bluebeholder bluebeholder

What is the goal of the AI players?

▼

What is the goal of the AI players?

from GalCiv2 Forums
Because I sure don't think that the AI players are trying to win the game. Which is quite disappointing since the game was touted as having a great AI.

I've been steadily increasing difficulty and am now at Tough + 2 (I forget their names so I just remember how far away the difficulty is from Tough- which is the difficulty where the AI plays its best game but doesn't cheat). I play with maximally abundant planets and habitable planets since it supposedly gives the AI an advantage.

Yet So far I have not seen a single game where any AI player has established a competant strategy. For instance:
-Military might is worthless without planetary invasion. The computer doesn't realize this and focuses heavily on military before anyone even gets close to invasion. A good strategy would be to tech up their economy while there is no possible risk of an invasion, then shift over after about 6-9 months (depending on aggression) once planetary invasion becomes feasible.
-the AI doesn't specialize it's worlds, the worst example of this is building starbases on every world, even on worlds where the starbase is essentially a lost tile since it won't be building a single ship ever
-It prioritizes farms and embassies very heavily. Okay, these things serve a purpose, but the computer seems to think it can beat you by making more food and out influencing you (without researching influence techs).
-It fails to create a decent economy. This is because even when military might has no impact on the game, the computer heavily focuses on military. Meanwhile the human player can get about double the computer's economy and steamroll as soon as the player decides to get planetary invasion.
-Computer players have poor research paths. It seems like all they ever research are weapons and shields. The other things it does research, it stops on the wrong levels. Examples: It stops on Research Academies which are worse than the tech below them (due to exorbitant build costs on the Academy and almost no boost to efficiency), it actually upgrades the farm techs, and on economy it stops on banking centers (which should NEVER be built under ANY circumstance, since the next tech up is the stock market which costs a third as much and is about twice as powerful all mods told). Etc.
-It builds starbases without upgrading them much. What's the point of that when the last upgrades are always worth the most?
-No AI players tech toward enhanced governments. This is ludicrously idiotic. They're quite cheap and provide insane bonuses. Not to mention that the path boosts your diplomacy which is also quite valuable.

In short, there is almost nothing that this AI does correctly from the persepctive of trying to win. I give the AI props for being "human"-like in that it's diplomacy is rather well done, but as far as being a good opponent, I give this game's AI a D-.

Anyone else have thoughts on this matter?
17,742 views 48 replies
Reply #26 from GalCiv2 Forums Top
I'm glad I'm not the only who thinks this way as the first few replies had me a bit worried!

Gauging from others thoughts, it looks like 2 nice boosts that the AI could get- that don't even involve modifying the AI- are:
1) simultaneous attacks (hooray!)
2) balancing the tech tree
[of course there are also several good AI changes being suggested in this topic, but I would like to focus on things that don't involve the AI for a moment since I have pretty much already said what I have to say about the AI]

So if we could get a list of the over/underpowered techs (and why) that oughta help the Stardock guys out. I was looking through the XML and apparently one of the patches aleady changed several techs, so I will just list the ones that I have used since 1.11.

Overpowered techs:
-Interstellar Republic (prereqs are pretty cheap and you get tons of bonuses going up to this. Having to get elected is not much of a disadvantage, since you generally want high approval anyway)
-Stock Market (the building is cheaper than the one before it, and it's a ton better too. The Stock Market should cost more to build (the banks should also cost less) and maybe the tech itself could cost more research points?)
-Neutrality Learning Centers (way... too... cheap. to research the tech, I mean)

Some underpowered techs:
-Any farm upgrade (this annihilate morale if you upgrade them too high even though the mods are rather minor. I see no way to make these feasible without adding auxiliary mods to upgraded farms like the Stock Market has).
-Seems like there's more, but I can't think of them...
Reply #27 from GalCiv2 Forums Top
Let me deviate to an analogy with chess. Many players benefit from book openings. Same is true for 4x game players. Certain set moves for the first 5-10-20 turns are far superior to what most people are doing at the start of the game. It may take some of the fun out of the game if someone mimicks another human player. I could write out step-for-step what I do the first five turns. Players that mimick my moves would probably do a lot better. However, I don't even think what I do is close to optimal. Half the fun is discovering it on your own in any case.

