To be honest, the AI is in a disastrous state at the moment. Even at the highest available difficulties, the AI is inefficient, suicidal and lacks a decent strategy to actually win the game.
If you browse through the posts, you can see more and more beta players complaining about that and in my opinion this is the worst point of the game currently. One can forgive if there is some small graphics bug, or if the UI is cumbersome, maybe even seldom game crashes. But one completely looses interest in the game quickly, if the AI is not able to challenge you and make such stupid things like
Sorry for being cynical, but you devs/Stardock invested a lot of time and money in the game development and now (are forced to) release too soon. I fear the game test scores will be much worse due to that fact and your ROI too. In the old Ascendency-times it was kind of ok that the AI was dumb, but nowadays in view of e.g. self-learning neuronal networks getting common (there are even open distributions of the available for C++/Python etc.), this has to change.
I'm looking forward to your reply and beg you to focus on AI-improvement, please!
Quoting Franco fx, reply 2973% of players are using the beginner setting.
Like many, this surprised me. I play on Challenging because that gives me the balance between "hard enough but not too hard." Always considered myself a casual 4x player.
Many have asked, "Show me a better AI." Civilization 5: Brave New World. GC 3 isn't there yet (though I have hope it will be better than Civ 5 eventually).
I've already mentioned the AI problem with pirates. (AI in Civ 5 deals well with even raging hordes of barbarians.)
Here's an AI behavior I don't understand: I far outclass my AI opponents in my current game. They have offered me several gifts. Why? I haven't threatened them or promised them protection or some other benefit. AI gifts should be based on more than just simple fear. I certainly don't do this as a human player. I always have a reason for gifts. If the AI is threatening, a gift might keep them from attacking until I have defenses in place. Or, if an AI is at war with a more powerful AI that is hostile to me, I might give them tech and ships so they can survive. Or, I supply the Drengin with ships because they will keep lots of other AIs busy fighting instead of researching and building up their worlds while I quietly build my empire and position it to win (I used this strategy A LOT in GC 2). The gifts kept the Drengin attitude toward me friendly and didn't undermine my relations with other less warlike AIs. (I always included a "hole" in ship defenses that was vulnerable to my ships in case they turned on me.)
The AI should have a purpose, a strategy, when they give a gift. "I give you this gift so you won't attack me" would be something an AI would do if I threatened them or have fleets near their border. "I give you this gift in the hope you will open your borders to me" is an example of a preemptive "this for that." You know what they want. Maybe they've offered open borders as a trade and you refused. Or maybe, "We give you this gift and will send more if you ally with us."
my core strategy was much the same. Constantly induce everyone else to be at war, while I quietly built up a big econ and tech lead.
How do you encourage an AI civ to go to war with another? I've tried bribing them but it seldom works.
On another note The release note they are putting together for 6.2 shows a lot of AI improvements and fixes that should make a big difference.
Yeah can't wait for the patch.
Sounds really silly, but I actually want to lose a game!
Like many, this surprised me. I play on Challenging because that gives me the balance between "hard enough but not too hard." Always considered myself a casual 4x player.Many have asked, "Show me a better AI." Civilization 5: Age of Kings. GC 3 isn't there yet (though I have hope it will be better than Civ 5 eventually).I've already mentioned the AI problem with pirates. (AI in Civ 5 deals well with even raging hordes of barbarians.)Here's an AI behavior I don't understand: I far outclass my AI opponents in my current game. They have offered me several gifts. Why? I haven't threatened them or promised them protection or some other benefit. AI gifts should be based on more than just simple fear. I certainly don't do this as a human player. I always have a reason for gifts. If the AI is threatening, a gift might keep them from attacking until I have defenses in place. Or, if an AI is at war with a more powerful AI that is hostile to me, I might give them tech and ships so they can survive. Or, I supply the Drengin with ships because they will keep lots of other AIs busy fighting instead of researching and building up their worlds while I quietly build my empire and position it to win (I used this strategy A LOT in GC 2). The gifts kept the Drengin attitude toward me friendly and didn't undermine my relations with other less warlike AIs. (I always included a "hole" in ship defenses that was vulnerable to my ships in case they turned on me.)The AI should have a purpose, a strategy, when they give a gift. "I give you this gift so you won't attack me" would be something an AI would do if I threatened them or have fleets near their border. "I give you this gift in the hope you will open your borders to me" is an example of a preemptive "this for that." You know what they want. Maybe they've offered open borders as a trade and you refused. Or maybe, "We give you this gift and will send more if you ally with us."
