Artificial Intelligence

Valgua

Chieftain
Joined
Aug 10, 2010
Messages
16
One of the best things in Civ5 will probably be the way combat is handled. Without the possibility to stack we will probably see many more battles outside the cities. The player will need to use the territory efficiently to defend his cities.
However, a question is nagging me the last few days. Will the AI cope with this added layer of complexity? Will we see the AI defending itself with competence, trying to stop our troops far away from the cities, perhaps taking advantage of choke points and natural barriers? If this is not the case the new military features risk to turn into a major drawback.
 
lot of testing is needed, I guess - I believe patches will be of the way for development in this field
 
lot of testing is needed, I guess - I believe patches will be of the way for development in this field

Possibly. I hope, however, that Firaxis has paid a lot of attention to the tactical AI and its defending skills during the development of the game. The issue is too complex and too important for the gameplay to be handled in a last minute patch. Without a very competent AI the game risk to become a major disappointment. While the AI has always been important in Civ, I feel that it is even more important now that a very easy and powerful tactic (stacking) has been removed.
 
One of the best things in Civ5 will probably be the way combat is handled. Without the possibility to stack we will probably see many more battles outside the cities. The player will need to use the territory efficiently to defend his cities.
However, a question is nagging me the last few days. Will the AI cope with this added layer of complexity? Will we see the AI defending itself with competence, trying to stop our troops far away from the cities, perhaps taking advantage of choke points and natural barriers? If this is not the case the new military features risk to turn into a major drawback.

I am optimistic that the AI will be able to handle combat. If you watch the E3 closed demo video that showcases combat, you can see that the AI positioned its units on hills and behind rivers to take advantage of defense bonuses.
 
I've always thought that when playing the harder levels, your secretly playing against Sid Meier.:lol:
 
Yeah, I hope they manage to create a competitive and fun AI. The AI was what kinda ruined SMAC for me. It couldn't handle the game mechanics.
 
So far, every preview of the latest beta version I've seen said that AI handled itself good enough during combat. But I am sure there will be many exploits/tricks that a smart player will be able to use in his advantage to level the playfield on higher difficulties.
 
I'm hoping the AI is very good out of the box instead of taking multiple expansions and user-created patches to be decent.

With Civ IV I thought the AI was OK but didn't really get really nailed til BTS and patches, and even then...

Yeah, I hope they manage to create a competitive and fun AI. The AI was what kinda ruined SMAC for me. It couldn't handle the game mechanics.

...IMO that was the problem with Civ IV too. I wouldn't say "ruined" but in a lot of ways the AI's inability to handle some game mechanics was a bummer.

Shafer seems to be in tune with things that needed improvement moving on from Civ IV, and I hope AI is one of the focuses. Since he was involved in BTS and modding, I'm hoping he's intimitely familiar with the evolution of Civ IV AI, even if he's not an AI guy, and that he's prodding his AI guys to nail it.

I personally think many elements of Civ IV and prior versions lend themselves to players exploiting the AI and/or weak AI handling of mechanics, and it appears that some of those elements have been addressed. These may very well be fun things to some players but they've always bothered me. For ex, I think the player has always had an advantage in tech trading (or the AIs end up acting like a research collective). Players use "strategies" the AI never considers, like early military rushes, sacking undefended capitals during exploration, worker grabs, taking advantage of chopping or whipping quirks, and so on.

It often seems like some player strategies include taking advantage of AI weaknesses and to me that's not really all that strategic. It's not really the fault of players - it's usually been about poorly designed game mechanics or weak AI in relation to those mechanics. If you have a better design and better game mechanic, it's easier to create AI that can handle it well. That is what I'm hoping for in Civ 5.
 
AI is a bit like Moddability in the sense that there's no way you can tell whether it's really improved or not (all developer hype aside) until right up to the moment when you have the game in your hands. And in the case of Moddability, sometimes not even then (as mod tools are often something that are published some time after release).

