AI: I'm so excited!

New Stuff!

It's been decided that we are going to do a 1.X.2 release. Don't know when. But there's little tweaks and changes we want to make. The team fixed some glitches that have been reported on the forums.

But for me, it gives me a chance to start having some fun with the AI.

Stupid AI

At lower levels, I want the AI to play stupidly. I don't want them to just be super smart but economically starved. There are a bunch of APIs in there for the AI that will cause it to make "newbie" mistakes.  But I want to do more of them.

Smart AI

Yay! They're making me an UpgradeShip API for the AI players. Now I can really bring hell down on people. Whohaha.   I also have a few other ideas too.

I've got a bunch of other stuff in mind as well.  Will have to make sure it can be done without slowing down the game.  One of the biggest things I want to work on is having the AI be more effective at controlling their territory militarily.

Stay tuned.

49,781 views 38 replies
Reply #1 Top
It seems you are still very much enjoying the development cycle on your baby, and I love to see that. It's incredibly frustrating to get a game that you enjoy, but which is seriously flawed and realize it might never really be fixed, much less improved from it's release point.

That's one of the things I dislike a lot about console games. While I know I can get decent games for a console system, normally with current generation top of the line graphics (as an example, Xbox 360 stuff, or the pending PS3 stuff), once a console game is burned to disc and shipped to customers, most of the development work is ended, and except for perhaps roster updates for a sports title, or some other very minor fixes that can be downloaded from a service like Xbox Live, the game code will be forever frozen in time, with no future enhancements, at least until the next year for the series (if it's a sports title, or other title that might be part of a series).

It's a huge advantage for PC games. If the developers are active (like Brad and StarDock obviously have been), then the game makes constant improvements, and everyone wins, or at least everyone benefits (as long as they keep downloading the updates).

Even with that advantage though, there are plenty of developers that put out PC games and really never improve them. They just don't see the economic sense in it, and just aren't that enthusiastic about their products, or, they get shifted over to another project very quickly and never have time to come back and address any misgivings they may have had over lack of features, bad functions, etc.
Reply #2 Top
So does this mean the AI will have less economic starvation at lower levels? I'd be all for this. I can win easily at normal, but have a much more difficult time at challenging. I can only win on intelligent in a 1v1 or if I can actually convince the computer to ally with me, which I usually can't. I've lost every time I've tried to play a 10 player game at challenging or intelligent. So I guess I wouldn't mind the AI at normal being more on par economically.

The AI progression for me goes something like this: too easy, meh, ok, hard, really hard. It's missing that "good" step in there.
Reply #3 Top
I believe that this is the first time where I've seen 'make the AI dumber and suck more' as a feature to develop.

Rock on!
Reply #4 Top
Im all for tweakin the "intelligent" area of the AI. It still seems too easy to dominate the AI at that level. It would be nice to see the AI build different "fleets for different purposes." My attack fleet usually consists of slow moving but high attack ships, while my def fleet is more manuverable but less power. I usually protect the space in my territories with 2-3 fleets near my borders/starbases. while i send 2-3 fleets out to attack the enemie's planets. I would also like to see the AI choose its targets more carefully (I usually see the AI chasing my flagship movement 4, and the AI ships only have a movement of 2 and it chases it halfway across 2 sectors) which leads me to exploit it. I end up beating a superior AI force by using my flagship as a decoy then i make it double back adn destroy the ships that were chasing. I would also like to see more of that kind of tactic comming from the AI. (it parks a bunch of ships near my border while it sneaks around the flank to take an unguarded/ low defended planet. anywayz exciting stuff
Reply #5 Top
They're making me an UpgradeShip API for the AI players. Now I can really bring hell down on people. Whohaha.


Hey Brad, just because a few ill informed muppets give you a hard time on the forums because they didn't upgrade their video drivers, there's no need to get nasty...

Great news though - have fun playing with it!
Reply #6 Top
Regarding the "make the AI dumber", I say that the AI (At low lvls) should try great tactics, but screw them up. ex: a flanking menuver... that underestimates your rear. And more taunting at low lvls.
Reply #7 Top
I'm glad to hear that there is a 1.X.2 on the way! I love Galciv2, but I absolutely suck at it. The Yor are stomping me on Normal level right now. I can't wait to see what the update includes. Go team! *Slaps Stardock on the rear*
Reply #8 Top
"I've got a bunch of other stuff in mind as well. Will have to make sure it can be done without slowing down the game. "

@Draginol

I'm sure I'm probably in a minority of one here, but I'd love an option that let the AI have more time to have its turn if it meant it would act smarter.
Reply #9 Top
Stupid mistakes.. like building a tech capital on a manufacturing planet ?


Reply #10 Top
I'm so excited, and I just can't hide it
I'm about to lose control and I think I like it
I'm so excited, and I just can't hide it
And I know, I know, I know, I know
I know I want you

(Next line for the song in this context is "to die horribly under the missiles of my AI's attack fleet" followed by maniacal laughter.)
Reply #12 Top
Don't know if you want AI suggestions, but here it goes:

Suggestion #1: The AI needs to make Cargo-based Sensor Ships (not just scouts) and use them properly.