Halimac, You mention Civ 5 Age of Kings. Ok... I have all the expansions and this must be a new one. So is Vanilla Civ 5 as good or better than Vanilla GCIII? The point being is that Civ 5 has had what 4 or 5 real expansions to get where it is. Personally I think the AI is rather good. I love Civ 5 but it has taken a Loooooong road to get there. I am pretty sure the AI here in GC III will far outshine Civ 5 when it has 3 or 4 major expansions under its belt.
/game on...
I actually can't comment on Civ5 much, I got it for free in some deal on Steam, tried to play it a couple of times but it was incredibly unstable on my system and crashed about every 10 minutes, also I wasn't having much fun with it in the bit of play time I could get between crashes so gave up.
I've never managed to get into Civ5 - find it too complex and the UI non-intuitive and rather ugly.
GalCiv3, on the other hand, is pretty easy to pick up (but hard to master)
Quoting Larsenex, reply 56Halimac, You mention Civ 5 Age of Kings. Ok... I have all the expansions and this must be a new one. So is Vanilla Civ 5 as good or better than Vanilla GCIII? The point being is that Civ 5 has had what 4 or 5 real expansions to get where it is. Personally I think the AI is rather good. I love Civ 5 but it has taken a Loooooong road to get there. I am pretty sure the AI here in GC III will far outshine Civ 5 when it has 3 or 4 major expansions under its belt.
Sorry. Brave New World Expansion, not Age of Kings (It was actually Gods and Kings; This stuff tends to merge in my head as I get older). And yes, it has been an evolution in the Civilization series but people were asking to show a better AI, so I mentioned what is still the gold standard for 4x games. That said, I have high hopes for Brad and his team.
I think the issue with developing AI on a complex game like this is an example of chaos theory, you tweak something to make the game AI play better in one area and it seems to work but throws something else way of kilter.
I think by the time GC3 reaches it's first expansion it's AI will also be much better. There are already lots of fixes and tweaks coming in 6.2 and Brad intends to keep working on it long after release. GC2 got AI improvements for years after release.
You also need to take CPU power into account, there are just so much CPU power you can give to the AI before people start complaining about turn times. If we all could accept five to ten minutes between turns then you could make AI pretty smart and more dynamic. We just have to wait for quantum computers and then you will get AI with nasty abilities that regularly beat even the brightest of human players...
In the future when quantum computer is a reality you will see games where you as the player act more like an administrator and each ship captain/admiral and governing body on planets have their own mind and psychological profiles. They perform whatever tasks you ask of them but you will not control everything to minute detail. You will even be able to have REAL fog of war with communication problem and lag in data-flow to you as the head of the government of your faction.
Now we talk real AI where you hardy know if the opponent is a human or AI automaton...
I won't get into the AI of the games mention (not just because the people who wrote them mostly work with Stardock now). I will say that the AI in GalCiv III on release day is certainly as good if not better than most 4X titles on their release days.
I mean, if you guys want it really tough, I can make a super tough AI in a few hours that will beat you. It'll cheat like crazy but you'll get a challenge. That is, after all, what most game AIs do (Civilization included).
The challenge here is to get an AI that plays by the same rules as the player and even on the harder difficulties is still playing by the same rules as much as possible.
Judging a pre-release AI, however, is not especially useful. What is useful is identifying specific things. I think people would be surprised at how quickly (think engineering minutes) it takes to fix dumb AI stuff.
It does this. It just apparently is not picking it well. I'll ask them to look more into that (I don't write that part of the AI)
It is getting better at this. The problem is that there is no such thing as "should" with AI. Should is a subjective opinion. Over the next year or so, this is an area the AI will get a lot better at since we'll all be better at the game.
