GalCiv2 AI vs other games' AI

Red Alert is not in the 4x genre, it is in the misnamed RTS genre.

I've read lots of comments on this forum as well as reviews saying that this game's AI is much better than other 4X TBS games... since I hardly played any other games in this genre (aside from CnC2 Red Alert), I am curious then, in what ways is the AI better? So for you veterans who had the luxury and experience to compare, do share your observations...

It is and it isn't better. The AI in this one is clearly a lot better than many of the older 4x titles. Some titles have too complicated of systems that are difficult to program AI for. Or ones that include tactical combat usually end up lacking in the AI department.

There are ultimately holes in the AI which allow for all sorts of gamey tactics, but it's getting there.

Red Alert is not in the 4x genre, it is in the misnamed RTS genre.

I wouldn't say it's a misnamed genre. Although the real-time is not true real-time it sounds a bit better than "Unpausable Strategy" game or "Non-Turn-Based-Strategy" game. If your complaint is that it's called strategy, most games that are strategy games are not strategy in the military sense, very few model any sort of "realistic" strategy, it's strategy in the way some would call a puzzle game a strategy game. And you can't get away with suggesting that a game like RA is "tactical" (again ignoring issues of realism) because it deals with issues that are abstractions of both the strategic level and tactical level.

RTS stands for Real-ime Strategy. It is not, it is Real-Time Tactics. There is little or no strategy in any RTS game I have ever seen, even the choice to invade one country or another first, or one country from one border or another, is borderline. Here is a link that will explain the distinction in terms of games and game design:

Link

Figuring out whether it is better or not is hard and somewhat subjective. There are no TBS games where the AI is all that smart.

In my all time favorite SMAC, the AI did a horrible job of colony development and force deployment. It built bad units and depended on rush tactics to win. I suspect that AI cheats were built in to keep the game competitive. That is the common way to keep the AI competitive in TBS games.

IN GC2 the AI plays on a level field up to tough but it still makes some gaffs in some games and does better in others. Brad keeps tuning it and sometimes the tuning has unexpected consequences but over the long haul it will get better (+2-1).

In Civ 2, every enemy would get mad at you at the same time. That was stupid.

In every game, the computer would build stupid improvements. Granaries in every city, propaganda centers in the middle of its empire. I guess that still persists.

The computer always throws away its defenses against unbeatable units, fighters versus armor and stuff. I rarely have unbeatable units but I don't see too much of this.

The computer in Civ 2 played awesome stalling defense on Deity, including nuclear weapons on the cities you had just taken from it. But you could sit there without playing forever and it wouldn't come kill you. This is something GCI did better than GCII but there's still some true aggression here. Basically Civ civs would always let you seek peace, even if you were very weak. They let you seek peace and then break it a lot too. GC doesn't do that.
You know, I can't remember much about the AI in MoO or Master of Magic. I just remember what I could do and how cool it was. I do remember that Civ III's AI would do awesome counterattacks where you would strike at its home and instead of just playing defense you would find a column of swordsmen in your rear.

I agree with Franco fx - in fact, a good example of this is recent developments in AI awareness of an impending lightning-strike invasion.

In previous iterations of GC2, the computer was exceedingly vulnerable to a well-planned and well-orchestrated sudden invasion of multiple planets. And while the AI is still vulnerable to this, I would drop the "exceedingly."

In several games I've gotten warning messages from AI's about positioning my transports near its worlds - and often the messages were not just warnings, they were declarations of war (i.e., "How dare you position your transports near our planets? We declare war!").

But at the same time, I've also had civilizations declare war on me because of transports when my transports were not all that close - and threaten me even though I'm gathering forces near one of my own worlds, which just happens to share a system with one of theirs. The AI can be a bit *too* sensitive now. As Franco said, +2, -1.

Mistoffeles is true.

The concept of strategy and tactics are both differents. RTS games are in fact tactics games.

The strategy would be taking place in the pentagon, not on the battlefield, if you understand

You could interpret that as paranoia... if an alien race landed on Mars, and then started massing troop ships nearby, what would you think? I'd rather the AI were a bit oversensitive than undersensitive. I've never had war declared on me for this reason. By the time they see that my troop transports are a bit near to their planets, it's already too late...

There's a couple of things that, to me at least, seem relatively simple, and would make a major hike in the difficulty of the AI...

1. Colony management: The AI should always start building a colony with at least one, preferably two factories. They should leave bonus tiles clear until last (unless the building they are planning to build is of the right type). Bonus Influence, farming and morale squares should sometimes be built over with something else, but Manufacturing and Research tiles should ALWAYS be used for their respective building types. If all they have left is a research/manu square and they really want to build something else, they should upgrade another building rather than build it on the bonus.