Instead, what I suggest is a player start a bunch of games and retire after the colony rush is over. Experiment with different slider settings, different patterns of rush buying, of initial tech research. If a player pays attention he/she will find a much better path than they are currently using. A diligent person might perhaps even save the game on turn 1, so the map is the same for each, and choose different openings. I did not go that route. It may seem boring to some, however, this is my single best suggestion for players that are struggling at the game who want to improve.
Multiplicity - Control multiple PCs using one keyboard and mouse
Reply #28 from GalCiv2 Forums Top
From where I sit, the AI seems to prioritize defending its planets surviving and gaining influence.


Actually I think they are concerned more with not losing. The way they develop seems to indicate they are worried about some early attack rush hence they start wasting resources going for tech and building defenders when in all likelihood this never happens.

I have suggested several things that would the game more interesting. Have some AIs bent on war from day one, no tribute, no gifts will sway them, only the sword and their bloodlust only satiated after one race is eliminated from the game (hopefully not the human). Have some AIs prefer fast ships, at least one engine on every design from day one, and increasing as the game progresses, with an average of 40% of ship space going to engines. The Yor with their miniaturation advantage would be extra scary if this flag flips for them. Have some AIs prefer resources to colonizing planets and build one early constructor for every colony ship. This would spice up the early colony rush.


Yeah, it would be nice if some of the Ai play more risky strategies. How about playing like the typical human style, of building ZERO military until they get threatened/ when it is necessary? Sure they could be exploited by a early attacker human, but it would be only 1 out of 9 AIs and it this Ai surivives it would be pretty powerful since it doesn't waste it resources on defender or on researching military techs. This is the strategy i use most of the time and it works way too easily because i gain a crazy lead...

The top 5% remark, I find hard to believe. There are a ton of experienced 4x game players out there that cut their teeth ten years ago. GC2 is not that different from those earlier games. Any of those veterans can wipe the floor at crippling and above. I am not one of the top players (in tourneys etc.) and I can beat suicidal the majority of the time. The top players would walk over me, just as easily as I usually walk over the AI.


The fact that there are players who can beat you doesn't mean you are not among the top 5%, the fact that you play in tourneys automatically makes you one of the better serious players. Besides People who are tops in 'tourneys' are within the top 1% to 0.01% probably. I can safely say you and them are not remotely considered normal players.

The interesting thing is that top 5% remark was about beating tough not sucidal! So they actually made a far stronger claim. If they said only the top 5% could beat sucidal, I might agree. 200% bonus is quite a lot, even if the AI is brain dead to work against.

But for sure saying only 5% can beat tough is totally off the mark.





Reply #29 from GalCiv2 Forums Top
I just had 2 more thoughts after that last post (he's right that it all boils down to your opening).

1) What if all players started with less money? Maybe 2000bc?
2) What if players did not start with a colony ship?

The first one is to help the AI out and hurt the advanced player. It should have no effect on newbies. Reason is that the advanced player spends his money, and the AI ends up trading his money away to the advanced player.

The second one is to help newbies and the AI, but should have little effect on advanced players. The reason is the initial colony ship has 100M people which will take forever to grow into a positive tax paying colony without 100% morale (unless you land first, which AI and newbie won't know to do). It's also slow as hell, thus setting a bad precedent for what a colony ship should be.

Good ideas? Or am I starting to stray too far off topic on my own topic? (Sorry, I just think a bit too holistitcally sometimes)
Reply #30 from GalCiv2 Forums Top
I think Delta is a bit unfair.