All too often, I'm able to send a fast ship ahead to bait the enemy into coming into range of my fleet. I can only assume that, if the AI were to 'see' the fleet, it wouldn't blantantly put the blinders on and doggedly pursue my 'bait' to its death. (Just on this point, the AI does seem to pay more than the usually attention to sensor ships. This is not a bad thing, but it doesn't make sense for them to ignore more dangerous targets, or to 'over' pursue my obviously faster ships).

On that note,
Suggestion #2: The AI MUST be taught how *crushingly* important it is to be the Attacker, verses being the Defender.

From what I've observed, when ships of roughly equal power meet, the attacker is about twice as likely to be victorious. This increases even more dramatically when fleets of equal numbers meet (in my experience, its like 3 to 4 times).

Suggestion #3: The AI doesn't seem to use fleets very effectively.

One example of this when my fleet was approaching a Drengin home world. The AI had *6* ships in orbit and no Fleet organizer. As it turns out, I (stupidly) had to spend a whole turn in range of his ships before I got to the planet. I gritted my teeth, expecting the ships to come roaring out into space.....but they just sat there, and I spent the next 2 turns taking them out one by one. (This may be due to the sensor issue, as well. I couldn't tell).


The one cravat in all this is that I haven't played on any levels higher than 'Challenging'. If you tell me that the higher levels will take care of these suggestions, then that's great.

Thanks again for this wonderful game.
Reply #13 Top
Binnister's Suggestion 2 and 3 seem to be very import factors for increasing the AI diffculty. Especially #2. Attacking is so important that I will purposely stop my ships out of range, wait for the AI to move into range, and then attack.

Basically I think it would be better for the AI if it also stopped it fleet's and waited for you to move. Then its a stalemate, if you move into range, you lose. If they move into range, they lose. It is is now, they always move (and lose unless they are somewhat more powerful) because they move forward headless of their impending doom.
Reply #14 Top
This link contains very good information about the mistakes the AI still makes :


http://apolyton.net/forums/showthread.php?s=&threadid=149664

Im copying/pasting what Gufnork said in this excellent thread (i hope he wont mind) : I repeat, this is from Gufnork, a very good TBS player from the Apolyton community :

1. Transports should be protected better. They mustn't always be protected, but if several of it's transports have been gunned down they should start taking precautions. I've also noticed that even though they have fleets one or two parsecs away they won't use them to cover their transports.

2. They need to utilize speed more. For instance, unprotected transports. Higher speed means fewer turns to intercept it. It also means that it can attack from further away, making it harder to spot it before it's too late. Why use the standard transports when their technology would allow them to add several engines for only a little extra cost? They should also have atleast some fast fleets to catch any of the opponents fast ships that strike at their freighters, defenseless starbases and transports. If atleats a few of the AI's could do this as well it would be nice (Korx seems like a good choice, for instance).

3. They seem to ignore planet square bonuses. I've conquered planets where they've build farms on squares that provide bonus to manufacturing and factories that provides bonus to food production. They seem to ignore these entirely.

4. If they see an unprotected transport approaching their planets, they should send out ships that are defending the planets to destroy it if they have the chance. Sometimes the AI could simply pop out, destroy my transport then go right back to orbit in one turn. Yet it does not.

5. They need to send enough troops to conquer a planet. They know how many people that lives on it. They know what range the advantages will be in. It can easily calculate the odds of success and should base the amount of soldiers sent on that. Different races could have different criteria. Altarians might not send any attack that isn't 100% guaranteed to succeed (don't waste people's life for nothing). Drengin might send troops as long as they'll kill more than dies. Having four civilizations weaken my planet so the fifth can take it with minimal losses is just dumb.

6. Minor races tends to spam influence starbases in areas where they have no influence. I don't know the purpose of this at all.

7. The AI should really care about the planet quality of the planets they colonize. Right now they colonize the closest planet regardless of quality. This lets me get the nicest planets before the AI since their first colony ship often goes to a planet of low quality they encounter on the way instead. It would also be good if they tried to colonize the planets furthest away first and work their way in, but that might be harder to do. Atleast the more peaceful civs should do this, warlike civs would probably prefer a more compact empire.

8. They need to research different techs. Right now they all ignore the same techs, meaning that the player can just research one of those and trade them with every single other civ in the game, including minor races. This lets the player keep up with the others with minimal effort.

9. They almost never build farms. This means it's very easy to conquer their planets and their economy suffers.

10. They are slow to eliminate minor races. Since the minor races don't colonize, they can produce other ships instead. Usually these are constructors, meaning they often control a significant amount of resources. They are very bad at protecting these however, meaning that the first civ to attack them will get these. This is generally me, so I love it every time a minor race nabs a resource. They're holding it until I'm ready to go get it.
Reply #15 Top
I second comments no. 2 and 3 of Binnister, the AI is building nice ships and forms good fleets with it (although a bit slow sometimes), but it is rarely taking the initiative. I'm almost never on the defending side in a battle because usually the AI ships are too slow for that and even when they have the opportunity to strike first, often enough they don't use it. From time to time the AI even seems to forget to move some ships, even at war (!).

