the difference between difficulty level.... and the intellegence setting on the AI... Is difficulty level just reflective of the choices you made to AI settings on the individual races?

What started off as a discussion of the AI in this thread seems to have degenerated into (tearing down someones hard work) Which didnt seem to be the case. I on the other hand have lost a couple of games but on the front side not gearing up for their rush, but have never found a good AI planet as well. When they put a special building on a tile you cant even tear it down to build the right thing, certain building cant be deconstructed, and those are the ones the AI puts on the good tiles... well its probably just the odds but hey.

I agree that some people here seem to expect too much from a computer game. The AI seems pretty good to me, considering what it is. I have seen it do some good moves, at least twice: one time I saw an enemy fleet and sent a ship to pursue, and lost track of it. It turns out that that fleet went around my sensor range and attacked me from the opposite side, trying to take out the sensorship.
The other time I had several sensorships covering my territory and moved one of them, which created a thin fow covered zone (one square large). I thought no problem. And turns out that the computer sent ships following that line, undetected by me, and attacked one planet in the heart of my territory that was lightly defended.
And I’ve seen the computer construct ships that are equipped to face my own. And that evolve if I change weapons on the ships. Of course it takes some time, but I’ve noticed it, which pushed me to build ships with 2 weapon types so that there is always one that the AI has no defence against.

There are only 2 things I don’t like and that seems to me weaken the AI a lot (mind you I don’t play big maps, always medium with all civs, mixed intelligence levels, but not above intelligent : after a year or two some civs have given up to others and it is always a different game that way – without me controlling it all with the start settings).
The first thing is that, indeed, the AI build a lot of very small ships that cannot do anything against my large designs in the late game. When medium is available, they should at least use it…
But what I really consider as a flow is that the AI never go attack my military starbases. Since I play a more defensive game, and fall behind in the start, I build one or two of them (depending on how many planets I have to defend) that give a lot of bonuses and allow my small ships to survive till the endgame with incredible levels of experience, thanks to the defence assist modules installed in the starbases. Most of these ships are worthless when first built, and would be easily destroyed by AI’s very superior ships (since they have more attack techs than me). If they attacked the starbase they could easily take out my fleets and then planets. I’ve had a single fleet of 3 small ships defend against 5 or 6 waves (24 ships +/-) and come out without a scratch. If they went for the starbase I could not have defended it. (no engines on my small ships).
That’s my only real critic. Maybe it can be corrected, since I noticed that the AI will always go for my sensorships. ALWAYS. Which is a good tactic as well since it takes out a huge advantage of mine that compensate for the slow speed of y ships. But still, the starbase would be a much better target.

But all in all, good AI, limited by this starbase flaw: I would not have won the 3 games I posted if the AI tried to take out my military starbases.

Look. nearly all the criticism in this post has been constructive. It is not vague unspecific mud slinging but observation of very specific flaws in the AI. It is just a FACT that the AI makes some very dumb mistakes. Biulding embasies on 300% manufacturing tiles for example.

These flaws interfere with my enjoyment of an otherwise great product. I payed $$$ for that product am I therefore not allowed to make observations about that product?

Most of the negative tone in this thread has come from the overtly aggresive unconstructive flaming of those who are mearly stating the FACTS about some of the AIs behaviour.

As for being an expert player, I have never played Galciv 1 and beat GC2 on tough on my third attermpt. I believed the hype about the AI, at the moment it dosnt live upto that hype. Though if stardock take on board some observation and are more open minded than some of the flame tards in this thread then the AI should improve dramaticaly as the foundations of a good AI have been put in place it just needs improving in key areas ti remove its obvious and dumb mistakes.

The planet I took over from the Dark Yor seems fairly well put-together, even without any reconstruction. Actually, it's probably my best planet at the moment.

You know, with all the comments about poor algorithms at higher difficulty levels, it might be worth checking that the AI difficulty selection didn't actually get messed up somewhere around release.

Overall, I think there is a good chance we'll have a much more competitive opponent over time. This is because the AI has a number of weaknesses at the moment that tend to reinforce each other, but each of the weaknesses seems addressable. I'll put up a couple of suggestions, but they might not be practical at all...

1. The AI has to be able to take the observed ship types of an opponent (and the proportions) and using the military strength score, make a reasonable estimate of the opponents total inventory of ships. Then as a hypothetical exercise, form those ships into fleets and fight a few (dozen) simulated 'wars' between it's fleets and the opponent fleets (just a series of fleet-fleet engagements until one side is dead...). Some reasonable proportion of the engagements in those simulated wars should be under the influence of military starbases on one side or another. The AI estimation of relative military strength should be based off the median outcome of these hypothetical wars, not raw military rating. You can imagine elaborations on this, such as fleet speed determining how often one side gets first attack. This is working on the assumption that you can simulate hundreds of typical fleet engagements per second (which I don't think is unreasonable).

