Hi all. Long time GalCiv player, rare forum poster.
I really like Crusade. A lot. I've put 339 hours into it, I'm surprised to find out, and am starting another game after this post.
But, it is too easy. When I got the expansion, I played on normal difficulty, and then raised the difficulty one setting each game until I was at "godlike." In previous GalCiv games, this setting was significantly difficult to beat. A lot of games I would lose. But with Crusade, I have yet to lose a game on godlike. Last night I won a game even though I could only colonize three planets at the beginning and had no durantium.
And it is not because I'm profoundly good at these 4x games; rather, the AI is making some major mistakes. When playing godlike, the AI has major advantages early in the game. It will have more colonies and far more resources than I do. It should attack me and end my upstart civilization. But that's not what happens.
Some of this material is probably mentioned elsewhere in these forums, so I apologize ahead for redundancy if that's the case.
1. AI does not seem to research improved weapons (deathray for example). If I research the higher weapons, a significant commitment of time, I can be assured to be the only civilization with this technology. In the late game I may face an enemy "huge" ship class with 500 hit points, but it will have six lasers and only do 6 points of damage. My huge ship, will have deathrays and 200 attack, plus loads of defense and assistance tech. I have yet to fight a battle in which the AI has end-game technology despite having a massive tech advantage most of the game. But it needs end game technology since there are a limited number of resources. The beam weapon track, for example, takes one elerium per weapon. If the AI only researches the first or second tier, then it is doing 2 damage per elerium. My deathray does 20 damage per elerium before any adjustments. Since elerium is limited, I will win.
2. AI does not attack with more than one transport. In the early game, when the AI has the advantage, I will often get attacked by a fleet with one transport and about ten escort ships. I see it coming. I rush two garrison units. I win the ensuing planetary battle. I may be left with no legions, but the AI won't bring a second transport even though it should/could immediately. If these early rushes were more deadly, then I would have to adjust my strategies: maybe I would be more diplomatic, or heed a threat and give an AI what it wanted, or get someone to help me. What I usually do is defend the first transport, while building my fleet. Eventually, because I lose no planets, I will win.
3. AI does not hire mercenaries that improve planets. The most powerful mercenaries are not the battle mercenaries, but rather the mercenaries that impact a planet. The tech mercenaries are especially important. But the AI never hires these. It only hires the mercenary ships that carry combat bonuses, leaving the best mercenaries for me. And I do not have to prioritize these hires because I know the AI will never hire them. Mine all mine. So I get the mercenary that gives raw production, then the one that gives percent increase in raw production, then the one that gives percent increase in manufacturing, and place them all on the same planet. The AI could do this earlier (it has the resources long before me), but it doesn't. Eventually, I will win, especially with the tech mercenaries surrounding my dedicated research planet.
4. AI does not attack my defending mercenary fleets. Those three mercenaries I hired? They should be killed asap. But the AI ignores them. This behavior is so consistent that even if an enemy fleet is next to my planet, I will not bother moving my defenseless mercenaries. The AI simply will not attack them.
5. AI does not appear to use espionage much. I've had one technology stolen, and four planetary resources hobbled by spies in the above 300+ hours of gaming. And I don't bother with defensive spies. Instead, I usually get about 8 or so spies, get ultra knowledge on all factions, then steal relentlessly. Meanwhile, the AI does nothing. It is not provoked by my constant stealing, it does not appear to defend against my constant stealing, and it does not turn the table and steal back. I grant that I have a huge expenditure in spies, with about five or six citizens committed to spying, plus a few free ones. But the AI would really only need one or two citizens to have a counter-espionage pool, reducing my success at spying. Maybe it does this. But it seems I steal a lot of tech, and it doesn't.
6. AI only attacks local starbases. When I go to war, my number one goal is to take a planet. But that might be too tough to do. My number two goal, a close second, is to take over resources, especially durantium. It's the most important resource in the game as it permits most starbase upgrades. I'll send a small fleet out into the black in order to take over resources, knowing that these are finite and that I can starve my enemy. If I have more resources, I can trade them, use them for weapons, use them for special modules on my ships, upgrade my planet resources, hire more mercenaries, etc. But the AI does not share this view. Sure, it will attack a starbase that is near my colonies. But any far-flung mining starbases are ignored by the AI. I don't bother upgrading these starbase defenses because they will never be attacked.