The fact that Human players know which techs are good or where to stop at, which wonders are useless and the AI doesn't, definitely IS the AI's fault! I disagree that there shouldn't be useless (or at least not so useful) techs , if all techs are equally useful (impossible anyway), then what strategy is there in deciding what to research or build? You might as well play rock, sissors, papers...

There can be bad techs or at least less useful ones, it's a "Skill tester", some humans who are not so good in the game may even like them! It's not a problem, unless the AI is programmed by one of such humans.

Some underpowered techs:
-Any farm upgrade (this annihilate morale if you upgrade them too high even though the mods are rather minor. I see no way to make these feasible without adding auxiliary mods to upgraded farms like the Stock Market has).


There is practically no point in building farms until you are near or at the population cap, I was spying on the AI, and it was building crazy number of farms one after another (on one with a manfacturing capital!), for no reason at all? Do you need 4 farms to produce 21 mt for a 5b+ population??iI's amazing that the AI doesn't know that! What's so difficult about hard coding absolute rules like that? To use the Chess analogy, it's like knowing the general principle of not bringing your queen out too early or that a queen is worth more than a bishop, except that the rule about not building farms unless you need is , is much stronger.

To be honest, I used to think I was beating the AI, mainly because of exploiting first strike and/or focusing on purely non-military techs while tricking the Ai so they wouldn't attack me. I could understand why these areas were hard for the AI.

People complained about execessive embassies, and I thought it was just them reacting (badly) to the human plaer's influence, some unexpected effect maybe...

But in fact, the colony management of the AI, which I would have thought be the easiest and best part of the AI because they could micromanage like no human, was equally bad. The first time i decided to spy on them (using cheats) , I was shocked at some of the dumb things they do, no human player who knows the game would.

I don't expect deep level strategic thinking like "this colony is near the boundary to the Dregin and it looks like they are hostile or going to be soon, so I better use this colony as a industry world to produce ships in defense..." but I expect a certain level of efficency in the way they handle their colonies, it's just pure math after all an area where they should be whizes at.

I expect them to know not to build manfucturing capitals on lousy low PQ planets, or to refrain from build them too early on colonies, when the 50% bonus doesn't do much, or to fill their manfucturing capital colonies with farms and whatnot. You might think these are trival , but these things add up.

They should definitely priorize on building improvements that finish sooner, so their effect kicks in earlier.









Reply #31 from GalCiv2 Forums Top
I must admit that I am a bit dissappointed overall in the AI thus far in GC2. Afterall a major mission statement was improved AI verses GC1. The games are actually pretty different so maybe the AI is indeed more sophisticated but there are still too many weak areas. For the record I play either at painful all or with 1 or 2 races also at Crippling.

I am also in the camp that this is a great game but the deveopers should be reminded that the AI should be the main focus for improvement over features. I 'm not sure how big this 'camp' is but I'm certain its not jus a one man tent

We can see in the thread that there is a fine line between someones cheese and anothers 'weak AI'. My 2 cents on this is that if players can use a speific exploit then the developers need to look at this and counter it as failry as possible through good code. The developers create the framwork in the first place.

Blueeholder is a player who has taken time to play the game on adavnced levels and has been able to find various ways of exploiting the AI this would not have been possible without many hours of play. This is a GAME and I think this is totally acceptable - a player pits their wits against the AI be it cheese or otherwise. At the end of the day it is up to the developers to counteract cheese (loopholes in code really) and to basically improve the AI in general for ALL players. Its like the cheetah and the Gazelle - simple evolution.

