Terror Stars

What difficulty it was?

Since the latest version (1.11) I have seen the AI's create several terror stars, but exhibit the same lack of knowledge of what they can do with them.


~SDC~

Mine was on masochistic, BTW.

kasualkid

~SDC~

Maso

~SDC~

I've seen the AIs create terror stars in Maso, but they have never used them.

They will still sell them relatively cheap too!

~SDC~

I've seen the AI build terror stars but never use them, they seems to treat them as starbases.

I don't personally use terror stars I must be too lazy and find conquering a better idea.
However I wouldn't mind if the AI had the potential and if the terror star took a couple of turns to implode a star. Adding to that, the targetted system was fully imformed (ie the player/AI knows they need to act).



~SDC~

never seen one built other than my own on any level.

Actually, I've never even seen an AI START to build a terror star.

~SDC~

I also saw the AI on maso build terror stars and it did get my attention at first. But they never actually used the terror stars they made. So like Ray says, the AI may be unable to use terror stars.

My use of terror stars is rare now except when something goes bad and I fear pending defeat (bad starts gone worse, bad random events, etc.). Terror stars can be a good way to snatch a victory from an otherwise lost cause.

Sometimes they speed up the inevitable because it is easier to build 5 constructors than build the military necessary to take down a strongly defended star (especially if one plays with no military). It would be interesting if the AI could use terror stars but I would not like to see terror stars go because they are kind of fun sometimes.

Very interesting and well-done experiment Ray. Thank you.

I've seen the AI build terror stars. Interestingly terror stars were banned in the universe and when I bought one it couldn't move...

I agree that the AI needs to learn how to use terror stars. Even if they aren't SMART about it, they should at least BE ABLE to use them.


While we are on the subject, the AI could (under the new rules) erect a rather effective defense against the terror star alpha strike merely by building one or two junk starbases next to each of its planets. Player would have to kill them first, which means having to take along a good fleet, too, or else player would have to buy them, and that could be too costly in some circumstances.

Then again, is that really a defense? Or merely the AI being programmed to exploit a flaw in the game design?

I'm having second thoughts about making it more difficult to let the player "alpha strike", at least with terror stars. Sometimes the end game gets quite boring, and this from the guy who mostly wins by alliances and can't be bothered to milk the score with full conquests if another kind of win can be obtained with less time spent. If the player lacks influence mines, taking over 90% of the culture without military attack can be a massive chore, while sometimes a good alliance victory can only come to pass with a terror star wipeout (or longer, slower, perhaps even more boring conquest) of one obstinate non-alliance member. Would the game be any "harder" if I had to escort all my constructors with overlords/excaliburs? Or would it merely grow more tedious and take even longer yet?

I need to give some more thought to the game balance to offer any useful suggestions. Some of the ideas that seemed good on paper a couple of weeks ago no longer look so good to me. One thing for sure, though: giving military conquest games more score, when the AI's ARE SO INCREDIBLY BAD at responding to terror stars, is a joke.


- Sirian

I've seen the AI do that to block culture flipping....

Alex


~SDC~

Are you sure it was a deliberate response to culture flipping, and not just a coincidence? AIs seem to build a lot of junk starbases, especially the Arceans.

Maybe in latest patch stardock has gotten the ai alittle wiser in using starbases?

I don't think so. It looks to me like all the AI's build a flood of constructors when a resource opens (due to war, surrender, etc). Then when someone lands the resource, the rest of the AI's vying for the resource have nothing to do with their constructors and tend to plop them down wherever they were at the time, or in the immediate vicinity.

Besides, how could AI junk starbases prevent a culture flip? Sounds like somebody is spamming the starbases like mad. If you are running out of room and still don't have enough to flip, then the new 3-tile limit would seem to be working as planned.

I've never built more than three culture starbases in any single sector, and not more than one in a sector where I don't have planets. Influence is all about the influence resources. If I don't have them, I either go after them or go after another victory condition.

Oh, and I'd like to see the scoring take more into account. Turning off victory conditions ought to reduce your score. Disabling cultural victory and "conquering" the galaxy with an endless flood of party palaces ought to be counted as a cultural victory. Why even allow the option to be turned off? The only reason some folks are doing that is to exploit the unbalanced scoring awarded to military victories.

