Why AI combat will never work in Civ 5

[PanzerGeneralli]

Straight off, I have to say there is nothing inherently wrong in principle IMO in having one unit per hex and having a paper rock scissors combat system. In multiplayer it might turn out to be excellent.

However, in basing Civ 5 combat on Panzer General, history was ignored. The AI in Panzer General and all the other titles that followed it was completely unable to handle offensive operations. In defence it didn't have to do much, the player mostly had to figure out how to solve the puzzles that the map setup involved.

In a dynamic game like Civ 5, there is no way that the AI will be able to handle paper rock scissors - unless Firaxis magically cracks the AI code that eluded the General series. So the AI is crippled offensively. That means it will lose lose lose even if it spams units.

The standard CIV stacking system avoided this problem because defensive and offensive units would mix, not giving a smart human player the opportunity to take apart the AI.

Hopefully I'll be proved wrong.

 

[Lord Monkey]

The combat system in Civ V wasn't based on PG, it was inspired by it. (At least, that's my understanding of the developer quote I read.. somewhere.) So I'm kind of lost as to how Firaxis ignored history.

Also, I don't think the combat is quite as simplistic as rock-paper-scissors. There's terrain type, bottlenecking, ranged attacks.. There's a lot of things you can/are forced to do now that stacks are gone. Overall I think it's a definite improvement in combat, even if the AI needs a little more teaching before it's ready for a huge base of veteran fans, most well-versed in ripping the poor computer apart.

Side note, consider the increase in processing ability of personal computers since PG (1994). While this doesn't necessarily translate to smarter AIs, it does allow for more complicated and numerous algorithms to run without choking your computer to death.

 

[boredatwork]

In fairness the fact that civilization is a dynamic game is to some extent compensated for by the fact the new combat system is still very basic, even compared to a simple wargame like PG - it's not like it has to worry about unit variety, or defensive fire, or suppression, or logistics, or 9 steps of entrenchment, or transport, or weather, or order of fire/movement, etc.


But yes, I would not place money in any major improvements - though I disagree that unit spam couldn't provide a challenge, as long as we're not talking 1 hex bottlenecks.

 

[rschissler]

That's a good point about the defensive and offensive units mixed in a stack. Obviously, you can't do that in Civ V. I know the game designers wanted to get rid of the SOD, but maybe they forgot how much the AI from Civ IV used it to their advantage as well.

 

[Fistalis]

The big issue is.. The AI can and did overcome its tactical stupidity with SOD. Now.. the AI has no such crutch to fall on so battles are fought purely based on tactical prowess.. of which the AI has none. I'm all for removing SOD.. problem is they didn't spend enough time invested into the tactical AI to overcome the AI losing its most useful tool. How much time would have had to be invested? I don't know.. but that should have been the first question asked before removing the SOD dynamic imho. I mean Ideas are great.. but it all comes down to implementation and it seems to me they didn't put near as much thought into that as they should have. It seems that rather than looking at the implications of design choices they just said "This is what were doing and to hades with the consequences". Now they have released a game and the question is is there any remedy thats realistic. The answer is probably no. Yes the AI will probably be improved but I imagine the majority of devs are going to be put where the money is.. thats in expansion packs and DLCS because after all it really boils down to a business.

 

[Lord Monkey]

I think there should be an option to crowd units into a single space, but at a detriment to their defense/attack/etc. Or maybe make them all take damage from ranged bombardment. Or set a hard limit on the number of units in a space (3?).

 

[KAuss]

Ive played a lot of games that involed 1upt. Most of which were even chinese titles. One big one would be the romance of the three kingdom series. Despite them having static maps, they work well on offense and defense. It will never be as good as another human but it provided a challenge and even some surprises at times.

I wouldn't go as far to say the game won't ever work right, it would just take a lot of work.

 

[gonzo562]

The combat ai is not as bad as i thought. My current game im playing
immortal difficulty.
inland sea
large map
reduced to 10 city states

Every other difficulty below immortal has been easy. and lead in all demographics and my mech infantry was fighting like riflemen and lower. But in immortal difficulty im not leading in all demographics

I just crushed the french army invading my territory about 7 units. After that I went and tried to counter attack and invade them. They had quite a few units to defend against me
Also I hope they keep it at 1UPT

Here is a screen shot of the French Defense after i Crushed their invading forces. After all saw this i decided not to go the offense

 

[ond_magiker]

Ironically, if they do improve the AI, the already long wait in between turns will probably take even longer.

 

[Earthling]

Hopefully I'll be proved wrong.

The game's already out, and you were right.

But a large majority of people (90% +) insisted right up through release and some afterwards that the AI was guaranteed to excel - some even took weird positions of claiming that the AI was so good/so hard to beat/*already* played like a human in games like the Panzer General series you mention, or the Battle for Wesnoth or something.

