GalCiv3 Forums
Right?
The developers are too focused on AI tactical issues. I believe this is misguided. Tactics don't matter if your strategy is flawed, as the Bush administration discovered in Iraq.
How else should it react to the players strategy? It already reacts if you're a war monger, if you're near victory, to your ideology, and to your military strength.
1. Gang up on a dominant player that is obviously a warmonger.
2. Respond to random floating transports in their zone of influence.
3. AI analysis algorithms are badly flawed. It responds to CURRENT MILITARY STRENGTH, which is totally wrong because what matters more is INDUSTRIAL/RAW PRODUCTION.
4. AI should build up fleets PRIOR to declaration of war and send those fleets out at you immediately post-declaration.
5. It shouldn't want to sell planets for 99999 credits.
6. Economically the AI just can't keep up against a decent human.
#1 happens already because of the large diplomatic penalties you take as you conquer other races. Though admittedly I play with a lot of custom civs so I'm not sure if it kicks in soon enough with just a few civilizations in play.
#2 I'm not sure about, but during war its pretty good at attacking transports that I send in undefended.
#3 I sort of agree with, but I wouldn't call it badly flawed. In beta civs would treat you like you were powerful without you ever building warships because you could get a really high power score based on just your economy and starbases. That led to games where if you weren't going for a military victory you'd never bother building warships. Going back to that would be bad, but I think it would be better if they weighed economy and and to a lesser extent starbases a bit more again.
#4 I seem to remember Brad talking about that. I'd have to play some games with fog of war disabled to know one way or the other. If it doesn't happen I think it's coming though.
#5 Why not? If the AI offered that to me for research planet #15 I'd take it.
#6 That's what the handicap is for. Up the difficulty level.
1. Large Diplomatic Penalties are pieces of crap. It doesn't actually change AI economic behavior. The AI is INCAPABLE of preparing for war. It can only begin war preparations AFTER declaration.
2. That's why you station transports next to their planets BEFORE DECLARATION OF WAR. (Duh). Clearly you have not thought of this. Upon declaration of war, you invade all of their planets in one go.
3. AI evaluation schemas are pieces of crap. The AI can't tell apart between a barn and a cow. If you aren't going for an influence victory, why the hell is the AI even bothering to consider your influence in your power score? The way they calculate power score is badly incompetent. For example, did you know that the AI calculates score based on INDUSTRIAL PRODUCTION, NOT RAW PRODUCTION? Consequently, if I change my economy wheel from 100% research to 100% industrial production MY POWER SCORE LITERALLY DOUBLES. The reason the AI declares war and then asks for peace 1 turn later is because OF HOW INEPT THE POWER SCORE IS AT ACTUALLY CALCULATING POWER. The AI declare wars on you, you turn to 100% military production, your power score doubles, AI asks for peace next turn.
In other words, the power score is rubbish.
5. Because if you can do it once on the AI, you can continue to do it on the AI. You can buy ALL of the AI planets with the exception of their home world with money. They'll sell ALL of their planets to you as long as you can generate the cash.
6. I play on Suicidal. I CAN'T TURN UP THE DIFFICULTY LEVEL ANYMORE YOU SILLY PERSON. It's much, much, much too easy.
It already does the first 4 things (though with a caveat that different ideology players won't gang up together on someone)
Ok, I'll give 1.2 with Naselus's mod a go later this weekend. Even with his massive nerfs to my playstyle (for example changing LEP to -0.2 money/planet and nerfing Thalan Hives and Colonizer) due to his mod I'm finding in the prior versions that the AI still can't match what I'm doing on suicidal. However, I haven't tried 1.2 yet.
Will play the game on a medium map first, then Insane next. Will let you know the results.
Your sentence order is wrong. I think you meant "AI should build up fleets and send those fleets out at you PRIOR to declaration of war." The "immediately post-declaration" is the part when you attack and invade. 
More seriously, the AI and/or the supporting strategy scripts ought to understand and use planetary specialization. I think, that's the biggest issue crippling the AI currently. Lots of other things are simply the consequence of that as the AI's economy and production simply can't keep up with that of the player. There's some effort at specialization from the build queues but most of that is wasted because the AI insists on running on global economy settings.