2. I have to go to work ...

I wouldn't say the AI is weak but on my first intelligent difficulty game I was outclassed in every way from the beginning to the midgame. After the Torians (the most powerful civ at the time) declared war on me and started sending a deluge of unescorted Transports that I easily destroyed with a ship built from a cargo Hull with one Laser and lots of engines (appropriately called "Desperate Measures" ). After some time I started taking their planets and soon I was a local power. From there I easily got the upper hand and I managed to won the game.

I admit the AI did some intelligent things. They adapted to my strategy. They started putting weapons in their transports and using fleets of them. They built ships with force fields to contermeasure my lasers. But at that time I already had better ships (Like fleets from a tiny well armed fighter called Betafish). The war was lost for them.

If the AI is building embassies all over then you're not playing at a high difficulty level. That's one of the low-intelligence algorithms on planet set up.

Although I don't support the tone or comments of the thread starter, I've been playing the AI on the intelligent difficulty level and both of the two planets I've captured in my current game so far were covered in embassies - maybe 50% of structures or more. One was Torrean though, who are influence-freaks I think.

May be something wierd is going on with which levels are associated with which algorithms?

Any way, I think we should give Frogboy the benefit of the doubt here as he seems committed to improving the AI ... we can always flame the hell out of him later!

Speaking of smart things...

In a recent game the torians declared war on me (the Iconians). The first I saw of their warships was a fleet trashing a rather expensively upgraded but undefended economy starbase- didn't even see them til they were right next to it. They almost pulled it off again with a single fighter.. fortunately I had a constructor nearby so that didn't come off. They also darn near got transports to several of my undefended worlds but fortunately I had some fast ships (<3 my mosquitos) that barely managed to stop them. In previous games, I've lost worlds deep in my own space due to enemies circling WELL around my front lines (thus bypassing my powerful military starbases and numerous fleets) and me not seeing a damn thing until the invasion screen popped up.

Suffice to say I've learned to start using the Sensor tree >.<

I will take exception to "the AI is pretty bad" though. Pretty bad compared to what?

Well, pretty bad as it the AI in GC2 is just as bad as in other games. But the thing is, other games usually rely on their mulitplayer or long campaigns for good gameplay while GC2 is really a single player game so any small AIs flaws are seen repeatedly throughout even single game.

Ok, here's something constructive (my first post touched on the basics of the problems with the AI):
The AI does not guard (or hell even attack) transports. Sometimes I see a transport in a fleet. Wtf? What is the purpose in fleeting a transport with combat ships? You have a gimpy fleet and it now only takes 1 move to kill the whole thing. Putting a combat fleet on top of a transport fleet is far superior. This is a simple fundamental part of conquering, without doing this, it is really hard for the AI to win. Now the problem may not be with the actual AI, the problem is when you group 2 fleet and Auto-pilot them to the same place, they move at their own max speeds so they get separated and destroyed. So you have to micro them and make sure your transports don't get ahead of your fleets (transports should have troop modules and engines only, no weapons or shields on transports silly AI) A uman does this by micro'ing his fleet each turn. A computer may be able to do this by not using Auto-pilot or by making a lot of waypoints = to the speed of the slowest fleet in the attack group.

So that we're on the same page:

I'm all for taking feedback from players on the AI.

And I have a long list of pet peeves with the game. But when someone acts like the AI is basically crap, it's hard not to start dismissing because I know it's not crap.

AI has to be compared to other games and I suspect a poll would say that we've done pretty good on the AI, particularly compared to other games.

I realize it's not good PR to be defensive. But I wrote the AI so naturally I take flammage personally.  There's some good GOOD stuff in this thread. But it's mixed between guys going "Yea, the AI is just broke. Hope they fix it."

Please. If someone's beating the AI at suicidal, go onto the Metverse and start demonstrating that. Become #1.  Play a large sized map with all the AI players at the highest difficulty.

Because I can tell you right now, if you're beating the AI with a decent mix of planets (i.e. where they have several planets) on a decent sized map then I suspect there's an exploit involved somewhere.  But on the off chance that you have beaten the game on the highest level, then well, there's no way the AI will be ever good enough.  There's no scenario where I can make an AI that doesn't cheat that plays as devastatingly as the current AI at the highest level.

 

From my point of view, the AI does have it's low point at LOWER DIFFICULTY LEVELS.

My main gripe is that even at intelligent, with 100% economy, the AI doesn't seem to be doing that much.

