To be honest ive never touched the wheel because I forgot it even exists. Im a noob, but I had such an easy time on normal I figured id give myself a genius level challenge and im still kicking ass.

Yep.  Soren (Civ III/Civ IV) and Brian Wade (Civ V) talk about AI stuff all the time.  The biggest mistake GalCiv III makes is that it tells the player too much about what bonuses it gets.

Way better to hide it.  Then I could have simply made "gifted" get more bonuses and strategically unhide planets.  People mistake good AI for challenge.  This thread demonstrates that clearly.  The AI IS doing a good job at the AI part.  What it's not doing a good job of is managing its economy well but that's less an AI thing and more of a scripting thing.  That can be easily cured by giving the AI a higher handicap.

I didn't write the AI in GalCiv III (I've helped do some stuff in it like in v1.2 I redid the way it reinforces its fleets) but I do know that the first thing I'd do if I was working on it would be to dramatically bump up the boosts.

It's easy to make an AI that beats players.  The hard part is making sure it does so in a way that doesn't make the player feel like they lost due to flagrant cheating.

Or by enhancing the scripting tools available, since they're presently very basic. Scripted AI doesn't have to be bad, and doesn't necessarily need massive bonuses to provide a challenge - take the Space Empires 4 AI scripts that the community came out with. They're brutally effective, and do not cheat or receive bonuses unavailable to the player. Of course, those were the product of literally years of work by an insanely dedicated community... which is something we actually have to hand; unfortunately, they can't do much when the script hooks aren't available.

We, and I imagine Stardock's own scripters, really need a lot more stuff that we can direct the AI players to do through scripts. Local production control is one of them, but also lots of triggers for things that might allow us to predict and react to game situations. These are presently barely available. I would like the AI to react at a strategic level to discovering it's next-door to an Aggressive civ - or at least, be able to to check that it has a neighbour at all. It's spending priorities should be different depending on it's situation. We need to be able to check map size; we could do with being able to check difficulty (godlike AI can have much, much more aggressive scripts than normal can cope with).

At the moment, we can check victory condition, present turn, personality, war status and tech age. I've written a dozen separate AIs, based on personalities; I've extended the AI to turn 400+ with 15-20 different 'phases' per personality; I've even given different personalities differing war strategies. I can't really do much more than that though (aside from tech choices, but tbh I'm dreading trawling through all the tech files ). And that's hugely frustrating; I can see 3-4 simple things that would drastically improve the AI's play, and which are so close to in reach. 

Please, give us more scripting options in the xml so we can help improve the strategic AI

The AI is bad, regardless of difficulty. That much is clear. I only play on Godlike now because the AI is just waiting to be steamrolled on every other difficulty.

The AI really needs to learn how to specialize WITHOUT neglecting their other improvements to become more competitive. For example, they need to stop building influence improvements in all of their tiles on all of their planets near ZOC borders. All I need is one planet on Cultural Festival with about 2~3K social manufacturing to overtake the influence growth of their entire faction. Influence improvement, besides the ones that generate other things like Ideology, are the worst improvements one can build.

Since they are also stuck with default ships, the core designs need to be greatly improved. I have never seen an AI ship with "The Bane" even when everyone has all the techs researched. Their weapons and defenses are also too specialized, making them extremely exploitable, and they never change their game. 4K Kinetic and Shields but 0 Point Defense? Let me just throw a few cheapo missile ships at them to mow down their fleets before they can even come into range to counterattack. I can solo the entire map with one ship by endgame. It is just that bad.

Research could probably use a fix too. The AI does not realize that research improvements become obsolete and should be replaced.

I vehemently disagree.  The UI must be fixed before anything else.  You can increase your difficulty and change the options.  (Such as a more crowded galaxy.  Putting yourself so far away from the AI makes it easier on you, not the AI).  You're making it artificially easier on yourself and complaining about it.

On the other hand, you can't undo the boredom of the clickfest with any setting. 

I thought there's pretty much where WoW has been for the last five years too, tbh

The problem is I crush the AI LONG before turn 100.  I mean, most of my games end on turn 50 when I'm like "ok the rest is just mop up, time to do something else."  As I see it, if you haven't won by turn 100 you've done something wrong.  

As long as I can colonize up to 6-8 planets I've won the game.  

Hi Marigoldran

Maybe you are just too good for the game setting.

Try:

• Medium galaxy spiral map
• Common star systems
• Common planets
• Habitable planets to a be uncommon
• Ultimate planets to be rare
• Around 12 CIVs on at least Gifted difficulty
• No minor races (these are just free planets for players)
• Regenerate map until you are located in the center of the galaxy
• Set resources to be rare as well so the AI snatches these from under the players nose and they do not give a big advantage in the beginning.

This way you can still get 6-8 planets but you will be stretched out if doing so (there should not be more then one habitable planet per star system). The AI will also have a blast if the player does not build up military tech and ships fast enough and if the layer is surrounded.