I'd suggest killing two birds with one stone: Implement planetary governors. If done right both the human player and the AI could use them, and this would simultaneously solve the problems of micro-management and poor AI economy.
However, the "done right" part is the crucial thing. Here the key is to notice that simple really is better. A planetary governor, as its name suggests, needs only to think about his singular planet and he needs have no other goal than "maximize target within set limits". In fact, he should never think about anything else than that one planet and how to optimize it for a given target. He shouldn't even think about build orders leaving that part to the player whether human or AI. The governor gets a "mostly ready" colony handed over to him and his one and only concern after that should be: "maximize target within set limits"
The basic principle is thus quite simple:
- The human or AI player builds a colony up to some level he thinks is "sufficient for now" and sets specialization target (manufacturing, research, economy, people, influence) for that particular world as well as putting the social/military slider to a level the governor will use as the "normal" level.
- The planetary governor would then each turn adjust the production wheel, social/military slider, and project so that it maximizes the specialization target output.
The output is a smooth continuous function of the wheel setting, and usually with just one peak, meaning that the max optimization can be found with simply just moving towards increasing value provided you try it with all available projects (if the build queue is empty). The projects are needed because in some situations "max production + project" gives a better result than "max wheel setting". The limit to the max optimization comes from the social/military slider's "normal" setting (from the player or upper-level AI) but the governor is allowed to shift it if production (whether planetary or yard) is overflowing a "built in 1 turn" situation.
It's simple and brutal but would be vastly more effective than the current AI planetary setup plus it would also eliminate about 95% of micro-management that the human player currently does in the "Govern Planet" screen.
Most of the game on the medium sized maps roughly boils down to where I can get 6 colonies (or 8 depending on if I use the constructor into colony ship trick) by turn 15.
On games where I can it's basically an auto win even with no tech trading. In games that I can't it becomes more difficult (but not that hard with Thalan tech).
Malevolence invasion line + Thalan tech is the key on those maps. On average I can wipe out two AI capitols with that. The other transports are returned home for a massive population boost.
Petri Kokko is basically right.
1. Better AI specialization of planets + Control Group of planets for the production wheel. The production wheel is currently ALL or ONE. Instead, allow players and ALSO THE AI to make control groups of planets, allowing the adjustment of the production wheel for all members of that control group. This will help with AI specialization and planning tremendously AND cut down on micro.
AI research worlds for example can be 50% social production (for the building of research labs) and 50% research. AI manufacturing worlds can be 50% social and military production for the production of factories and constructors. Once the AI can't build any more on that particular planet, it'll turn it to 100% military manufacturing or 100% research depending on planet type. This would TREMENDOUSLY improve the AI and make it competitive with humans.
The rest of economic management is just a matter of finding the right balance of research and production worlds for a given AI strategy. The AI strategy could be:
a. Rush attack, in which case research will be prioritized towards weapons and planetary invasion.
b. Expansion.
c. Starbase spam and economic buildup.
2. Improve the AI blueprints and prioritize them to build ships with movement speed greater than 8. OR an even better idea would be for the AI to have TWO or THREE different ship blueprints- some ships for planetary defense (almost no engines) and a second class of ships for offense or space defense (more engines). I'm tired of looking at AI ships moving at an average rate of 5 spaces/turn. By the time any of those pathetic ships reach my worlds I've already got more than enough defenses online.
Once those two basic features are fixed, the rest is just a matter of writing specific strategies for the AI.
Interesting topic the AI of Galciv3..it seems to me that the AI is based on a buildcapacity of a race in social/militair building .what can i race produce in terms of social engineering and militair engineering ?
Militair power is the most direct way to rule the galaxy... when you can destroy enemy fleets and shipyards and do planetary invasion to invade the enemy planets ..you regain influence and enlarge your territory ( zone of influence )
A conquered planet gives extra money too.. as soon you get one conquered planet your empire is growing
So the Ai for races is focussed on maximize on what ? : to get as many colonized planets as possible ?..or in getting a militair strong fleet ?
Welcome Guest! Please take the time to register with us.