The Gal Civ formula (and why it needs to be addressed)

from Stardock Forums
Preface: I play on intelligent and I've played about 10 - 15 games at this level on huge maps.

I've read a number of threads and talked to some other people who play the game and it appears that a lot of people are coming to the same conclusion: There is a tightly defined formula for winning Gal Civ.

This formula is esentially,
1) expand like crazy
2) research straight to dreadnaught
3) build up military to at least par with the next strongest civ
4) sit back and wait for tech or culture victory

Now of course there's trading, starbases, and economy management in there, but that's basically how everyone seems to be playing. To me, this seems one dimensional. The problem is that even though the game has multiple ways to win (tech, culture, etc), having the strongest military is a pre-requisite to all of the victory conditions.

Additionally, if I can get to dreadnaught without suffering too many losses, my winning is a foregone conclusion. This obviously takes much of the excitement out of the game for me.

There are a few things that I think should be done to fix this:

The computer should request less money, perhaps for longer periods of time, when it's trying to be a bully. Asking me for 1100 bc when I only have 100 is stupid. Doing this would keep all civs (not just humans) out of war more often since they are actually able to pay for peace. This way, if I wanted to play as a peaceful economic power, I could just be paying 100 - 200 bc per turn indefinitely to the dregnin to keep them off my back (obviously they would eventually need to detect that I'm becoming a threat and attack.. but just not as quickly as they currently do).

My second item I believe is being addressed, but it's so bad right now I think it deserves more attention. The computer (all of them) is horrible at starbase management. Computer starbases built off of resources are invariably never upgraded and those built on resources often are only 1/3 to 1/2 upgraded. If the computer was better at managing starbases, perhaps the period after acquiring dreadnaught tech wouldn't be quite so easy (the premise is that while I'm building up SBs all over the galaxy and extending my power, the computer is not).

This last one seems trivial, but it does make a big difference. I ALWAYS have at least one survey ship cruising the universe picking up anomalies. I have never seen a computer do this past the first year or two. Presumably they commit their survey ships to battle at their first opportunity. After a few hundred turns, all of those +1% bonuses add up.
3,299 views 97 replies
Reply #1 from Stardock Forums Top
*BUMP*

I don't agree on the survey ship point, though. With my games, the Torians and I always have a "race for anomalies" going...

~SDC~
Reply #2 from Stardock Forums Top
I think the problem is that the AI is just too one-dimensional. They play virtually the same way every time, and the way to beat the AI when it plays it's style is to do just what you described. If the AI changed it's strategy, then it would require the player to also adjust. But it doesn't and so when the player figures out the AI's strategy and what beats it, the player can do it every time and not worry about the AI changing strategy.

Now, I do think that the AI in GalCiv is MUCH better than in any other strategy game and it does adjust it's strategy for minor issues, but not it's overall strategy. It may just not be possible with today's technology to program an AI that is more dynamic.
Galactic Civilizations IV
Reply #3 from Stardock Forums Top
Good topic! Maybe brad can get some ideas for AI improvements from this topic. I agree especially on the Starbase management.
Reply #4 from Stardock Forums Top
My formula's a little different, but I play on smaller maps...

--Rush for trade goods
--Sell the goods to economically break the AIs
--Finish them off when I get to dreadnoughts

Very easy to do. It works because the AIs seem to place a great economic value on trade goods, but they never race to get them.

I can win the game on "Intelligent" without building a single starbase.
Reply #5 from Stardock Forums Top
My thoughts:

1) AI must be more random (and it's must be saved in a savegame so no save-loading will help). Say, some AI (not specified, so no weird things like breaking alliances etc.) declares war w/o any treats (to a random race, not just to a player).

2) Make conquests easier.

3) Gigantic galaxies are unbalanced. Suggestion - give penalty to metaverse score for gigantic(and maybe huge too).

~SDC~
Reply #6 from Stardock Forums Top
>> 3) Gigantic galaxies are unbalanced. Suggestion - give penalty to metaverse score for gigantic(and maybe huge too). <<


You wanna justify that outrageous statement?
:(

~SDC~
Reply #7 from Stardock Forums Top
Don't look for a good justification from the metaverse liberals. Apparently some people think that doing 36X the work as a tiny map on gigantic shouldn't result in a score a few times greater in magnitude.