If all this does not help, simply create new CIVs with aggressive opportunistic traits (just like a humie ), give them a lot of penalties for economy related stuff (they will get these back with higher difficulty anyhow) and give them more research, logistics, hit points and speed

All this will not help if you tech up mid-game like discussed above but should build up quite some sweat for a player until around turn 200

These settings are begging for culture flip.

@wpkelley41 I've been tweaking the map settings a lot with my mod, and I can tell you from experience the difference can be huge. (but keep in mind that lots of stuff has changed due to patches) The number of starting planets is probably the best "booster" you can give the AI. However, that does not change that game is currently somewhat broken, and how you play the game will massively effect how efficient the AI is. This means you can get completely random results depending on the map as well as your play style.

Back on topic, I think we can complain about the AI all we want, but that does no good at this point. There are way too much broken stuff that is unbalanced breaking the game, making beating the AI too "easy" for us, not the AI being underpowered or stupid or anything. The 1.1 patch "buffed" our potential power too much compared to the AI sort of. 

These are the issues I think needs fixing, before we even start thinking of boosting the AI with resources and whatnot:

-Tech trees needs rebalancing, both in cost and power. Thalan tech compared to the others is insane. I think the Thalan design was intended to be balanced with the negative growth, but that essentially does nothing due to abilities being useless at the moment. 

-Starting abilities are a complete failure at the moment, and not balanced towards the current resource design. Hence the negative growth not working etc. Custom races as such makes a huge difference for a competent human player vs an AI.

-Ideology paths are too important at the moment. The Malevolent Death furnace is an insane raw production boost when built correctly, and nothing in the other trees comes close. (except for the flip thingie in Benevolent maybe) The AI has no clue how to utilize this stuff correctly, and is as such heavily penalized compared to the human player. I think this is fine for normal AI. We should not expect most people to be fully exploring this mechanic until they approach harder difficulty levels, and as such making the AI do it makes no sense given that it scales with bonuses. We need a linear growth in difficulty, and if the AI gets scripted to this, it gets exponentially harder instead. Some rebalance is needed here, but it is probably the least important on this list.

-Endless projects is most likely the real offender here. Birthing subsidies is too powerful, and the basic mechanics too weak. Same with influence. You don't even need those fancy culture buildings, just raw production and a project. This is probably the most broken part. When a birthing project creates more citizens per turn than the ultimate Yor assembly line, something is very wrong. Fixing these, and a review of growth mechanics would probably fix a lot of the current broken stuff. 

-Weapon choices are too impactful, the right specialization techs, coupled with the right modules, makes the fights ridiculously lopsided. The AI can't compete at all as OP shows. The basic resource less weapon modules should probably be nerfed given their massive effect. (either by cost or impact, there is no "real" decision to be made regards to them) Boosting your weapons that much should cost strategic resources or something to balance it. Again, massive human boost if you design your own ships properly, AI should not be expected to compete here in my opinion, the balance is just off. This should be the place you go to to defeat the AI as you increase difficulty, so some possibility of min-maxing should be retained, but the current gap is too big. 

Until these issues gets fixed, we can complain about AI all we want, but that is not the problem here. The game and us are. We are currently abusing imbalanced stuff to steamroll the AI. I think the best example of this is the 200 turn win steam achievement. I guess that was meant to be a challenge to play a fast game, rushing on a small map or something? I've usually researched the whole tech tree by then, and that was probably not intended to happen from a design point of view? I think someone on SD turned on godmode for us by mistake with the 1.1 patch, and most people haven't noticed it yet, and here we are complaining about the AI instead of the core issues? I've seen progress with the AI, so something Frogboy has been doing must have been right.  Fixing the stuff with a mod is possible, but would be a massive undertaking, so hopefully someone is still in charge of the game over there able to get this game on track again.

Um... not particularly massive, I addressed the majority of this list for IAB weeks ago. For example:

Takes approximately five minutes to fix, provided one has spent a couple of hours learning the structure of the xml and understands how the game works. Give yourself maybe half an hour to think of suitable bonuses, though tbh if it takes more than 10 you probably don't understand the game well enough to be making changes to it.

The trees are actually pretty well balanced already tbh - there's about a dozen abilities which are either too poor or too good, but fixing them is again a five-ten minute job. Death Furnaces, for example, can be brought broadly into line with the other ideology traits just by nerfing the level bonus down to 1%/level. Much of the rest of the three trees can be made workable just by flipping a few flat values to multipliers and visa versa, or just changing how many planets are effected.

Took me most of the evening one day after work a couple of weeks ago to plod through this - we'll call it four hours. Which included writing up the spreadsheets, changing the necessary values, re-balancing Defenses in line with the weapon changes, adding in an extra 9 weapons, figuring out how the localizations work, and re-balancing the blueprint defs.