This should hopefully be reduced in the next udpate. The idea was to try to soften them up but the combat ratings are not doing the softening.
Only one of the last two issues can be true. That's not suggesting that you're not seeing what you're seeing. It's a matter of the AI deciding what matters in different cases.
Over time, it'll keep getting better and better.
I've said this before Frogboy - if you fix the economy issues it will have a whole bunch of knock on effects that will make all the other problems seem less serious. As such, I'd drop what you are doing with the other stuff and fix this pronto!
Ie, lets imagine the AI does something really stupid with a colony ship - running it right up next to a pirate ship and getting it killed.
If it's economy is weak, it could take 5-10 turns to recover the population and produce a new colony ship to have another go.
If it's economy is strong, it only loses 1-2 turns.
Same deal with dumb military decisions. An AI with a ship decides to attack a star base.
if it's economy is poor, the AI tech level will suck and it wont manufacture big ships quickly - if it has them at all. So the chip it attacks with is likely to be tiny/small with small attack/Def numbers. The ship will die without doing significant damage. The decision will appear to be stupid.
on the other hand, if the AI has a strong economy with advanced tech, that same ship (for a given number of turns) might be large with reasonably high attack/Def. The star base will get creamed and the decision will appear to be smart.
Fixing the AI economy will go a long way!
And I love the fact you are reluctant to resort to cheating - it's what made galciv2 so good!
Only one of the last two issues can be true. That's not suggesting that you're not seeing what you're seeing. It's a matter of the AI deciding what matters in different cases.Over time, it'll keep getting better and better.
First, completely agree that the AI will get better. I saw what you guys did with Gal Civ 2!
To your point, this is what I think is happening as I have watched the AI war with me:
If we think about "Warfare for Dummies in Gal Civ 3", you have 3 objectives
1) Take enemies' planets.
2) Kill Enemy starbases
3) Kill enemy fleets.
What it looks like is that the AI is heavily favoring strategy 3 right now. It will chase my fleets around, even bypassing starbases that it can take. Also, I rarely see the AI building transports and going after planets. Ultimately the AI's priority should be noted as above as a 1st level. There are obvious exceptions to that (don't kill a starbase that will cripple your fleet when their is a strong enemy fleet nearby), but as a 1st level pass I think it would make the AI more competitive.
Now when the AI does finally decide it wants to kill a starbase, I think it should always go for a kill shot. The soften strategy is not as effective in general, and often the fleet gets obliterated. My thinking is:
1) If I'm near an enemy starbase, engage it.
2) If I can kill it, kill it.
3) I can't do 2, switch priorities to other target.
4) If no good targets are in "sight", acquire greater fleet strength.
Once again, this is very crude, and there are a number of exceptions that will be worked out over time. But even this simple logic puts pressure on the player, if they don't react to the AI before its fleet builds up, starbases start to fall.
Lastly, for the initial release of GalCiv 3, I do think the AI should cheat at the highest levels, in a graduated way. I saw a similar technique used in a Civ 5 mod that I thought was useful, the AI was giving more bonuses gradually over the course of the game. This ensured the AI was competitive without overwhelming the player with an early death rush from the AI.
Now long term we all want an AI that doesn't cheat. But we aren't there yet, so for the 5% of us that want that ultimate challenge, I say put in the cheats for now and then scale them back as the AI gets smarter.
How actually do you program the ai? I mean, is it only kind of simple if then else statements or what else? I also ask because you say it can learn and second because you say you dont need suggestions and we would be surprised how easy you could fix issues...
I keep seeing it. In fact, he's goddam annoying at attacking my undefended planet, bypassing my fleets
Brad, I don't think you should take whatever criticism you get about the AI as a knock on what you're doing. I think most of us recognize the huge challenge you're taking on. I think we'd also acknowledge the strides the AI has taken to becoming a more viable opponent.
And yes, you could just whip up a cheaty AI and have it challenge players with sheer numbers. That would be easier (though probably not satisfying for most of us here, which you know).
