Hey, at least they made a fleet - for this AI, that's progress.

I've put GC3 on the shelf for now. I don't declare war because mowing down the AI is like drowning kittens. I don't research because I'll unlock something that 'lets' me update a hundred starbases - one by one. I don't... well, I just don't enjoy playing it anymore. There's only so many pretty ship designs I can make before boredom sets in and I think I'm at that point.

Try Godlike....

No but it'll be more of a challenge till they fix it possibly?

The AI put together a larger fleet than you put together but your complaint is that you designed better ships than it did right?

Try playing on a harder level.  

The team is looking at improving the AI unit design system but as you can see, it's not nearly as good at mix/maxing as good players.  That's why there are multiple difficulties above gifted).

Do the levels above gifted make it smarter though or does it just get more resources to do stuff with? Its not that I built a better fleet but that my fleet is 20x better and I have no resource advantages.

The AI seems to get a lot of flak for many issues not really connected to it at all. I would argue that you creating the optimal fleet with Missiles is the main issue here, not the AI stuff. Missiles are OP when you pick the right techs/modules, and should be nerfed. And lets not even start with carriers

Research is also out of whack. The last military techs are way too cheap and OP, and too easy to get with a research focused empire, leading to exactly to what you see there. If they tweaked some of this stuff, the AI should do much better. I also think the AI should get a scaling logistic bonus tied to difficulty to create bigger fleets to compensate for player micromanagement with ships. (I will probably mod this myself eventually if they do not fix the late game)

I was talking to Naselus the other day about giving the AI more bonuses to make it more difficult.  Naselus said that it won't work because the AI DOESN'T KNOW how to utilize the bonuses that it gets.  

The problem here is simply the AI's inability to set production at a planetary level. It will never be able to keep up with the player late-game, even with the huge bonuses on offer from higher difficulties. 

Say both you and the AI have 3 planets, each with 100 production. You give each 900%'s worth of buildings, one specialized on econ, 1 on industry and 1 on research.

The player specializes each planet to use 100% on the correct output type. So you have 1 planet producing 1000 industry, 1 producing 1000 research, and 1 producing 1000 econ.

The AI uses the empire-wide wheel, and set itself to 33% on each. The industry planet produces 330 industry, 33 econ and 33 research; the econ planet produces 330 econ, 33 industry and 33 research; the research planet produces 330 research, 33 industry and 33 econ. It's total production from the 3 worlds combined is 396 manu, 396 econ and 396 research. 

If it tries other settings, then the results are not improved. If it goes for 50% manu and 25% elsewhere, we get the industry world on 500 manu, 25 econ and 25 research - a nice increase for industrial output. But the econ world is now producing just 250 econ, 50 manu and 25 research; the research world is only producing 250 research, 50 manu and 25 econ. It's total output is now 600 manufacturing, 300 econ and 300 research. It's still stuck on 1200 output, where the player is pumping out 3000 total output.

You're producing more than 2.5 times as much output as the AI from the same buildings, planets and population. Even with the tripled production that the AI enjoys on Godlike, it is only just able to match the basic specialization that the average player manages using the wheel alone (even without any advantages from optimal improvement placement etc). This means it starts falling behind the moment the player begins specializing planets.

This is actually exacerbated by some of the other features - so, for example, the AI avoids going into debt if at all possible, by setting the econ slider to match outgoings (when set to breakeven, or when reaching the 'breakeven' point on a spendtilbroke setting). This is entirely useless on manufacturing worlds, which would be better off sticking with industry and just pointing it all at an economic project. The AI here is doing the right thing - avoiding bankruptcy - but going about it in the wrong way, by re-allocating highly productive workers into tasks they lack the infrastructure to do usefully - it's making the guys on the factory floor sit down and do the accounts while the machinery sits idle. Then there's some basic other problems - the way the AI picks which build queue to use on planets is based largely on what special the planet has, and research specials are rare, so it under-produces research-dedicated worlds; it's research choices are (in most cases) sub-optimally weighted; many of its blueprints are outright broken; the entire BestDefense thing is conceptually flawed.

As to the bonuses the AI is granted; the scripted nature of the strategic AI, combined with the lack of difficulty triggers (so the Easy AI uses the exact same script as the Godlike one), means that the AI doesn't make efficient use of the handicaps. It cannot mass-produce colonies in an effective way on Godlike without scripting it to rely on bonuses which do not exist at lower difficulty levels, and so would be completely useless on Normal or Easy. 

This is very frustrating, because the actual scripting of the AI is not difficult. Most players could likely figure out how to write a decent AI script given 30 minutes or so and a very basic thread outlining what the different script options do. That means that, as a community, we could probably produce challenging AI scripts for 500+ turns on all different difficulties in the space of a week; and Stardock could take this work and use it in an official patch - but we are unable to do so based purely on the primitive strategic AI triggers in the xml. Just adding in a difficulty check would mean that we could make them take advantage of the bonuses more effectively.

TL;DR: The AI will continue to perform poorly until it understands planet-specific economic settings, and handicap bonuses do little to help the situation because it cannot be made to understand that it has them. If the AI is given a means to specialize planets, then we will likely immediate see it performing 2-3 times better than it presently does.

I've enjoyed Gal Civ 3 but have parked it since 1.1 primarily due to the AI.  It's good to see there are more improvements in 1.2.  Hopefully Gal Civ 3 will eventually reach a similar standard to Gal Civ 2 and once it gets a lot better I'm looking forward to plenty more time in-game.