2. Troop transports: I understand that getting a computer to coordinate multiple fleets is processor intensive. But if they just coordinated one fleet of troop carriers with one fleet of warships, they would be much better at taking planets. If the troop ships just follow the warships everywhere, then invade the planet when the warships take down the defenders, this would make a huge difference. At the moment, there's just too long between the warships hitting the defenders and the troops arriving. The computer could then stop wasting resources on those useless Escort ships, which only work against the 'Cargo hull full of engines plus one weapon' exploit.

I compare it to moo games for fighting. The combat bug here has been fixed. The attack first exploit that made defense impractical should make a huge difference. In the moo games you had tricks in fighting by using missile ships and playing the running away fighter. This means you can only compare the ai based on the ability is has at building and diplomacy. This game is amazing at this. There are a few problems but everytime you notice one they fix it in the next patch. Example: ai building too many culture buildings. The diplomacy works well but there are some problems that allow the player to dominate this area but again they are continually trying to fix this in the patches. The ai started off maybe not as good but it is getting better and better and I have found it very challenging. I still win but unlike some of the other 4x games I find I have to think more.

I said this in another post, but it bears repeating. The GalCiv2 AI works best when it is doing high-level thinking. Who it should attack, perhaps what techs to research, making proxy wars to soften up an opponent, etc. It's pretty good about that sort of thing.

It fails when it tries to actually execute some of those actions. It doesn't have a clue how to fight, for example. It doesn't know how best to use attack ships, or how best to prepare defenses. And it only has a minimal clue as to how to build a good world.

Even the SMAC AI was better at attacking and defending (when they had tech pairity with me) than GC2's AI.

The concept of strategy and tactics are both differents. RTS games are in fact tactics games.

Not really.

Resourcing and building units are strategic concerns. RTS games put them into tactical situations, effectively shrinking a strategic scale game and putting tactics and strategy next to each other. RTS games are an entirely unrealistic hybrid of strategy and tactics.

An RTS game that didn't involve any resource management or unit building (like the MechCommander series) are tactical games, not strategy.

RTS stands for Real-ime Strategy. It is not, it is Real-Time Tactics. There is little or no strategy in any RTS game I have ever seen, even the choice to invade one country or another first, or one country from one border or another, is borderline. Here is a link that will explain the distinction in terms of games and game design:

As I person with a large collection of military theory and history books and having spent a long time studying it, I can refer you to better places than a link on game design (which fails on many accounts at understanding the difference between the two).

Production, research, resource allocation, etc are all strategic. The RTS genre abstracts two levels of warfare into one setting (it abstacts the strategic aspect into the tactical battlefield).

That link is way too simplistic and number of the things it claims as strategic aren't - or they apply to all three levels (if you subscribe to the theory of strategic -> operational -> tactical). What a RTS misses the most is in fact the operational level.

I said this in another post, but it bears repeating. The GalCiv2 AI works best when it is doing high-level thinking. Who it should attack, perhaps what techs to research, making proxy wars to soften up an opponent, etc. It's pretty good about that sort of thing. It fails when it tries to actually execute some of those actions. It doesn't have a clue how to fight, for example. It doesn't know how best to use attack ships, or how best to prepare defenses. And it only has a minimal clue as to how to build a good world. Even the SMAC AI was better at attacking and defending (when they had tech pairity with me) than GC2's AI.

Good way of looking at it. The AI definitely does best at what to do not how to do it. I've finally realized I can stop panicking when the computer declares war on me, even at the higher settings it has trouble actually planning an efficient offensive.

the diplomacy part of the AI is better in GalcivII (in the demo at least, I'm sure its revamped in the main game through patches, etc) and I find that its more interactive and somewhat more feedback.

the diplomacy part of the AI is better in GalcivII (in the demo at least, I'm sure its revamped in the main game through patches, etc) and I find that its more interactive and somewhat more feedback.

1. Colony management: The AI should always start building a colony with at least one, preferably two factories. They should leave bonus tiles clear until last (unless the building they are planning to build is of the right type). Bonus Influence, farming and morale squares should sometimes be built over with something else, but Manufacturing and Research tiles should ALWAYS be used for their respective building types. If all they have left is a research/manu square and they really want to build something else, they should upgrade another building rather than build it on the bonus.

They already do this in general. 

great, the forum wierded out on me when I posted and had to back up and now I can't edit my origional post

I meant that the diplomacy part of GalCivII is better than the diplomacy part of the AI in Civ4

edit: OOPS! I see my double post up there, no idea how that happened, I was posting and then it gave me an error and I backed up, and dunno...