The galactic resources in general are the controlling force. Influence and military resources rule all. If you don't have any, winning a cultural or conquest victory is made immensely harder. And yet there remains all kinds of cheese in the game for getting around this, such as AI's willing to trade away their resources and starbases and unable to defend them or even to TRY. Then there's the "park ships next to target starbase, give ships to base owner's enemy". I don't know how to close that one down without breaking three other things. Yet the influence resources are SO powerful, and the military are right behind them. With the AI mishandling them, there are all kinds of ways to con the system. Short of the annihilation strike, sometimes just blitzing enemy influence and military starbases is all it takes, but at least a strike against starbases brings a war. Some of these other tricks have no costs.

The AI doesn't prioritize the resources. It doesn't defend except by pumping up the starbase defenses, which is nothing to a determined attacker. Defense in this game is a joke. GalCiv is just as prone to the SoD (Stack of Doom -- mass your units into a pile and roll over the AI's scattered forces) as any other 4X game I've ever seen. The GalCiv AI only targets bases the same as any other target, beelining to whatever is closest. That is probably the biggest factor making huge maps easier: the AI's unintelligent handling of starbases really shows on a larger map. On a small map, it's "beeline" mentality doesn't handicap it so badly because the distance from its production centers to the targets is so much smaller.

Essentially, though, the issue with the AI and terror stars is merely part of the larger picture of the AI's ineptitude with starbases in general. It does a good job going after resources and building up those bases, but it never protects them with ships, and it has no strategy for using bases beyond either range extenders or resource gathering. Those two points, it does well, but that's the limit of its ability.


Not only is the AI bad at using terror stars, it's bad with invasions period. Does it calculate how many forces it may need to take a world? Does it even try? All I see is pure beeline. Build a transport, send to nearest enemy world. If not for the massive boosts given the AI on Maso, the much-hyped GalCiv AI would not be putting up any fight against the top players.

In the early or middle game, the beeline tactic works OK for the AI. It's massive production boosts and other bonuses can leave it with unbeatable tech and production advantage, and the beeline can roll you up. That should not be mistaken for clever tactics. There is no organization to its battle plan. There is no battle plan. Later in the game, when the planetary defenses are in play and populations soar into the high tens of billions, the AI is still puttering along with the same old same old "one transport at a time" beeline strategy.

In one of my early Maso games, I rode out a situation where I had been completely crushed militarily.

http://www.network54.com/Hide/Forum/message?forumid=237344&messageid=1066152493

I sat and ignored the inept AI invasions while I raced on to a tech victory. That was a heroic game, but the main danger posed by the AI was all the influence mines they had. With four or five in the hands of one AI, as soon as it got a planet within two sectors of anybody else, the flips started to happen. Militarily, the AI was hapless on the invasion front. They did keep targetting the same system until it fell, but only because it was the closest and thus the one it would beeline all its new transports to.

Is the terror star equation any different? The AI is good at expansion, but only because it plays with full spoilers and can immediately beeline its ships to the best stars, and even then it spams the colony ships like mad. The AI is good at trade, but again, only because it cheats (by knowing the map, not having to scout). And again, it spams the freighters. The AI always sends out way more freighters than it can have trade routes, and the extras bounce from star to star until they are wiped out or a new route is enabled or an old one is shot down. The AI is good at building up a military, but that is all it ever does, beeline up the military branch of the tech tree, because any AI who falls behind on military will soon be dead, while the player can either trade for military tech, wait for a sweetheart trade offer, or outplay the AI on diplomacy, often by bribing the AI's to attack one another to keep them too busy to notice the player is fairly well defenseless. If an AI ever started to build cultural starbases along its borders to try to stretch its influence, rather than to build military like it always does, I'd faint straight away from the shock! LOL The AI is decent at research, but always takes the same path -- and anything that is too predictable is strategically exploitable.

Overall, this AI is better than the average, but the best players all migrate to Maso, where the AI's massive bonuses hide a multitude of strategic weaknesses. I guess we'll have to wait and see what improvements the AI gets in the next patch and go from there.

Kudos to Ray for noticing the AI's complete inability to do anything useful with terror stars.


- Sirian

I enjoyed your AAR, Sirian.

I played a similar game months ago when a rampant Arcean (IIRC) empire mowed everyone down.