Takes less than five minutes. Just change them from multiplier bonuses to flat at a 10:1 or 20:1 scale, so that 1 manu gives 0.1 research/econ. Influence can be given 0.01; birthing subsidies 0.1%. The real problem with projects is not the conversion rate, but rather a matter of manufacturing, research and econ bonuses being silly-plentiful and it not being particularly hard to get Food over 100. I suspect the devs thought people would build 2, maybe 3 farms, rather than 8 or 9. They need to either harshly limit food - so getting over about 30 population is nearly impossible - or aggressively slash building bonuses; either would help the game enormously (I chose the latter). 

Seriously, none of the above are big issues. This one:

Is just about the only time-consuming thing on your list, and that's largely because there's ten files to go through.  Though tbh, even in this one's case it's not actually hard; most of the big advantages that the Thalans get are just OP buildings, which it's very quick and easy to change to something more balanced. You can deal with the worst-offending bits in the course of an evening - which again includes taking the time to spreadsheet them - and it might take several hours to do a proper re-balance. 

In sum, this whole list is about one weekend's worth of work. For some context, when I was modding Victoria 2, getting the game to a remotely balanced state took around 2 month worth of man-hours, and I suspect the total time sunk into the project by the end was probably nine months to a year. Most of them mine; I was working on it 18 hours a day for about three months at one stage, and the mod was in development for three and a half years with a team ranging between 2 and 10 people. THAT was a game with some serious balance issues. GC3 does not begin to compare on that; what we have here is a largely working system with some fairly minor balance problems and a strategic AI that can't play it.

There are some serious structural issues in the game mechanics, which will take a lot of time to work through - I touched on a few in the post above, but also things like the way growth works, the fact that building bonuses are too big, the fact that level bonuses are insanely overpowered compared to the already-overpowered buildings, the fact that LEP is presently little more than a flawed concept poorly implemented, and the resulting imbalance between 'tall' and 'wide' - but nothing on your list really qualifies for that; it's more like one day's worth of work for a junior scripter. The strategic AI scripts are a way, way bigger problem, which is way, way harder to deal with, and will require way, way more work. Worse, fixing some of these real problems actually requires fixing the strategic AI - to introduce any meaningful penalty to expansion or reduce building output, for example, will cripple the AI in it's present form and basically requires re-writing all the scripts from scratch (which, incidentally, took around 6 hours to do, including expanding it from about 20 scripts to about 200 - probably about the same amount of time it'd take to do the biggest item on your list).

Tweaking the numbers in the xml isn't the hard part. Finding the right numbers is. You would need to balance it for different map sizes, total habitable planets etc. The current fixed tech cost is probably not working properly, and some LEP tech cost scaling might be needed to make research harder and the snowball effect less. Testing stuff is probably by far the largest workload of such a mod. I'm not sure it's even possible for us to balance this version without making some major game design decisions as well, which would change the game a lot, and make it more of a standalone mod, and not a balance patch. Hence the massive part. 

I disagree about the minor balance problems. They are major, and are the cause of the current "AI" issues. I made the list simple to understand on purpose, but there are lots of room for details to be discussed on what to tweak and how much. I see exploits (missiles and carriers), and raw production scaling gone exponential breaking the balance of the game. We do not want the AI to play that same game with us and get bonuses to compensate (it would be a random mess depending on map settings with loads of variation in performance). As you mentioned, many of the issues are minor on their own, and easily fixed, but together they affect the game in a major way.

The AI scripts/behavior is working okay at the moment. (not saying there isn't room for improvement) The current iteration of scripts/behavior is working somewhat close to what we should expect by now. Specialization is the key to victory for humans, the AI goes more macro and generic, and does micro mistakes we can abuse to win in even with their bonuses. You know what to expect from the AI when you up the difficulty, it get's faster tech, more aggressive and more ships. This is how it should work in my opinion, and is probably getting close to what the CiV AI does. We don't need some crazy smart AI to provide a challenge for the basic game. (though being able to mod it would be sweet, but this is not what we should expect from SD itself)

However in this game and current patch, the human gets research output going too early and too easily if you know the right path to take, the weapon specialization techs and modules are too good when min-maxed (trading specialization techs intended?) and makes space fights a joke, and projects makes for easy victories, whatever the condition you are going for. This makes the game to easy to beat by itself, with no relation at all to the AI whatsoever, and that is the main issue here.

All this needs to be scaled back for the AI to provide the challenge it has potential to, and that is what my list try to do. (small steps) Try editing a CiV map to give you a crazy start, and just notice how much easier that Deity AI is too beat. This is the game we are currently all playing. We have been given ridiculous bonuses we do not need and are worse off for it, and somehow the AI gets the blame. Anything on my list should not affect the AI much, since it currently makes no use of these "exploits" at all, and I do not think we should expect it to either, that is how we are supposed to beat them with their bonuses after all, with smart micromanagement. The rewards for doing this is overpowered currently though, and that is the root of the issue here.