I'd say the AI is about the same as other 4x games. Perhaps a tad better, but not much. Unpredictability would add a lot. Sometimes have a race do a military rush right out of the gate, after building a set number of colony ships depending on map size (0 for tiny, 1 for small, 2 for med, 4 to 6 for large, maybe 8 to 10 on gigantic). Sometimes have them keep their transports with their massed fleets. Sometimes have them separate. Sometimes have them build fast armed ships, sometimes with one engine, sometimes with two engines, sometimes none. The current escort ships tend to be too predictable to be valuable against human players.

Keeping the human player off balance would add a lot to the game.

These could be preprogrammed for certain races, if that would help players enjoy the game more.

The concept of strategy and tactics are both differents. RTS games are in fact tactics games. The strategy would be taking place in the pentagon, not on the battlefield, if you understand

I don't agree with this, since you do choose a lot of what you do in the game. You aren't just given a set objective and set resources, you often choose what to use, and what you are doing. This is strategy. To say that Tactics are not strategy is not really true. A strategy in simple terms is a plan, and that is the basis of the RTS games that I have played. Tactics are the way that you go about your plan perhaps, but that doesn't mean that you you can't plan out your actions before you make them.

Well its still isnt it, i bbelive it will get better eventually but for now t cant compete with human players... sure ppl that start have great fun but after you start to play on masochistic the Ai is so predictable and the only challenge is to use all exploits ( ai weaknesses) to win vs a 200% economy cpu. Until something isnt done with the tech trading i stopped playing it ( also becouse i have oblivion ^^).

The GalCiv II AI is quite a remarkable piece of work.
Sure, a good human player will always be able to beat it, but an average player will have a tough time on settings where the AI doesn't cheat, which is an incredible feat.

Civ3 and, even much more, Civ4, also have excellent AI. Soren Johnson gets all my deepest respect for this work (having programmed this AI single-handedly).
Even an average but experienced Civ player will have to play correctly to win a Civ4 game on normal settings.

A strategy in simple terms is a plan, and that is the basis of the RTS games that I have played. Tactics are the way that you go about your plan perhaps, but that doesn't mean that you you can't plan out your actions before you make them.

Actually, it's even simpler : strategy is whatever doesn't include battle itself, and tactics is battle itself.

There are no TBS games where the AI is all that smart.

The Problem is that just nobody knows the lil game called "Arena Wars". It has the best RTS-AI I've seen so far.
It was capable of defeating even the best players without any advantages.

It is not scripted and will learn while the match is going on and adapt to your strategy. Your best bet to beat it is if you play only a short game to a score of 3 and try to surprise the AI before it adapts.

I find that sometimes the AI (in most 4x- or RTS-titles) does things that amazes me. I've had my ass handed to me by many an AI that I thought I had mastered, but who suddenly perform the most brilliant feats of military planning and execution.

In GC2 I had this experience when I attacked the Iconians and was repelled. First off, the grey bastards refused any kind of ceasefire. Then they attacked me, and focused all their resolve on my core worlds. I thought I had the situation well in hand, redeploying my fleets to the core, but suddenly not one but two flanking forces of really fast ships appeared in unison from out of sensor range at different ends of my empire and crushed my outer worlds. I simply wasn't prepared for such an attack, especially whith so little warning as I had come to rely on the eye to keep me safe from the usually slow AI ships.
The fleet assaulting my home worlds retreated and regrouped around the latest members of the Iconian refuge. I had lost a third of my worlds to what was really quite an unsignificant fleet.

I've never seen the AI do anything even remotely like it since. My guess is that it's really nothing but blind luck for the AI, but I can't help but feel humbled by the potential I saw on that faithful night.

Well since the only space turn based strategy games (i REFUSE to use the Quicksilver MOO3 made up term 4x) in recent years have been Space Empires 4 and Master of Orion 3 I think this game has incredibly smart AI. Compaired to those games its near humanlike.

If I compaire it though to Civilization and Alpha Centauri and those games i'd say the AI is simular to SMAC if not slightly worse but FAR better than any Civ game.

There are no TBS games where the AI is all that smart.

Wrong. I've played a few... SMAC, Silent Storm, Heroes of Might and Magic, Fallout Tactics... the Tactical games usually have good AIs.

by JamesACG Thursday, May 11, 2006 1:27 PM : As [a]person with a large collection of military theory and history books and having spent a long time studying it [...]

My mom has a large collection of cookbooks and has spent a long time studying them.

Everybody in my family (except her) find her cooking quite amateurish.