Anyway I am ranting now - Lets just say Stardock promised vastly improved AI. I dont think that are even close to this yet. For what its worth I will repeat or add to some of the existing comments.
- The AI overall still fails at building planetary improvements optimally - Wrong tiles, wrong choices, wrong order, etc. this should be imo the easiest to code even a rule od thum would be better eg Alway build correct type on squares with bonues. Never build capitals or trade goods on bonus squares etc. Many ppl have stated this is an area to focus on - come on stardock!!
AI is GC2's bread and butter. Graphics is for Far cry and upcoming Crysis (wow!), evolution is for CIV4. AI is (should be)GC - please dont loose track!
I personally have not seen the AI react (as promised in journals) when 1 AI starts rampaging. I have had only had 3 games on >= 1.1 though.

Final comment, am I the only one that thinks that plabetary defense is too weak? Planet defenes should give much higher bonuses imo. Plenetary defense is at least an area for possible improvement in the game. At the moment I never build Planetary defense improvements - just not worth it. Easy Fix - Ships in orbit should fleet automatically as you have all the logistics needed planet side!

In Summary - Work on AI, if you think it simproved work on it again and repeat...









Reply #32 from GalCiv2 Forums Top
if all techs are equally useful (impossible anyway), then what strategy is there in deciding what to research or build? You might as well play rock, sissors, papers...
There's a big difference between being equally useful, and being equally powerful. Take for instance the 3 tech lines of weapons we have, they are equally powerful largely but they are not equally useful in the same situation. Here lies the key to this, not every improvement/tech is meant to be useful in all situations, but in theory they can be equally powerful. A simple example is the 3 types of 'centre' you can plonk down on a planet; they are largely equal in power but on certain planets one can be a lot more useful than the other two.

The problem is that certain improvements are not useful in any situation that can feasibly happen, or due to further game mechanics the benefit they produce pales in comparison to their cost. Thus they are deemed useless.

The rock/paper/scissors analogy falls down on its face because it is pure dumb luck, there is no overall strategy involved and your choices are not dependant on outside factors. While improvements are dependant on outside factors, such as your overall gameplay strategy, your short-term economic situation, what your neighbours are doing, and so forth. Ergo even if all improvements are made equally powerful, they may not be all equally useful in a given situation and vice versa. It is not impossible to achieve this, especially in a closed environment where the developer has control over most of the variables with only a single human to mess it up. It's hard sure, but if MMOGs can achieve a close approximation to balance when there are thousands of humans to mess it all up, I'll bet the coders here can do the same for GC2.
There can be bad techs or at least less useful ones, it's a "Skill tester", some humans who are not so good in the game may even like them! It's not a problem, unless the AI is programmed by one of such humans.
Except all you're doing there is handicapping those who don't sit down and min-max everything. Sure those of us who like to analyse every facet of the mechanics to squeeze out every ounce of power can simply skip over these, lord only knows I did that for years playing MMOGs However, if you then program the AI to do the same then what exactly have you acheived that wouldn't be done by simply removing those improvements with no replacement? If on the other hand you improve/replace the offending techs/improvements, surely you have diversified the game and made it just that little bit more interesting? Isn't that A Good Thing(tm)?
Reply #33 from GalCiv2 Forums Top
After easily winning the Dread Lords campaign on level Tough, I'm currently playing my first Metaverse game on level Painful (1 above Tough) on a gigantic map with everything abunant and tight clusters. The Torians are performing the best among the AIs due to their high population growth. I'm currently taking out the Korx and expect to easily win this game, despite the bonuses given to the AI. The AI is building up its' planets very well (see original post) and not making effective use of economic starbases. With tight clusters economic starbases make a world of difference!
Reply #34 from GalCiv2 Forums Top
I usually play on Painful or Masochistic on large galaxies. While I understand the game mechanics more than most people might, I don't take special pains to take advantage of special circumstances and I don't worry about falling into the disadvantages of certain techs either. I still win virtually every game! I still take over planets in the late game that only have 4 tiles built, or planets with a manufacturing capitol but only one factory, etc, etc, etc.... I see all the problems outlined so well by everyone on this thread.
Reply #35 from GalCiv2 Forums Top
DeltaGorno (#24) wrote:
So declaring the AI 'weak' when you've sat down specifically to find out where it is weak and exploit this, seems a little ridiculous.