Another weakness of the AI lies in the defense of its home sectors: The AI should be VERY suspicious about sneak attacks (on the higher levels, of course), and should keep more ships in orbit. (By the way, I think the AI should not park unarmed ships in orbit, they're useless for defense and only take up space for more powerful ships.) The way it is now, I saw (and conquered) planets with 1-3 ships in orbit while dozens of ships were nearby - forming fleets, I think.

(Don't get me wrong, I'm enjoying GalCiv a lot, I wouldn't bother to comment here otherwise. I hope I can give you some valuable hints for improving the AI - it still has a lot of potential.)
Reply #16 Top
It is a testament to how good the AI of this game is that one of the features they're adding is a stupider AI for easier difficulties. I have never seen an AI that is this smart beforehand, in fact, most people use AI and Stupid synonymously when referring to difficulties in games.

Good show, Stardock, good show.

However, I must third the comment on Fleets and Attacking. I too rarely see the AI effectively defend itself. The AI should understand the concept of escorts for vulnerable ships, sallying forth from besieged planets to fight transports and lone ships and returning to orbit, and the idea of relieving besieged planets with other fleets to take pressure off (something I rarely see).
Reply #17 Top
First of all let me say this is a great game. A game where the AI both doesn’t cheat and plays to win instead of only trying to defeat you is a groundbreaking endeavor that should be commended and encouraged. Kudos.

If you are tweaking the AI here is what I have noticed. (apologies if they have already been mentioned ad infinitum)

I have seen the AI place multiple starbases with no upgrades around the map. I am not sure that a bare bones starbase even provides any kind of bonus. One starbase with a few key upgrades would obviously be better if this is the case. Mining starbases on the other hand are usually fully decked out with defensive upgrades.

The AI doesn’t seem to pay attention to population at all. I have viewed game ending graphs where all the AI’s plateau early and stay constant. There are reasons perhaps to do this but it seems it would limit not only taxes but also the number of troops available to attack/defend.

Here is something that was also in GalCiv1. Once an AI has ships guarding a planet it doesn’t seem to upgrade or add to them until they are wiped out. This can lead to a race’s homeworld being guarded by scouts and colony ships, etc. It is odd since once you clear a planet if you don’t invade immediately the AI will then usually buy a decent defensive ship. This makes you just wait until you can take over a planet before attacking since there is no point to wearing down a planet’s defenses. It is not as extreme as it was in GalCiv1 but it does seem to appear from time to time. I think it is has to do with how the AI evaluates ship strength on a planet. It may see 4 scouts and 3 colony ships as enough protection just by looking at the numbers even though none of them are armed. That said I am pretty bad about putting little more than token defense around my planets so maybe this is just mimicking human behavior.

I have only played on small maps (at all AI’s at intelligent level) so this may be the cause of these behaviors. Also this is the X9 patch and not the X10 patch (haven’t been able to play since last weekend) so the latest may have cleared things up.

Keep up the good work.
Reply #18 Top
I have to second the 'not escorting transports' thing. They don't even hold them back until the area is clear: I've seen them drive transports waaay across my empire to get the one undefended planet in the middle. They must be hoping I don't pop a defender out and blow them up on the way past! Transports - and population - are expensive, they should be protected. The AI (on normal, or one up from normal) often bleeds it's planets white for no gain.
Reply #19 Top
I'm glad to hear that there is a 1.X.2 on the way! I love Galciv2, but I absolutely suck at it. The Yor are stomping me on Normal level right now. I can't wait to see what the update includes. Go team! *Slaps Stardock on the rear*


I started a game on beginner when I first loaded up the game and got the snot smack out of me. I had to go to easy to get the hang of things then slowly progress up. I just got done with a technology victory on beginner. It was a LOOOONG game. Next i'm going to normal.
Reply #20 Top
Looks great. Any thoughts about including new AI regarding the tech trading exploit?
Reply #21 Top
And to think, all the time spent on making the AI a tiny bit better could have been spent fixing the economy.
Reply #22 Top
Pffff. The economy works fine. And Brad isn't the only guy who works on the game, he's just the only guy who codes the AI.

Better AI is always welcome. The one thing that distinguishes GC from all other 4x games is it's AI.

People need to get a grip.
Reply #23 Top
Agreed - I may be new here, but it's nice to see someone just focusing on the AI side of things. Thus far, I've been quite impressed.
Reply #24 Top
I might be wrong about this, but I think that some of the influence starbase spamming might be for the purpose of increasing range.
Reply #25 Top
LoL, Scumbag mentioned the GalCiv2 economy like it is broken. I don't have any problems with it. Is there an issue, or was he just tossing out garbage because he doesn't have a handle on the game yet?