I don't play gigantic (or at least haven't yet at any rate) because I'm not that patient - and my scores *rightly* reflect that difference.
Reply #8 from Stardock Forums Top
I think this requires machine learning of AI. I don't think the game currently supports this feature. It's quite complicated.
Reply #9 from Stardock Forums Top
That formula may well work for 'Intelligent', but the fact is that the higher AI levels actually play differently to the lower ones... it's not just a matter of the AI getting more of an economy bonus.

If you want a tougher game, try playing a tougher AI :)

~SDC~
Reply #10 from Stardock Forums Top
Ryan Clifton:
"I've read a number of threads and talked to some other people who play the game and it appears that a lot of people are coming to the same conclusion: There is a tightly defined formula for winning Gal Civ.
This formula is esentially,
1) expand like crazy
2) research straight to dreadnaught
3) build up military to at least par with the next strongest civ
4) sit back and wait for tech or culture victory"

I disagree. Your statement more or less true for the big
galaxies, but on tiny and small the play is much tighter and varies more. I mainly play tiny/small vs incredible and there are many differencies:
1) where? With AI knowing where yellow stars are you can't count on grabbing more than 1 or 2 planets.
2) how? AI has big bonuses and at the start I'm usually nearly broken and behind in tech. I don't remember ever getting dreadnoughts before AI. To get them I usually have to build/buy some trade good and exchange it for basic techs and then research dreadnoughts myself. If I'm too late to get trade goods I have to open battle axe trade to get basic techs.
3) This is hard in the beginning, because of the same things as in 2). Besides it's hard to maintain military power and trade battle axes for tech at the same time.
For me it's usually sort of balancing when to pay off
and when to go to war
4) After I've conquered somebody and took 2-3 more planets I can usually choose what victory I want, but to reach this point it takes plenty of maneouvring.
Reply #11 from Stardock Forums Top
The original poster basically came to the conclusion that this is a 4X game, which isn't that stunning of a revelation. ;)

Ellestar, the last thing I want is bizarre random AI personalities/decisions. I had my fill of that in SMAC, where Miriam would declare an undying Vendetta against me when she had 1/1000th of my strength. I'll take the GalCiv AI hands down over that silliness.

I believe Brad added in some different AI decision paths in 1.04, but since I don't have access to that yet I can't comment on it.

I agree with the AI's use of starbases, they upgrade the resource ones pretty well (adding resource modules and defense modules) but the non-resource ones always suck (at with a 5 BC cost each that's an expensive paperweight). I don't know if this is the result of code not existing or a defect in the existing starbase AI code that keeps what is written from being applied correctly.
Reply #12 from Stardock Forums Top
I'm with Alex.

I generally play medium and large maps. I'm up to Normal at the moment.

I find every game, I have to adjust how I play. Especially on medium size and under. On those sizes, the resources you have nearby tend to determine your victory path.

My overall strategy is to go for trade goods, build up SBs in strategic locations and on strategic resources, and work toward the morale boosting and battleships. In half my games, my battleships are capable of standing up to the computer's dreadnaughts, as while they've ran straight for battleships and dreadnaughts, I've picked up a lot of extras on the way.

The computer does use it's survey ships. It just isn't very good at exploring unless they pop a scout or drone. If they get a drone, they'll plough through the map, uncovering it all. If they do not, then they end up just sitting around. The AI never seems to research any of the sensor/drone techs until after it has achieved its primary tech path.

Change your galaxy size, and you will find your formula for victory fails.
Reply #13 from Stardock Forums Top
I think the concerns are right for one given size and difficulty.
For instance, on small maps the strategy of selling trade goods works up to but not including incredible ai's:
Good luck getting an incredible ai give you a penny (except a few minors).
My questions are: What can change the equation?
I see two answers:
-Propose an alternative to dreadnought in terms of military deterrant. A turtle unit which makes the ai not want to attack you but is useless in attack.
-Propose an alternative to general hostility: Make trade more efficient in terms of keeping buddies. Since you can hardly trade with everyone, someone will always bully you, but you might hope that they would take into consideration that you are friends with the big guy next door, who wants to protect your trade routes. Currently, the big guy thinks it's better to conquer you than to trade with you if they have the military.
There is also the point of 'expand like crazy'.
There are options to prevent that, for example playing tiny/rare stars. I don't know if anyone tried to win a game by using a single planet and building lots of starbases instead. I am afraid the upkeep of starbases would kill that option, but it might be worth trying on a bigger than tiny map.
Reply #14 from Stardock Forums Top
Isn't the point of the metaverse that we submit our games that we have won and if there is a pattern that the majority of people are using to win then the designers will try and counter it in updates to the AI? Perhaps I am just naive but thats how I understood it.
Reply #15 from Stardock Forums Top
I read from the gamespot interview that the game is supposed to be able to download not only updates to the AI, but you're supposed to be able to download an AI that is supposed to resemble the tactics of the player that submitted it, in a sense allowing you to challenge the strategy employed by the player submitting the game.