Realistically of course you're right, tactical AIs are extremely difficult to create, the developers (should have, at least) knew this, and the chances it will be completely overhauled now are pretty low, though some minor improvements probably occur over time.

 

[BjoernLars]

Computers will never be able to compete with the human mind in reasoning skills in the near future. I do think that the AI script can be improved over time to make it more of a challenge.

 

[Cyan Leader]

I play a lot of Advance Wars, and the AI works fine with 1upt there. It might be easier for it if it's squared based though.

 

[gonzo562]

In a dynamic game like Civ 5, there is no way that the AI will be able to handle paper rock scissors - unless Firaxis magically cracks the AI code that eluded the General series. So the AI is crippled offensively. That means it will lose lose lose even if it spams units.

Hopefully I'll be proved wrong.

France is spamming units since forever. I have been fighting the french to a stalemate and America Minute men special unit blows.


Naval invasion first time i seen it in any of my games. They sent 3 musket men and longswords man. I doubt they will take my capital but if they AI actually uses their ship of the lines to bomb Washington i could lose it but i doubt it will happen


To me The combat AI is not as bad as everyone make it seem to be. Levels king and lower was really easy. But immortal and deity I don't think its that easy

or maybe i just suck

 

[ElephantRider]

I have seen the AI put up a good fight and place units nicely on the battlefield. I have also played 1upt games from the NES days that had good AI as well.

I'm starting to think that maybe some of "oh the AI is sooo bad at war" comments come from people who go to war against someone like Ghandi. Ghandi was not a war leader. It only makes sense that he would not know how to place units on a battlefield. Obviously from the above screenshots, France puts up a pretty good defense. As they should, Napoleon was a great general. Perhaps, the developers are trying to do this with the AI and make an AI civilization only as good at war as leader should be.

To be honest, I've only played through about 1 1/2 games. The first full game, no one ever made any units to fight me. None of the other leaders in it were ever known for being a military leader. The game I'm currently in is showing some decent AI fighting ability from leaders that you would think would have it.

 

[aimlessgun]

I have seen the AI put up a good fight and place units nicely on the battlefield. I have also played 1upt games from the NES days that had good AI as well.

I'm starting to think that maybe some of "oh the AI is sooo bad at war" comments come from people who go to war against someone like Ghandi. Ghandi was not a war leader. It only makes sense that he would not know how to place units on a battlefield. Obviously from the above screenshots, France puts up a pretty good defense. As they should, Napoleon was a great general. Perhaps, the developers are trying to do this with the AI and make an AI civilization only as good at war as leader should be.

Nah I'm pretty sure the AI is actually horrible. They are easily baited, easily outmanuevered. They will suicide into set defenses all day every day.

The AI needs some chess-program type methodology that allows it to simulate at least a couple turns into the future. Simulation of future events is a key part of the human advantage, something we do without even thinking about it. Simulation of future events is how Deep Blue beat chessmasters.

The AI will never be remotely competitive without the ability to simulate a battle a a couple turns into the future and determine what will work and what will not. A computer needs to be able to realize that attacking my defensive stand will result in death of all their units, and death to none of mine.

 

[Biz_]

civ5's minimum system requirements are more than enough to provide a good AI for the situation

it's just a matter of putting resources into developing it instead of just letting 5 percent of the team work on it

a 20 year old strategy game that needed the only CPU available to do all the other in-game stuff isn't exactly the best example of what is possible with AI

 

[Mîtiu Ioan]

It will be hard to balance a "hex-combat AI" - in Panzer General series there was some limited number of predefined maps for each could be established from start a troops arrangements, predefined defensive points and so on. Also the fact that the human player should complete a level in a certain number of turns ( quite limited ! ) was another pressure factor ...

Main problem in Civ5 is the fact that maps are randomly generated everytime and strategical/tactical level are too mixed.

Best regards

 

[slowjames]

So, CiV's AI will never be competent because some other game with the same format had different gameplay goal? Sounds like the opinion of an expert.

 

[HeavyTwenty]

Maybe a 2 unit stack would work better? One melee and one ranged? I guess it would be unbalanced to have an siege unit defended by a high shield melee.

 

[falconne]

I'm starting to think that maybe some of "oh the AI is sooo bad at war" comments come from people who go to war against someone like Ghandi. Ghandi was not a war leader. It only makes sense that he would not know how to place units on a battlefield. Obviously from the above screenshots, France puts up a pretty good defense. As they should, Napoleon was a great general. Perhaps, the developers are trying to do this with the AI and make an AI civilization only as good at war as leader should be.

I'm pretty sure that, like with Civ 4, the actual AI code that's used is the same for all Civs. There is no handicaps making Gandhi less capable in tactical battles than anyone else, they will all be the same. Every one of them is a pushover during battles no matter what difficulty you play.

Even the different flavours of play that leaders had in Civ 4 seem to have been toned down (or at least being discarded during the game) because all leaders play the same.