Simple Ways the AI Could Be Better

[PhilBowles]

Hey all, I'm new to CF and this is my first actual post; been a major civ fan since Rev. I seem to be one of the "fortunate ones" in regards to AI conducting wars. I have been on several occasions since G&K came out, sucked into a flank ambush. Mongols were a tech ahead of me, I was Aus and lacked iron. I went for the only source, they hit hard ruined it and when I sent my units in he pulled back and hit hard on the other side of my empire.

One thing I will say that needs fixing is Pikemen. It seems when there is no iron the Pikes are your best counter, which I think its unfair that they can go toe to toe with LSmen. Just a thought but maybe find a way to balance that out?

Pikes are a lot weaker than longswords, much moreso than pikes are stronger than swords.

My only concern with pikemen is that now they're so strong as well as the 'go-to' unit for most AIs, their bonus against mounted units should be reduced, since mounted troops become very weak for most of the medieval era as a result. This hits knights particularly, since pikes tend to be viable for most of the knights' lifespan, and the transition to muskets finishes knights for good.

 

[nokmirt]

I agree, that is why I had noted "developers".
I would like to think that the promised source code etc will be released but I can't help but think with the Steam method of distribution and the way that the game loads we will not see that happen and thus be stuck with a flawed AI.

I agree that Steam is a real issue lately. I've already had an issue where a mod updated automatically, causing me not to be able to load my save game. From now on any mod I download gets backed up. Much of the time I stay offline, when playing.

I hope we do not get stuck with a flawed AI. The best hope I have is that the devs read our complaints and try to fix things. They could at least make the AI a little better each time. It will never be perfect. And I really do not expect it to be so. Just having it have a bit more common sense, and prioritize threats better would be enough for me.

 

[VimesUK]

Releasing the source code could well do more for the longevity and improvement of this game than a succession of more patches, with the skill and generosity of the modders in this community. However I can't help but think that it will not be done.

 

[dennisdc]

If the AI would simply target siege>ranged>melee units when firing from a city either with bombardment or a ranged units, it would be a lot harder to take over enemy cities.
Often they will target a close melee unit which you can then keep well fortified and healed, while the rest of your army pounds on the city.

 

[Avatan]

I've had Marie Therese attack me with spaceship parts and she was on another continent too!
I repelled her puny invasion and was very disappointed to survive as I thought all was lost. she had 3 times my score and a huge army (I was OCCing, immortal).

But I've seen the AI put up some decent defensive fight, and I sometime lose a city or two settling too aggressively. It's way better since G&K though. I especially enjoy the surprise trireme rush on archipelago maps! It drives me to systematically build an early navy which I never did in almost 15 years of playing the franchise.

 

[Avatan]

If the AI would simply target siege>ranged>melee units when firing from a city either with bombardment or a ranged units, it would be a lot harder to take over enemy cities.
Often they will target a close melee unit which you can then keep well fortified and healed, while the rest of your army pounds on the city.

Having all melee units destroyed can be annoying and slow an attack down, but it is not an excellent defensive strategy because you can always have a mounted unit waiting outside of fire range to take the city.
But I think that's what the AI is going after, preventing you to take the city with melee. It makes some sense because that's pretty much how the AI take a city, by suiciding melee against it.

 

[krc]

If the AI would simply target siege>ranged>melee units when firing from a city either with bombardment or a ranged units, it would be a lot harder to take over enemy cities.

The problem with this idea is that it probably works for the AI against humans, but not against other AI. (And so would be more complicated to program.) When my cities are being attacked, I often find it useful to eliminate the AI's melee units first. Then they cannot capture the city, and tend to rush their siege and ranged units right up adjacent to my city, where I can eliminate them as well.

 

[Xyuta]

Simple thing to improve the AI :

Peace treaty should be asked at the end of turn.

 

[Splify60]

Watch Sidon's Youtube videos for how to destroy AI units on Deity. I think there are many simple changes to the AI that could be made to improve the difficulty of combat. Here are a few:

Do you have a link for this? I have been unable to find it on youtube... Sounds really interesting!

 

[dennisdc]

Having all melee units destroyed can be annoying and slow an attack down, but it is not an excellent defensive strategy because you can always have a mounted unit waiting outside of fire range to take the city.
But I think that's what the AI is going after, preventing you to take the city with melee. It makes some sense because that's pretty much how the AI take a city, by suiciding melee against it.

The point of shooting the siege and ranged units first is that the melee will take damage while attacking the city while the ranged and siege units will not, therefore making their attack constant powerful while the attack of the melee units will diminish in strength quickly.
Also I do not see the hurtful aspect of targetting the siege units first. They are after all the units making a city attack a lot easier and therefore taking them out first thing can cripple the entire attack while taking out a melee unit will at most prolong the attack while the siege units are bombarding.
Regarding the pure ranged their attack will not be as hurtful as the melee units, so I can see how taking out melee units might be a good idea, but surely the siege units must be number one priority.

 

[smallfish]

1) Prioritize ranged over melee military techs. Get crossbowman, cannons, artillery before infantry.

This depends on their mood. Sometimes, AIs like Gustavus or Catherine will tend to go for Artillery before anything else.

Other than that, I do agree on them needing a bit more of a spine at times. Especially in the case of team games, where your team mate will declare peace at inopportune moments.

 

[poopfeast420]

Have a limit as to how much time your getting a diplomacy penalty from lying to a civ about whether or not you're going to declare war on them with your units outside their borders. It's really annoying when thousands of years later I can't be friends with a civ on the other side of the world because I attacked some civilization they never met early on in the game.

 

[God of Kings]

Do you have a link for this? I have been unable to find it on youtube... Sounds really interesting!

His YouTube channel is sidor1982, not sidon. That is why you cannot find it.

 

[Splify60]

That explains it, thanks so much!

 

[omnishakira]

The question is ; Do you think we could train a monkey to play his units better than the AI ?

 

[clif9710]

I think the ultimate AI would be one that learns your style of play and then plays future games against you with that knowledge. We all have favorite techniques that work pretty reliably. An AI that learned your modus operandi would be very hard to beat.

Think of that scene from Patton where George C. Scott yells at Rommel, "You magnificent son-of-a-, I read your book!"

 

[Minor Annoyance]

This thread seems to be mostly about unit AI and now less about specific suggestions, but mine is about the AI and religion.
Most of my games so far my religion end up being the biggest, because the AI doesn't often do a great job of spreading it. I think the AI should prioritize taking the beliefs that help the passive spread of religion. This isn't really improving the AI, but while it can't compete effectively it should at least compensate in the easiest way possible.