That's it for now. Thanks, Stardock, for your great game and truly rare support and dedication. I'm a big fan from back in the day and a founder's edition player. Next time you release another founder's edition...I'm buying that too.
I have not seen all of these behaviors yet, but I can confirm they are all consistent with a pattern of the AI not doing things it seems smart enough to know to do. I am still working at Normal, and may step up to Gifted. We'll see what happens. But preliminary indications are that these are issues deep in the AI and its web of priorities. That is also a sign that other portions of the AI are getting clean enough to reveal long term trends. To my understanding, that is how AI development works and why it can be a frustrating long term effort.
Good points. Good descriptions.
on #1 its something i have been crying about on the forums since almost day 1. Basically an unmodded game the AI#1 is awful at playing the game esp at higher difficulties. Im sure people who play normal barely care they are fighting an ai with lasers and starter weapons at turn 300 but shrug kinda stinks.
My rants is more pre 2.3 so what im saying may be outdated/fixed as i am taking a break from trying to make crusade work but from the battle viewer being borked and the AI not being capable of playing the game i have spent more hours modding and testing than enjoying the game.
Since basically we have enhancedweapon and weapon. With "weapon" being the starter stuff and enhanced being the researched stuff that takes resources. Since there is always a shipblueprint pattern available that doesn't use resources and since the AI is apparently incapable of actually planning and ends up just building the cheapest ships it can spam which are always ships filled with lasers or basic kinetic or basic missiles. So esp at higher difficulties where the AI has tons of research and tons of manufacturing and is not limited by maintenance and since they need to spam ships they will just end up spamming tiny and small cheap basic weapon ships with sometimes a medium and a large one here and there. Even if they have all other techs researched they just build crap.
Only way i could get the AI to actually play end game was to
#1 remove the enhancedweapon tag and make everything the weapon of its class. Change all the blueprints to just be weapon and no longer require enhancedweapon because if the AI cannot fullfill the required they cannot build it and since resources do not scale with difficulty the AI will always be resource starved but always be building so pretty much all they see im guessing is the basic blueprint as an option.
#2 this is the part that really sucks the only way i can get the AI to play the entire game is to remove resources all together by ship components just not cost resources.
#3 this may not be required but i noticed the AI has an issue with the two begining lasers and even if i did #1 and #2 they would still build crap. So i had to remove the two redundant starter lasers and just push them back into a required tech that cannot be researched.
By basically making the way worse i can now get the AI to build ships that will be best weapon but sadly it is not the way crusade was intended or is really that fun since taking our features of an expansion just to get the game to be playable is zzzz. Also dont forget that the AI will just spam small ships every turn from every starbase at higher difficulties since it builds stuff basically for free.
I have not completely figured out why the AI spams small cheap ships all game/late game. I assume it has to do with the AI wanting to build something that it can afford to rushbuild over actually building something good. I tried to change the priorities in the xml but the AI just wants to rushspam the cheapest ship. So sadly the only want to make the AI build good stuff is mod the game where the cheapest stuff is good stuff, which is usually by making it so the AI cannot see/build the true cheap stuff.
I really hate to agree, but I completely agree. The ai clearly knows when I am weak, Might even declare war, but never follows through with the curb stomping.
good post +1
I actually didn't even know #5 would inform you when an enemy spy stole a tech. I'd always assumed that I just had really pimp counter-espionage.
Agree with all points, I'll add another few:
The AI doesn't seem to understand the value of combining specialisations. My usual research strategy is: Let the AI research one or even two of the specialisations, then go ahead and research the remaining one. Trade the AI for the one it has, preferably without giving it valuable specialisations in return. I never accept a deal with the AI which give me a specialisation I haven't already researched, so I don't block out the option to research it. I've never seen the AI do likewise which cripples it in late game, only one example is the weapon range/power boosts and miniaturisations combinations.
A lot of late-game power comes from ideology by going after all three trees with missionary, preparedness and intimidation centers on most planets. I've rarely seen the AI build any of these, although those three techs are probably the most important ideology choices as they allow you to gain points passively, every turn.
Precursor anomalies can be game changers, and their survey should accordingly be prioritized by the AI, but it doesn't seem it is.