I retreated all my forces to a sector which was anchored by my Terror Star with military buffs and made a last stand. I won via tech with just 3-4 ships remaining, even rebuffing an Excalibur attack with weaker ships (at that time, I didn't realise that I was assisted by the fleet defense bug). However, it was evident to me then that the AI would need years to capture all my planets (it was a Gigantic Abundant Maso) so I actually had a lot more time to finish my research.

I had a really good score for that game though as that was the time where tech victories scored the highest.

I never use terror stars myself, so I can't say it matters much to me, but the demonstration is well done.
Either remove the terror stars or teach the ai to use them. Currently I only research the techs for the +10 influence they bring.
I would say that the ai also totally misunderstands the power of AMMs. They are cheap and more effective than any ship (since you always want to be the attacker, it doesn't matter that your ship has 0 defense).
Sirian's points are also correct, but then I tend to beeline with AMMs and transports too because it just works. A rough evaluation of the number of transports needed, and maybe an escort (though escorts need to be offensive instead of passive defense) would help the ai a lot in its invasions. But then again, the ai never reaches the point where it can invade me late game.

Did a similar experiment some time ago using 1.09 I think it was . . . gifted multiple terror stars to other races (I built too many constructors so went terror star crazy building bon bons all over the place).

Anyway as you mentioned Ray the TS just tried to make it back to thier nearest sector ownership space. Hopefully 1.12 will nip this in the butt . . . any word yet . . .

Now that you mentioned AMMs, LDiCesare, I've not seen any built by the AIs for a long while now.

Has anyone seen an effective use of the AMMs by AIs?

Yes in my last game the Dregs used them extensively against the Torians. I occupied the space between them so the salvos were quite obvious . . . the Torian surrendered soon thier after.

Sirian, I agree with you that the AI could be improved so that i would handle some of the game options better, but still... sometimes it seems that you expect the AI to make all considerations you do and though i don't know that much about how code is written - my guess is that this would take an awfull if's and when's....

I guess what i'm trying to say is that instead of wishing for an AI with the capabilities of a Human maybee we should try to find flaws which can easely be corrected and make suggestions for those corrections.

Please don't take this an assualt (i don't wan't to start a war here)

As an example the terror star:
As Stydracos noted the AI seems to handle terror starts as starbase instead of as combat transports.

Again i have no idea about how hard it would be to change that

Cool. That's something I've yet to see. The AIs in my games just seem to mass-produce the latest warship.

I've always assumed it took 10 constructors to build a terror star . . . seems it only takes 5 (or so I've been reading on the boards). Opps . . . can anyone tell me the build order?

Regarding modifying the behavior of AI's TSs . . . it mustn't be all that simple. I would've thought just translating the coding for combat vessels into the completed terror star 'things to do' protocols would work (can you tell I know nothing about programming?) . . .

Totally off topic a few games ago I had an AI take a Minor's system which was in my home sector of Sol where I had like two other systems and moral of 100%.

People generaly say don't worry about them invading minors in your sectors of high influence (I had 128 influence from Sol and about 80 a piece comming from the other two systems in this sector) so I didn't put up any palaces to defend thinking the newly conquered system would flip (low PQ and all the rest of it) . . . but it didn't!

I left it for a while (we both finished the tech tree) and they were pretty broke due to a UP vote regarding trade routes for evil races . . . next thing I know there are three little Skull and cross bones on all the systems in my home sector . . . ! ?

This is inexplicable! Any theories . . . ?

It only takes 5 constructors to build a terror star. I can't remember the actual names but they are all under the military module tab.

Hmmm... first is the starbase, two allows you to actually see the partially built structure, then you add the third and fourth constructors until you finally get to constuctor 5, which is where you get the pop-up asking you if you really want to do this and it will cost 5000 bc. If you hit yes, you have that big silver basketball - or your bon-bon - the completed terror star. After it sits a turn it will get a movement of "1" and you are ready for action.

In summary, I send out a fleet of 5 constructors right next to the planet I might want to destroy. Then one turn before the finale, I build the terror star, let it marinate for one turn to gain a movement point, and then it's ready to serve.

I don't know what to tell you about the attempted culture flip of your home sector that you don't already know. I agree that the AI influence boost on maso is really powerful until after you research all the influence tg/wonders and capture a few influence resources.
[Message Edited]