You're the one who set yourself this goal of making a challenging AI that plays the same game we do. So we're all on the same team. You want to do it, we want you to succeed.
This! My recollection of GC2 might be a little hazy, but my impression at the time was that the AI was good because it did 2 things:
I'm sure there was a lot more to it, but from my perspective those were the elements that made the AI stand out from other strategy titles.
now the AI in GC3 is way more complex and involved and a lot smarter in terms of the information it has to work with, the clock cycles it has, and the strategies it can employ. But if it doesn't do #1 it's only going to magnify its other mistakes.
Brad, you posted a "Me vs. AI" story earlier and, unless I am reading into it too much, the narrative you describe suggests the AI is really getting a lot better. Yes, it failed to appropriately respond to your initial thrust into its territory, but when you overextended yourself it came back with better ships than you could field and started eating your lunch.
I'm actually a bit confused as to how it pulled that off, since at that point you had taken all but the Drengin homeworld and sister planet. I think you might need to talk to your team about upgrading Frogboy's AI
I got the say the AI is much better now. Still not to my liking, but better and more agressive.
There's goi to be another opt in tomorrow that's even better.
there is one problem I would like to see fixed.
Let's say you build teams at the start of the game. Ally with 1 AI races, put 1 other AI races as allies, leave the other independant. As you explore the world, all your allies have shared vision.
Let's say I play Terran. Allied with Altarians. Drengi is allied with Yor, all others independant.
Drengi is close to Altarians. Drengi is very war like and he hates me. As the Altarians discover the Drengi, so do I. First contact is so-so, a few turns later the Drengi declare war on me.
Problem is, we're playing on huge maps. So the Drengi, he's in his little corner of space, I'm at the other end of the map. He's trying really hard to get at me, but he can't reach me, and he probably won't be able to reach me until very late in the game. And to reach me he would have to either cross a vast expanse of empty space or jump over 2-3 AI nations. While trying to war with me, the Altarians will declare war on him, and the poor Drengi is fighting a war on two fronts, trying to outmatch me while his real ennemy is the Altarians very close by. What eventually happens is that he gets slaughtered by the Altarians (or any other nearby race). Meanwhile, the Yor, right next to me, will declare war on the Altarians they can not reach without going over the Krynn first (and conquering their planets to extend their range). They build their economy for war, devote resources to building ships that won't really serve and when they advance in technology, they are unable to upgrade their ships because they are either strapped for cash or bankrupt or now they are at war with another neighbouring race wich has different weapons tech for wich they were prepared.
It is a common behaviour, as in, nearly every 4x game.
Before the AI declares war on another player, there should be some kind of check as to wether or not it can actually reach that race. Simply being in contact or even being able to send a scout or survey ship over there is insufficient. Starbases are a good way for the AI do extend its range (and that bastard will never frown upon it! ), but they require resources to upgrade (as in, constructors instead of military ships) because otherwise, any mid-range fleet will attack the starbase, destroy it, and cripple the AI's ability to wage war on the player.
The problems here with starbase is that AI rarely upgrades it's starbase to their full military potential. Fair enough, I don't do it until later in the game either. If the AI wants to attack me and needs starbases to reach me, it should prioritize upgrading and/or protecting these starbases that are necessary for its war.Also, maybe the concept of white peace should be implemented. If you don't attack each other for X turns, than peace is automatically granted after that time frame. It would let the AI concentrate on its closests neighbours that it can conquer without a chain of starbases.
I believe this would make the AI much harder to beat.
I know Brad has mentioned that such a check exists. Its probably either too weak or its conflicting with another priority which is why it doesn't look like there is.
yeah, he said that if we thought of something logical the AI should be doing, it's already there. I must have missed that particular thing.
In one of the patch notes, it was said the AI was more focused on its traditional ennemy. I think this is what is going on. The Drengin probably hates the pragmatic or benevolent races more than others, so they declare war on them as soon as they see them and can attack (military research).
There are many great features available to you once you register, including:
Sign in or Create Account