Godlike is too easy? Wow. I just about survive on Normal. Guess I should have expected that one from someone sooner or later. You don`t happen to be constantly reloading whenever something doesn`t go perfectly your way perhaps?
I agree on some of this. I've spent a lot of time coding on all these points.
The biggest problem is they don't let the AI design it's own ships. I only can choose among the crummy ships that it generates. It's very very frustrating.
At some point we will probably write a full on ship designer AI. But that's like a solid 2 months of work.
Even just updating the Blueprints is easiely a few days of work if you want to do it right
(you have to actually take a look at how the AI researches and keep in mind how it gives out ship-roles)
have done this for both the Basegame and Crusade and its a frustrating work -.-
DO IT!
Even make it a paid DLC if you must... which is something I don`t say often, that`s how much I want AI to always be better.
On my last game on godlike, large, 7 civs, the ships they were building were actually okay. The problem is that they never put them to good use.
When I play, my general thought process is:
The ai seems to be doing 1 and 2, but I never see the war party coming to visit me. What I end up doing is quietly stealing everyone's planets out from under them until I've snowballed into victory.
One thing that would easily fix this is if the game in general would recognize this behavior and form a coalition against me. I can divide and conquer, but I probably can't fend off the entire galaxy at once.
I have over 600 hours of game play and I have only seen 1 or 2 of my own designs being used.Why doesn't the AI make better use of the Millions of Player made designs? or even my own designs?
I use to think that it wad because I would get the Hyperion Shrinker and totally maximize the space available, but that isn't so much an issue in Crusade....
Underlined for emphasis. I would like to see this.
As a player, if I see a AI gaining snowball momentum and running away with the power creep, I will do what I can to destabilize their position. Pay for proxy wars, form an alliance, or just outright start a war if I'm in a strong enough position where I can make preemptive strikes and cut down their overall strength; a human player can modify ship weapons systems and defenses to exploit the AI fairly easily. This means a human is much better at punching above their weight against an AI. The last game I played, I spawned in a favorable position. I quickly gobbled up some prime resources and real estate, and quickly wiped a weak Drengin and Kryn player off the map. This was all in the early/earl-mid game. I consolidated and rode the rest of the game out completely unchallenged because my position was so strong no one would touch me (I kept getting vid mail of the suckers praising my vast p... prowess). The galaxy otherwise steadily devolved into warfare around me as the AI fought wars between themselves, with me receiving the occasional request to fight proxy for another's interest.
Still, I would think if the AI wanted to win, they had ample incentive to form a coalition, shut me out of UP leadership, etc. Should I instead believe that my immense military stick and economic might was seen as a good thing by competing galactic powers?
--Duplicate--
Hi Frogboy. Thanks for chiming in. Longtime reader of your posts, including your recent one about the AI being better, in your opinion, in Crusade. You know more about AI than I do. But because I play your games and try hard to win, I experience the AI and pay very close attention to what it does.
I want to push back on that comment about the Crusade AI a little bit. In GalCiv2, I did not win on the highest setting (can't remember what it is called; suicidal?) consistently. In fact, I had to play the Drath only, leverage their economic model (based on galactic war) to eventually pull off the upset. Most of the time I didn't win, as some civ had snowballed past the point where I could make a comeback. Most of the time, I was attacked early to mid before I could defend enough. These games were the best for me, actually, because for me to have fun I must have a significant challenge. No challenge, no fun. I might play the game on lower settings initially to learn it, but once I figure out the rules of engagement I crank it up. That's the fun for me: to almost lose, to be on the edge of loss, and then to pull off the win. And in GalCiv2, the highest setting was truly difficult.
I'm wondering if the AI has too much to manage in GalCiv3? The game is far more complex than GalCiv2. Here's what I mean:
1. In GalCiv2 hardest setting, I didn't have to focus a planet on research, money, industry, or influence. Indeed, I often had all four on every planet, spreading out my empire broadly. I could have a planet specialize, but I didn't have to, and I could still win. In GalCiv3, this does not work on godlike. If I want to win on godlike, I must specialize some (most) of my planets. One of them specializes in food (usually a poor planet), one in money, one or two in industry, the rest focus research. I ignore influence unless I try to convert a planet. So I must specialize to win. But when I take over AI planets, most of the time I see a cacophony of choices. It is rare to see a specialized planet, although it happens. Most planets are generalists. I'm wondering if the AI has too many choices for planet tiles.