I'm not sure what happened between the interview and release, but I was rather disappointed when I didn't see this feature in play

SeanB
Reply #16 from Stardock Forums Top
Laugh, the tactics I use to beat the computer, a human
player would walk right over. Playing a computer vs
a human is an entirely different affair. I can do
either equally well, but when I do one, its to the
exclusion of the other. It would be amusing to see
how easily a human player can beat a computer simulation
of my style.
Reply #17 from Stardock Forums Top
To Vorlin Varelse

Vorlin Varelse: --Ellestar, the last thing I want is bizarre random AI personalities/decisions. I had my fill of that in SMAC, where Miriam would declare an undying Vendetta against me when she had 1/1000th of my strength. I'll take the GalCiv AI hands down over that silliness.:

(Re to) Ellestar: --1) AI must be more random (and it's must be saved in a savegame so no save-loading will help). Say, some AI (not specified, so NO WEIRD THINGS like breaking alliances etc.) declares war w/o any treats (to a random race, not just to a player).--

1) No weird things is a must.
2) More important here is that comp can POSSIBLY start a rush. Even 10% will be good.

~SDC~
Reply #18 from Stardock Forums Top

I think much of this boils down to: Ctrl-N is too powerful.

I think one solution is not to wimp out Ctrl-N but rather keep how many times a player restarts their game in mind when doing scoring.

If you get a great starting position, you can usually win against opponents who aren't cheating.

Reply #19 from Stardock Forums Top
Instead of attacking the "Ctrl-N" issue, why not create a score modifier based on the number of habitable planets with X sectors of Sol/habitable planets on map? Another way would be a cumulative starting PQ total. That way you might get the superstar players using Ctrl-N to get to an awful starting position to get a higher score.

I think this thread started with a non-Metaverse/Ctrl-N issue, the rush to Dreadnaught. Either randomizing techs or better balancing defense ships so Dreads are not so strong seems like a solution to that issue.
Reply #20 from Stardock Forums Top
"I think one solution is not to wimp out Ctrl-N but rather keep how many times a player restarts their game in mind when doing scoring."

Frogboy: the problem with your proposal regarding scoring and Ctrl-N is the casual player who doesn't think about these things gets penalised while the guy who does just starts a new game from fresh.

I think I've used Ctrl-N about twice. If the start is real bad, I'll play it for a year or two just to verify my suspicions, then I just plain old start over.
[Message Edited]
[Message Edited]
[Message Edited]
Reply #22 from Stardock Forums Top
You broke the board :)

On scoring, CTRL-N'ing: I also think it shouldn't be too hard to calculate starting position into the score.

Wonder how this will come out :)
Reply #24 from Stardock Forums Top
Lets say that control-N was taken out of play, you're still going to have people who will quit all the way out to the main menu and then come back in. And it would be very hard to detect the number of restarts there. Either way, a person patient enough can defeat the Control-N "nerf".

It would probably be much easier to score the game based on the number of class 15+ planets within a certain radius of the player's starting position and go from there.

SeanB
Reply #25 from Stardock Forums Top
--1) AI must be more random (and it's must be saved in a savegame so no save-loading will help). Say, some AI (not specified, so NO WEIRD THINGS like breaking alliances etc.) declares war w/o any treats (to a random race, not just to a player).--

I believe I covered this problem in an earlier thread. They are far too predictable in that a tactic I use against one empire behaves identically to another. All empires seem to research the same family of technologies. The all rush the Dreadnaught. They never trade military techs on a player initiated trade. The only time I've seen them give up planets or tribute based on a player demand is when you're at war and you drastically overpower them.