On the other hand, playing a higher difficulty levels (where the AIs get an economic bonus) they can be seen in their true splendor.

I think it might be a bug or something, but shouldn't it be doable to allow all algorithms without an economics bonuses?

Apart from that, the AI -is- truly remarkable when set at its hardest. I just think "Intelligent" isn't at its hardest.

My AI suggestions for things non-military :

Have it focus more on influnce techs. What I'm finding is that if it falls behind in influnce, it then builds embassies on all the border planets to make up for lack of tech, using up valuable squares.

Focus more on population growth. I'm always leaps and bounds ahead in population. I see the AI with many 5 pop planets.

Adapt to changing borders. Less need for orbital defense and ship production on planets far from any other race. Be able to switch a planet's focus with changes in the game.

Be more planet focused. It seems the AI treats the squares on all it's planets like one big pool to use from and doesn't look to develop planets individually. It needs to be more aware of specializing, like a player does.

Maybe I am just dumb, but I think GalCiv2 has the best AI of any game I have seen (well, maybe not Massive Assault, but that was just basically Chess so you can hardly compare the two). You want to see shoddy AI that basically just blinks at you - Play Rome Total War.

If GalCiv2 has such crappy AI, then PLEASE - enlighten us - WHAT similar 4x games have BETTER AI? Its the nature of the beast that AI is going to have some bumps on it, that is what the 'A' in 'AI' is for

I have seen the AI do some amazing things. I have seen it match me, ship for ship as I built up forces for an invasion while our relations were still cozy (and this was three levels below bright!). I have seen it realize that it is about to lose valuable mining starbases and either sell them to another empire for desperately needed cash, or sell them to an empire to get their support against a common foe. All in all I think the AI alone is worth every penny I have paid for the game.

Its all very well and good to say 'The AI should anticipate my attacks more' or whatever, but context is everything. HOW exactly in concrete terms do you teach the AI that? Is it just distance between your ships and their planet? What if you inhabit the same system? What if you are moving through their system? What if you are allied? What if you just bought ships and they appear there, does that count? What if you are gearing up for a war, but this AI personality is set to a passive personality or something? How should this strategy be cashed out across the MULTIPLE AI personalities that exist in the game? Something that is obvious to a human with intuition, is not so easy to spell out to a formal system like a computer. I am not saying it can't be improved - ANYTHING can be improved if you pour enough resources into it - but from the mere fact that improvement is possible (which is always a given) it doesn't follow that something is shoddy, or SUXORS.

The ONLY weaknesses I can find in the AI (and I am still only playing on one level below Intelligent, and fighting tooth and nail for my victory too) 1) Some of their planets are wastelands, or are filled with useless structures, and 2) The AI tends to play a defensive battle even if it has more ships, which is suicide in a game where attacking is everything.
I think the AI is much better than any other game I have played where I have to think hard about what the AI does WELL.

Just my two credits

Dano

Because I can tell you right now, if you're beating the AI with a decent mix of planets (i.e. where they have several planets) on a decent sized map then I suspect there's an exploit involved somewhere.

I wouldn't call these exploits, I would call these bad military judgements.

* The AI often stack extremly many ships in one parsec, and there is at most one fleet in there. Makes it really easy to destroy a big chunk of their fleet in one or a couple of turns. They should form as many fleets as they can in the stack.
* The AI often have no extra engines on their designs. I always have at least one. This makes it very easy to always attack giving the AI a big disadvantage.
* The AI sometimes mixes ships with no extra engines with ships with extra engines, making the extra engines useless.
* The AI sometimes moves really close to my ships when we are at war but don't attack. I guess he thinks my fleets are too powerful. This of course only makes it more easy for my fleet to destroy his since I get to attack. It would be nice if he could realise that I will probably attack him anyway.

I would be so happy if you could improve these. It would make the AI so much more deadly and the game much more fun for me. The Yor doesn't feel satisfied when it's too easy to eradicate all the "meat".


Now that you will include the no-tech-trade option, I would also really like to see the AI:s research more different techs. For example, I have never seen the AI:s research Xeno Ethics, they have always gotten it from me in some way. And that's a really good technology.


I made another post about military AI in another thread a while ago. Don't know if you saw it so I'm reposting it here. It's basically about the same problems.