I fully support the great posts above from naselus and marigoldran, you've hit the nail on the head, if Frogboy focused on those aspects next it would have a massive impact.

Speaking of specifics as suggested by RevTKS, this video series from Macsen is a good reference to mine ideas to improve the AI.  Macsen shows in detail how to hammer the AI on Suicidal without breaking a sweat and as far as I can see there is nothing that could be considered a cheat or excessive exploit either.

https://www.youtube.com/playlist?list=PLHMuUHsmNFlQEvNuuDLZ2uBLJ0HNEY2si

One last point ... have Carriers been balanced yet?  They make the game strategically shallow.

Yes. It did this through massive MASSIVE resource bonuses.  

Hence: Increase the difficulty level. 

The AI isn't making dumb decisions (as the original screenshot showed). It simply didn't have *enough* stuff.  He's playing on 1 difficulty level above normal. 

Full disclosure: I work in the same building as the AI coder of Civilization IV and Civilization V. 

Seriously? Holy crap, Batman! 

With all due respect, this is not entirely the case.

One could play against the Civ 4 AI on normal difficulty (i.e., equal playing field), and still find that it could match you economically when playing 'normally' (i.e., no massive min-maxing). Even when playing GC3 with minimal effort (not caring about adjacency, never replacing buildings, no rushing techs, no special colony ships or sensor boats etc), the AI does not provide anything close to an equivalent challenge on normal difficulty. I used to play normal Civ 4 when I wanted a bit of fun in between 'serious' games, and I'm not ashamed to say that it occasionally managed to get the drop on me. Playing normal GC3 isn't fun; it's dull, because the AI is utterly ineffective by turn 100. The whole game is basically just an exercise in mopping up.

There's a bunch of reasons why. The way production worked in Civ 4 made it harder for the player to ramp up his economy as dramatically as you can in GC3 (for example, growing your cities in Civ4 usually required that you sacrifice some hammers for food, and became harder to grow as time progressed; in GC3, it largely grows of it's own accord at a steady rate). The maintenance system has been written about a lot regarding this too, and for good reasons - it worked very well in a way GC3's limiting system (LEP) simply doesn't. Sending out a new settler was extremely expensive - which, once again, it just isn't in GC3 right now. And, of course, simply the structure of the economic system, where research and economy were closely related and industry less so made it easier to prevent rapid growth on multiple fronts at once.

But overall, point 1 is that the mechanics of Civ4 limited the rate of expansion far more effectively, preventing the player from getting a huge lead early on. You simply couldn't colonize half the map by turn 100, whereas this is easy to do in GC3. This is a fairly big problem in the present game mechanics, rather than an AI point; 1.2 has clearly introduced some efforts to contain the player (the increases in killer tech costs, for example), but these actually act to nerf the AI, which was already pretty weak at picking up those techs. Mechanics need to be introduced which limit the upper end of the power potential, without damaging the lower end. This is hard to achieve but things like nerfing adjacency bonuses into the floor, a tech-spread deflation effect (so the second Civ to research it gets it for 90% of the cost, the third 81%, the fourth 73% etc), and removing the artificial production bonus on colony capitals would be effective.

Point 2 is that the AI was able to specialize its cities effectively. While its improvement placement was often weak, it still visibly created industrial powerhouses and economic supercities. This meant that the AI getting increased production levels on higher difficulties genuinely mattered. There is a colossal gulf between the relative power rating of a Deity-level AI in Civ 4 and a Godlike AI in GC3. Deity-level AI was defeatable, but required the player to min-max aggressively - he had to be able to squeeze every last drop of economic power out of his empire to even stand a chance. This is not the case against a Godlike AI. As shown above, even with it's 3x production bonus, the non-specializing AI is barely increasing it's economic output over what the player is getting from the exact same resources simply from moving the wheel. This is a very, very basic part of learning to play the game, and even most new players will start doing it from their third or fourth game.

Point 2 is the big glaring AI problem, and it means that the AI in GC3 as it presently stands can only match the player on production at the very start of the game on the highest difficulty level. In Civ4, the highest difficulty AI had similarly enormous bonuses, but at the very start of the game it had 500-600% the player's output of EVERYTHING, rather than a miserable 20% higher. As the player sets up more and more specialized worlds, he inevitably overtakes the AI's gimped economic control, regardless of difficulty; the only danger from a Godlike AI is that it might just manage to colonize more planets than you before it runs it's empire into the ground.

There's some issues with the tactical AI still. But at the moment, the planetary and empire economic AI is the by far the more pressing problem, and upping the difficulty to the point that the AI is no longer really playing the game anymore is not a solution. If just a few of these points can be tackled, then we'd be looking at a really impressive AI already - in fact, that's what's so frustrating about it. There's a great many things which the AI can do really well - hell, as you noted before, in the case of this very thread the AI was making good decisions but had been crippled by lack of resources. And that lack of resources is entirely artificial. It was not because its been out-colonized or because the player was better developed on all his colonies, but because the AI cannot not use it's planets to remotely the same effect as the player can. I fully expect the player to use planets 20-30% more effectively; the margin is currently more like 250% more.

... the fact that you're playing 2 difficulty levels above normal without ever touching the economy wheel kinda illustrates my point about the AI not leveraging it's handicaps. Playing without ever looking at your economy should be something that's barely sustainable on Normal, let alone Gifted or Genius.

EDIT - and seeing some of the actual production graphs - population, raw production, and research, for example - would be more useful for identifying where the AI screwed up. I usually see them tailing off hopelessly in research fairly early on.