Well, I agree more or less with your reply up until this. It seems to me like this is exactly how one goes about declaring an AI weak. You sit down specifically to find its weaknesses, prod at them and test them until you understand what the AI can and can't do, and if after all this you find that the AI consistently makes bad decisions in almost every aspect of the game, then yeah, you can say the AI is weak.

Take something like the defenders not leaving planets. Since I know this, I know that it's safe to leave vulnerable fleets near defended enemy planets, just not "mobile" enemy fleets. This allows me to play more aggressively when invading my neighbors, since I know that these enemy ships will just sit around and wait for me to kill them. Once I understand the AI behavior well enough to predict it, I can use this to gain an advantage. I'm exploiting a weakness in the AI, but I can do this because the AI is weak. It's building bad, slow ships to perform a task (defending planets), but that can perform that task only poorly (especially pre-1.2, or without an Oribtal Fleet Manager), and is never assigned a new task (opportunistically attacking invaders). We all draw a line between what's an exploit and what's not, and there's a lot of grey area. But if taking advantage of this kind of AI weakness is deemed an exploit, it seems to me like the word loses its meaning.

The AI and the game mechanics are distinct from one other, of course, but in a sense they're inseparable. It's up to the designers to get both right... in my mind, the AI isn't off the hook for building bad wonders just because bad wonders exist. The bottom line is that the AI simply does not play GC2 well.

(This isn't to say the game can't be fun for people. That's a separate issue.)

Reply #36 from GalCiv2 Forums Top
The fact that there are players who can beat you doesn't mean you are not among the top 5%, the fact that you play in tourneys automatically makes you one of the better serious players. Besides People who are tops in 'tourneys' are within the top 1% to 0.01% probably. I can safely say you and them are not remotely considered normal players.

The interesting thing is that top 5% remark was about beating tough not sucidal! So they actually made a far stronger claim. If they said only the top 5% could beat sucidal, I might agree. 200% bonus is quite a lot, even if the AI is brain dead to work against.

But for sure saying only 5% can beat tough is totally off the mark.


I like the idea of some AIs not building any military ships. This would be cool if they met the bloodlust warmonger race early. What fireworks! The outcome would not be certain either, depending on the map, and how fast the attacking ships are. A human player stumbling on to this early war situation could role play. The human could aid the weaker race with techs, ships, money, or befriend and then backstab, or join in the plundering and speed their demise.

I don't think I am exceptional as a player. I am fairly casual in the way I play the game. Most players that want to improve will read tips and try things. Those that play the same "stupid" way and struggle every game are the exception in my mind, not the other way around. I've been playing the game a few weeks and believe that any 4x game veteran will find it relatively easy to win if they go in with a good plan and have looked to improved.

For a game with a truly excellent AI this would not be true, so I see a lot the AI talk as marketing hype and/or talk from people that are relatively new to 4x games. If the AI were really good, it would be a struggle and a human player would have to be much more disciplined. Again, unpredictability adds a lot and makes certain tactics less than 100%. Sometimes that is all that is needed to make throw the human off balance and make them lose, while still having a fun game.

Like I said, a top player mapping out his/her first few turns (like 20 turns) that would help most of the lower level players immensely if they mimicked those moves. If a player is told that certain moves are good, and still fails to at least try them, they are not interested in improving. Nothing wrong with that, but I believe they are exceptions and that almost all humans want to get better at the game.
Reply #37 from GalCiv2 Forums Top
Bad wonders has come up several times, and while this might not be what the op meant, the first impression I got was of tactical issues. As in This wonder is generically good, it's just that it no longer has any use because everyone now has technology [whatever]. Like I said before, that might not have been the original message, but it is an interesting message: AI should be developed to play a dynamic game of what's going on right now, as opposed to a static game of this is always better than that.
Reply #38 from GalCiv2 Forums Top
I've only had this game for a week, so it is pretty arrogant of me to say this ... but I agree, the game's AI just does things that are stupid. I have seen it build three or four economic starbases next to a homeworld, but not upgrade a single one of them. Does the AI even know what an economic starbase does? As for planetary improvements, I think the AI does specialise ... the wrong way. The homeworlds seem to be covered in research facilities, while one planet I saw had like four farms and two entertainment centres, but it only needed two farms to reach the Planet Quality population cap. Now, this has been on Normal level, since I have only played a few games, but I would expect the computer to be playing decently at this level. Or to put it another way, I would expect to beat the computer because I have a better strategy for victory, not because it has a strategy of failure.