Military AI: This is the only problem I have with the game at the moment, all the others IMO are minor compared to this. I never feel threatened by the enemy. Doesn't matter that they have double the amount of ships when most have minimum movement and are quite bad. The AI needs to design better ships. It cares too much about costs and builds too many smaller hulls when larger are available. The beginning of the game is especially problematic. In my latest game, I noticed that all 9 AI races built their heavy fighters with only one component, a weapon, to minimize costs and build time. At the same time. I had heavy fighters with double or triple the speed and more than one weapon per ship. Naturally, they were outclassed and almost never had the opportunity to attack. The AI should be hardcoded to at least add one of the best engine available to all ship designs that are small hull and larger (except their Defenders). Don't know if it's doable with tiny ships since I always play default Yor with 25% miniaturization bonus and never build them for anything except scouts anyway. They need to realise that it's much better to attack than defend. They need to form fleets better, not mix in a heavy fighter (for example speed 3) with a couple of frigates (speed 6). They shouldn't make huge stacks of solo ships in one parsec, especially not 2 parsecs from one of my planets and then declare war. They should always declare war in the beginning of their turn, not at the end, so that they get first strike. Their huge stacks of solo ships should be formed into multiple fleets. In my current game when I attacked the Altarians, he had his huge stack of ships outside of Altaria. He had one fleet in it. I went there with one of my fleets consisting of 4 medium-size ships. I destroyed the fleet and a couple of ships. Pressed end turn. What does the AI do with the rest of his solo ships in the stack? Nothing. He doesn't attack, it's possible that he made one fleet in the stack but that's all. He should have made as many fleets as he could and then destroyed my fleet. Instead, I destroyed the rest of his stack. I don't remember exactly but at the most, I lost one medium ship. And the rest got a huge hp-bonus. I really like this game but the way I play it, evil conqueror, I have no problem to win over the AI at all, this at intelligent AI. Im sitting hoping that the fundamentalists should conquer half my empire or something to give me a challenge. The most difficult fight I have had in this game was with pirates. They actually killed one of my fleets. Then I sent a veteran fleet with extreme high hitpoints and crushed them.

The AI should anticipate my attacks more' or whatever

The thread has contained criticisms of the AI MUCH MUCH more specific than this. Basicaly the people coming to defend the AI have not exposed its weaknesses yet via gameplay. You will eventualy. Then you will be in the same boat as the rest of us of being able to beat the game on at least tough every time.

To reiterate there are at least 4 specific things the AI does very poorly.

1. It spams un upgraded starbases which serve no purpose.
2. It makes terrible use of planets
3. It never makes ships faster than 2 (and also often dosnt use larger hull sizes even if available)
4. It has no clue how to deal with Player military star bases and the fleets protected by them. (These are the key to victory by the way)

Yes it does somethings excellently but at the moment all of these things are totaly undermined by the above achilles heals.

Dont get me wrong, I think this a great grat game. That dosnt meen I shouldnt call a spade a spade. If these basic errors are fixed then this game will be truely excellent.

As for comparing it to the AI of Rome Total War, does that mean that all a developer has to do is create an AI better than the POS AI in RTW and we should all clap our hands. We should never use that AI as a yard stick of succses.

The games I know of that have the best AI are indy titles made by small development teams on small budgets. Namely Highway to the Reich (Amazing AI) and Dominion Wars 2. Is it a coicidnece that indy titles have great AIs and mainstream titles have poor ones, not if the mainstream is using the AI of RTW as a yardstick.

As for comparing it to the AI of Rome Total War, does that mean that all a developer has to do is create an AI better than the POS AI in RTW and we should all clap our hands. We should never use that AI as a yard stick of succses.

Well True enough. I guess it just irks me that, that game got 90+ percent scores, when the most fundamental part of the game was an abject failure.

I haven't played those other games you mentioned ... are the choices for gameplay (Diplomacy, influence, Resource securing etc, etc) as varied as in GalCiv?

Dano

No exploits are needed the AI just isn't as good as GC1. Next game I play will be on suicidal. (been playing painfull)
I think the root of the problem is the way military power is measured. I think the AI's tend to waste resources in a massive arms race every game, but the ships they build can be wiped out by a much smaller good navy.

This gives the human player massive advantages :
1) AI ships are slow, and under powered
2) lower maintenance costs. (build your top of the line fleet when you need it, and it's cheaper)
3) you can build up your planets much too long. (knowing you can arm up latter)
4) you can pretty much always get a tech lead.
5) you can pretty much always grab the super-projects that you really want - (xinthum, -training, eyes, research )

My last game the korx actually had some dam good ships.
But after grabbing they're mil resource, and the torian's, (both used starbase defenses, no fleets nearby) they were easilly defeated.

Perhaps if the AI just evaluated military strength in a more realistic way, it would build more capable fleets and not waste it's resouces.

The thread has contained criticisms of the AI MUCH MUCH more specific than this. Basicaly the people coming to defend the AI have not exposed its weaknesses yet via gameplay. You will eventualy. Then you will be in the same boat as the rest of us of being able to beat the game on at least tough every time.