2. In GalCiv2 hardest setting, the AI would research the higher weapons. This meant I had to research the higher defenses, and focus on whichever defense countered the AI's preferred weapon. In GalCiv3, I have yet to research a defense beyond tier 3. It is unnecessary since the AI is not researching the higher weapons. Perhaps they are not researching the higher weapons because the AI does not understand about using resources (ie, using elerium to make higher beam weapons) as Goxwerd noted above. This complexity is fun for me as a player, btw, but perhaps is too complex for the AI.
3. In GalCiv2 hardest setting, the AI would very quickly target the weakest civs in the game. That was always me, but maybe a few others that had a bad start and only had two or three planets. The rule was, attack the weak. In GalCiv3 godlike, I'm still the weakest in the early game. I get threats from a lot of AI civs, but they don't attack usually. I ignore their demands and threats. Eventually a few will attack. But in the game I am currently playing, the Krynn (most powerful) own half the universe and are at war with all seven other civs! So are the Altarians (second most powerful). And since these two are stretched so thin, when they declared war on me I knew I didn't have anything to worry about. They could not bring the full might of their navies on me, since the AI would target too many things. The rule about declaring war seems to be, attack the civs that you don't like in addition to attacking the weak. That may be too much.
4. In GalCiv2 hardest setting, the AI invested in movement technology, which was smart. In a 4x turn based game, movement is always the most important trait for winning. It's more important than attack or defense, for too many reasons to bother listing here. GalCiv2 AI would regularly have ships that could move 15-20 in a turn (such a pain). In GalCiv3, I'm the only one who builds the navigation wonders. I used to think it was because the AI did not have techopod hives, but that's not it. Even with the resources, the AI does not know how important those wonders are. It ignores them. So in the above game, the Krynn's massive ships move 2 or 3 per turn. My ships currently move 28 per turn (without putting an engine on the ship). I can't take the Krynn out just yet, but in the 30 turns it takes the Krynn to reach my best worlds, I will research the weapons I need and build the ships. So even if they did come, which they won't because of #3 above, I would be ready. I'm wondering if mobility has fallen off the AI radar as an important tech.
5. And then there is espionage for it to consider, and leaders, and anomalies guarded by precursors, and tactical resources, and precursor relics, and ideologies, and ship designs, and a larger tech tree, and...and...and. Ironically, the complexity of Crusade brings a great deal of enjoyment to me as a player, but may be overwhelming the AI. I don't know, but I suspect.
Lastly, I lost my first game on godlike yesterday. The Drengin. It's always the Drengin.
Love your art; keep it coming!
Oh my, no, Seafireliv. That is verboten. No reloading allowed, as a personal rule. If I mess up, that's on me; learn from it.
As a player, if I see a AI gaining snowball momentum and running away with the power creep, I will do what I can to destabilize their position. Pay for proxy wars, form an alliance, or just outright start a war ...
Good point.
@Brad: if human players can turn the galaxy into a froth of diplomatic hatred, AI players should be able to do this as well.
Example: In my last game, the strongest empire declared war on me. I wasn't ready for that. So i went to everyone else and paid them top dollar to declare war on the civ that declared war on me. Now they are all occupied destroying each other and i continue my handiwork.
This inability of the game world to recognize that I, the player, am about to work my malign will upon it is a defect of almost every strategy game I play. It's particularly curious if you are playing "evil," and the other races all just fall over to you. I mean, if you win through cultural domination, that's one thing, but you'd think that all the collective civilizations would object to their eventual slavehood and second course as dinner.
In fact, putting it this way, that should be one of the major disadvantages of playing evil: the other civs are likely to gang up against you.
Anyway, just keep in mind, as a player, you do need a foil against this sort of thing. I.e., there should be steps you can take to prevent the gang up, or to diffuse it. "Sabotage" the conferences, perhaps? Uses spies, and sew confusion amongst your enemies? I dunno, I'm just saying that you can't include an everybody wants to kill you game mechanism if there isn't also some game mechanism included that gives you a chance...
There are many great features available to you once you register, including:
Sign in or Create Account