And AI is something that is frequently tweaked in post-released patches.
 
BTS AI as far as combat was pretty good. They did a lot more amphibious assaults and actually attacked your more vulnerable cities, even using spies to destroy improvements on strategic resources. The AI would also make peace with you if they were losing or winning, or make themselves vassals of a rival if they feared an imminent destruction.

I agree there can be a lot of improvement, but I doubt this is Total War, where graphics come first and AI comes last (seriously, you start a war with anyone it in Total War and you almost always are fighting them until you kill them all).
 
One thing I hope is that they balanced the game mechanics so that the AI can just play fully for the win without it being limited by certain rules that the player doesn't have. One example is the early rush/worker grab. The AI in Civ4 was programmed not to declare war on the player in the first 50 or 80 turns or so. Because it would be rather unfun for the player. The mechanism was there though, so players would abuse it. Same with religion influenced diplomacy. The AI (mostly) couldn't backstab a player with whom they had good relations, no matter how weak the player was. The player didn't have such a limitations. So rather then preventing the AI from using such tactics, I hope they have made sure this time around that such tactics are not there or very risky to pull off.
 
AI is a bit like Moddability in the sense that there's no way you can tell whether it's really improved or not (all developer hype aside) until right up to the moment when you have the game in your hands. And in the case of Moddability, sometimes not even then (as mod tools are often something that are published some time after release).

And AI is something that is frequently tweaked in post-released patches.

Jon Shafer says that they have been working on this since "June 2007". As a modder, and having input from the vary start, it better be a "finished" product.
 
The BTS ai was excellent compared to any other strategy games AI. I expect it will further improve in civ 5 - especially with the removal of BTS's complicated economic micromanagement that the AI didn't understand.
 
I could agree with that, I think most of the city management has become further simplified and streamlined, and specific things humans abused and the AI didn't get in civ4 (specialists and great people, and some wonders) aren't there so much.

However, there is no indication the AI in combat and warfare will be better, and several reasons to think it could be worse, but it is a wait-and-see thing. I suspect it has been made harder to engage in offensive warfare in general, so maybe humans can't entirely run over AI, but on the flip side that probably means the AI will never ever be able to threaten a human playing defensively in any way without tremendous advantages, suiciding its troops hopelessly.
 
Well the devs have stated that Ed Beach and his team put alot of effort into the Tactical AI, and the modular AI code in general. Yes likely some fine tuning will need to happen, I don't think a PC game has been invented that was perfect on day 0.

CS
 
To the original question as to whether the AI will be able to cope with the change to a stackless combat: the AI never handled stacked combat well. They would consistently send streams of small stacks, rather than build up one big stack to attack with. Their stacks were often odd - such as using horsemen and chariots to escort a stack of catapults.

Not sure if this changed later, but part of the problem with the Civ IV AI was that it flagged units as "attack" or "defend" and wouldn't use them any other way, even when doing so would be advantageous. It would seem that that would be less necessary, and easier to avoid, in a 1upt system.

In Civ V, much will depend on the interaction and authority of the strategic AI and the tactical AI. If the Strategic AI only has control over units that are not involved/assigned to a given front and the Tactical AI has control over all units that are, then the problem will be getting the Tact AI to pass units back to the Strat AI when they are no longer needed. Otherwise, one front may have too many units, while another is getting overrun.

If the Strat AI has the authority to pull units from a front, it will have to be able to evaluate each unit's value within its current tactical context: even if the "mission" of a given front is to hold the line, a cavalry unit in position to roll up the enemies flank or otherwise weaken the enemy's position, may be more valuable where it is, even if another flank needs more cavalry.
 
I think stackless combat will be easier for AI than stacked combat simply because there are less options to calculate. Stacked combat offered almost infinite options, so is difficult for computers. In civ 5 the tactical ai may even be able to 'brute force' a few turns ahead. Thyrwyn has a good point though, multiple fronts may be a good way to confuse it.
 
Top Bottom