Sure, I might. But you're just playing the aggressive approach. This game is not just a battle of military might like a RTS, you're losing all the fun. It's the subtleties of it, all the other viable victory approaches. And besides, winning isn't everything, you know, it's getting there that's fun.

That being said, I bet, though, that on the hardest difficulty levels the AI at least gives you a run for your money, and you don't win EVERY TIME with your eyes closed and one hand behind your back like you make it seem. I think you're blatantly exaggerating. And you do it unfairly, and really not in a constructive manner, too.

As for comparing it to the AI of Rome Total War, does that mean that all a developer has to do is create an AI better than the POS AI in RTW and we should all clap our hands. We should never use that AI as a yard stick of succses.

Yup, but you bought a copy of GalCiv2, you didn't buy Stardock, yet. Don't go all-demanding-like and behave like a little brat (pardon me, but that's how you behave) that thinks everybody owes him something, and NOW!!

You've got a great game, with a great AI. And instead of just shrugging and ignoring you altogether, the devs actually GIVE A CRAP about what you have to say, and are willing to regulary improve the AI. Show me ONE game-company that might do it. ONE.

Don't go on pretending to be naive, the way you behave is not constructive. Being constructive is something like: "Brad/Stardock, I found out that the AI does this, and that, when it shouldn't, and also I managed to trick it by doing this and that, too". It's really NOT something like: "This game's AI was just hyped! It sucks, it's lousy, and I beat it without even trying! Look, the AI doesn't even do this and that! It's ridiculous!". Even if you are some sort of a genius, or you got DEEP-BLUE installed in your brain and you happen to actually humiliate the AI on the most difficult of the difficulty levels every time, it doesn't mean you have to demand or insult. Work with the devs, and stop bashing like children.

They need to realise that it's much better to attack than defend.

I dont think the AI should be changed here. I dislike the game mechanism that the attacker has the first shot with his WHOLE fleet. IMO the game mechanism should be changed here. Seek and destroy and engage at any price isnt really an interesting aspect of a strategy game. I like to have a more difficult choice between engage or not engage.

Actually moo2 had initially the same "attacker has 1st shot" problem. They solved it in the 1.31 patch with their ship initiative rule (ini=combat speed+beamattackvalue/10).

Greatmirror:

Sure, I might. But you're just playing the aggressive approach. This game is not just a battle of military might like a RTS, you're losing all the fun. It's the subtleties of it, all the other viable victory approaches. And besides, winning isn't everything, you know, it's getting there that's fun.

Some of us prefer to play the aggressive approach though. So far I haven't had any interest in a diplomatic, influence or research victory and I have always played evil going for a military victory. But the getting there isn't as fun as it should be because of stuff mentioned by me and others.

moo2niac:

I dont think the AI should be changed here. I dislike the game mechanism that the attacker has the first shot with his WHOLE fleet. IMO the game mechanism should be changed here. Seek and destroy and engage at any price isnt really an interesting aspect of a strategy game. I like to have a more difficult choice between engage or not engage.

I agree but if they don't want to change this, the AI needs to learn to attack more.
If they changed this however, then engineless fleets would suddenly become more dangerous because they wouldn't be crippled before even getting to fire.

Some of us prefer to play the aggressive approach though. So far I haven't had any interest in a diplomatic, influence or research victory and I have always played evil going for a military victory. But the getting there isn't as fun as it should be because of stuff mentioned by me and others.

Great! Then help improve it, you actually have the opportunity to! But really, nobody promised you a perfect and unbeatable AI. Nobody promised you anything, for that matter. Stardock just tries to bring you a fun game, and I believe they succeed. Charging here like someone LIED to you and there's really no AI in the game at all, just because you found some flaws, is...well...not constructive? To say the least.

By the way, I'm refereing to the nay-sayers in this thread in general, not anyone in particular.

WTF are you talking about.
Have you actually read my first post (#39) in this thread?

If Brad finds it offensive in any way, let him tell me himself and I will gladly apologize.

Great! Then help improve it,...

Have you read his previous posts? It doesnt seem so.

ignusss, I didn't actually mean you in particular. You're quite alright. I meant the nay-sayers in general in this thread.

I want to say thanks to Stardock for making a great, highly addictive game and especially to Brad for creating an excellent AI.

I think the people with destructive criticism have not the slightest idea of programming. Reading some (negative) comments here makes it seem like it's the easiest thing in the world to create and balance a challenging AI regarding every possible aspect of a complex game.

I'm looking forward to V1.1 . Keep up the great work, guys.