On the Galactic Banks vs. Stock Exchanges issue: each level of building should actually be a worthwhile improvement. Every other 4X game I have played follows that rule. The Banks seem to be the most obvious display that a lot of the 'improvements' are not worth the building cost. It is not a flaw that perceptive players have worked this out and told us all. It is a flaw that it is in the game.

RAS
Reply #39 from GalCiv2 Forums Top
The problem is that certain improvements are not useful in any situation that can feasibly happen, or due to further game mechanics the benefit they produce pales in comparison to their cost. Thus they are deemed useless.


Thank you I fully understand the difference between improvements that are useless in almost all situations and those that are merely less useful in some.

The rock/paper/scissors analogy falls down on its face because it is pure dumb luck, there is no overall strategy involved and your choices are not dependant on outside factors


If you think I'm seriously proposing that, you are crazy, I was just mocking your attempt to blame the poor AI, on the rules of the game.

If on the other hand you improve/replace the offending techs/improvements, surely you have diversified the game and made it just that little bit more interesting? Isn't that A Good Thing(tm)?


I have no problems with such a move, the problem is you stated that the fact that the Ai uses lousy wonders is "not a fault of the AI", which I think is strange.

I'm simply pointing out that the Ai fails not because there are unbalanced techs, but because the guy who programmed it didn't teach the Ai properly.

You seem to think the solution to Ai picking lousy techs is to change the rules until all techs are equal, which as I shown is futile and unlikely, unless you change it into literally rock,sisscors paper stone!

MMORPH aren't relevant by the way.


There can be bad techs or at least less useful ones, it's a "Skill tester", some humans who are not so good in the game may even like them! It's not a problem, unless the AI is programmed by one of such humans.

Except all you're doing there is handicapping those who don't sit down and min-max everything.


Knowing the rules of the games, is the essence of being a good player , demonising it by calling it "min maxing" is silly. Besides how can it be minmaxing, if it's so obvious a tech is useless in most situations?

Is knowing in chess that a queen is worth more than a rook in most situations min maxing ? How about something more subtle like knowing 2 minor pieces is typically superior to rook + pawn unless in the endgame where the pawn is advanced?

Right now the AI sucks , that is a failing of the AI, to blame it on the fact that some techs are not useful as others is pretty dumb. In fact, given that some techs right now are clearly useless in most situations is actually GOOD for the AI, because it is simple enough to code.

If you change it to a situation where every tech is really balanced and requires subtle reasoning skills to figure out what to research, the AI is screwed even worse.

Sure those of us who like to analyse every facet of the mechanics to squeeze out every ounce of power can simply skip over these, lord only knows I did that for years playing MMOGs However, if you then program the AI to do the same then what exactly have you acheived


Your achievement is clear, you created an AI that knows how to play the game better than some players?

What's the alternative, change the game, until it doesn't matter what you pick? The Ai fails because the Ai is bad , changes the rules won't help it, unless the rules strengthen the techs it likes to picks. But that doesn't lead to more balanced techs! It just means the techs it prefers now become unbalanced.

I have no problems making the game a lot more balanced, but it will hurt the Ai even more, believe it or not. As I said at the start, the failing of the AI, is not a fault of the rules of the games!

Reply #40 from GalCiv2 Forums Top
1) What if all players started with less money? Maybe 2000bc?
2) What if players did not start with a colony ship?

Looks like what the higher difficulties are all about, giving the AI bonuses.

The AI should be able to beat most players in this fixed frame, where rules and stats are known and micro-management as well as seeing the overall state of the game (eg knowledge of on-the-whole social capacities etc) are important.
I think the main problem of GC2's current AI is that it relies mainly on its "thinking", and not on any opening book or pre-set plan. It just plays with some basic development rules, chooses what to do in response to what it lacks according to the galaxy's numbers. In this regard, the AI is really quite good and adapts well, I guess that's how it was designed.
But it's not human and its basic adaptation capacity cannot make up for its lack of memory. The game is too complex for it to rely on its superior calculation ability. It would truly be unbeatable by most if it could successfully mix its computerish advantages, a strong memory system, and lots of pre-set plans made by people who know the game and played it thoroughly, chosen randomly and depending on the situation and their previous successes.
You can make a blunt chess AI that will calculate so many turns ahead what would give it more central chessboard control and save it more pieces (most basic rules, rely on blunt "thinking"), or you can make an AI that includes chess strategy, starts with good opening books, can recognize what can lead to victory, that remembers what kind of moves led to a loss, can recognize true long-term threats.

You don't rely on the percentages in the "victory" screen to choose your next move, you know that military potential matters and not just actual military stats, you know that government types can give you outrageous bonuses, you know that in some cases large fleets of small ships are better than expensive ships, you know that small planets are better suited for civilian specializations... The AI doesn't. For Christ's sake it often ignores bonus tiles.

The "blunt" one is a noble choice, and if I were to design an AI I'd go for that because it's the most interesting / least tedious way to do it, like creating a sentient being and making it grow, and it isn't version-dependent. But it doesn't go for the win, it just plays and strives to be on top in the stats.


Not saying GC2's AI should change though. It would require lots of time. Lots. And it's quite good currently. Just some more rules could help (what to do in this or that precise situation) as well as balancing the tech tree (or adding more specific rules), which I guess is what StarDock is doing.


Just my 2 cents...
Reply #41 from GalCiv2 Forums Top
I had a nice reply all typed up, and the forum ate it
You seem to think the solution to Ai picking lousy techs is to change the rules until all techs are equal, which as I shown is futile and unlikely, unless you change it into literally rock,sisscors paper stone!
No no no, and no. Firstly you didn't 'show' anything, anymore than I 'showed' anything, we conjectured and postulated with no absolute truth found. Secondly, if you can't see how things can be made balanced without being R/P/S then I'm afraid there's little point continuing. Many, many games have balance through diversity and succeed excellently at providing a lot of options that can be equally powerful, yet they are not anything like R/P/S.
Reply #42 from GalCiv2 Forums Top
Veblen:

Vanir wrote:

If the AI doesn't try to win, it's probably because the game won't allow it to win. The AI can't win a tech or alliance victory even if it fulfills the requirements. So the AI gets discouraged.


You'd think fixing this would be a priority...


No kidding. It seems obvious that one of the easiest ways to make the AI better is to allow it to win.

Reply #43 from GalCiv2 Forums Top
there is a lot of space between what we want, what we have, and what we're gonna get.

i think the devs will work dilligently within the budgetary constraints of their business plan. i'm sure the AI will improve incrementally in the next couple months; however, i don't feel the AI will ever truly be as "dynamic" a player as a human. there's no budget left for that--the most recent journal stated their roadmap for the next couple months and implied as much.

so this is pretty much what we have.

i think a better contribution might be to consider more "creative" options for improving the AI. has anybody hacked the text files? can the AI be "improved" independently of the devs in some way?

it's been mentioned that perhaps the tech file has "AI weights"--can that be modified for the AI's benefit?

it's nice that there is forum to document and discuss what is problematic with the game. but perhaps it's just as much our responsibility to find other, more creative solutions to the problems if we want something the devs don't have the time and budget to provide (to our high, high standards).

i'm just sayin.

-0.
Reply #44 from GalCiv2 Forums Top
I had a nice reply all typed up, and the forum ate i


Or maybe you just didn't have any reply to my devastating arguments , and this is just an excuse. In any case you didn't engage any of my arguments except to attack the following strawman argument .

Secondly, if you can't see how things can be made balanced without being R/P/S then I'm afraid there's little point continuing.


Stop attacking a strawman argument, I didn't say that. I was saying the AI is weak is the fault of the AI, changing the rules to be 'more balanced' will not help the AI unless every option is exactly the same (R/P/S)!

Your thesis that the Ai is weak because it chooses weak techs is not the fault of the AI is laughable.

Reply #45 from GalCiv2 Forums Top
Interesting read, I hope some of things here get SD's attention.
Reply #46 from GalCiv2 Forums Top
I am pretty new to the game, I've had it for a week and I'm still really into it. I must say that i find this thread interesting, but I cannot say I agree with the general opinions of this thread. I have played quite a lot of 4x games and I still think that the GCII AI is very "human" in lots of ways. Have you ever played against a good AI? Interesting, creative, humanlike, inventive...

Suggestions are good in general, but I think most people are underestimating the difficulty in creating a program that plays such a complex game as GC2. Changing small details could have immense implications. Code that make the computer play "better" oftens stiffens up the gameplay, making the computer boring and very predictable. Obvious in games as Civ4 where i eventually just don't want to play anymore.

The people who have been developing the game have regular jobs (probably working on a sequel ), wives, husbands, kids and real lives. I think they did a solid job before the release of the game, but you can't expect people to practically spend all their spare time to make the game challeninging a bunch of hardcore 4x gamers. Be thankful that they are still releasing patches and keeping momentum while most game developers today just don't bother at all.

I think one of the big strengths of the game is that is doesn't get very lengthy on smaller maps. You can finish a game while it's still somewhat interesting. I have not experienced this in any other game of this genre.

Many of you guys think that you have a solid perception of the AI. The concept of the AI, I think is more linked to the overall playing experience. I agree with you guys that the AI makes very obvious mistakes. But the AIs game is very casual and in some way more "humanlike" than many other 4x AIs which are smarter but insanely boring. I kinda like the bent tech tree also, I wouldnt prefer a balanced one because it's very fun to find out what it takes to get the upper hand and uncover these loopholes.

Most of us really liked the game, and those of you who have played a lot are now getting disappointed because at some stage, replayability is not perpetual.
Reply #47 from GalCiv2 Forums Top
[/quoteNot saying GC2's AI should change though. It would require lots of time. Lots. And it's quite good currently. Just some more rules could help (what to do in this or that precise situation) as well as balancing the tech tree (or adding more specific rules), which I guess is what StarDock is doing.
/quote]

I can not disagree more. It would not require lots of time, just some more creativity, and adding a random element to decision making.

I just posted (I think for the 5th time) my suggestions in this other thread:
Link

The gist of it, is to have flags, Bloodlust, Resource Hungry, Tech Lover, Speedy Ships, and enable them randomly at 25% to 50% at the start of the game to change what happens early. This would spice up the early game and absolutely wreck certain staid human strategies. The human could adapt and still win, but would require early and mid-game adjustments. It would also open up a host of new strategies to try. In general, it would not make the AI any "better" however, it would be unpredictable, and the game would be much more fun overall. The amount of time would not be huge, probably less than 10% of the time already invested, and the payoff would be very big.
Reply #48 from GalCiv2 Forums Top
I wonder how much harder the "Intense" AI will be? Will it make the AI strong enough to be effective?

When the AI becomes moddable I would like to see if people can create good AIs themselves?