AI not conquering each other kills this game for me

[Oberinspektor Derrick]

There was an interesting experiment conducted by Marbozir - he created a scenario where one AI had a giant death robot (and some uranium) from turn one. Did the AI use this huge advantage to roll up all the other civs? No, because not being sentient it had no way of thinking that it had a huge advantage. It just followed its estabished routines - build new settler - escort settler with strongest unit available - and so on.

The coding behind the AI's decision to DOW is completely off. It adds up the individual combat strength values of each army unit and summarizes them, and it will usually only DOW if it has a huge relative lead (3x as much for example).
In the case of a GDR, that's 130 army value, whereas deity ai standard starting bonus is 5x warriors at 20 combat strength each.
Add this together and the difference, from the ai perspective is 130 vs 100 army value, and so the difference is not big enough for the ai to DOW the other, even though there is no scenario ever where a GDR has to be afraid of warriors.
Even if there were 20 warriors (400 army value), they would get absolutely flattened by a GDR, but due to shoddy coding the ai doesnt understand this.
Good players can abuse this quite heavily to stay safe or trick the ai into declaring first.

 

[aieeegrunt]

I know it's a little late but the " no walls " mod helped me a little. now ai is better at taking cities.

Cries in Mobile/Console

Walls only slow down the AI, any competent Human player can go through them without trouble

The coding behind the AI's decision to DOW is completely off. It adds up the individual combat strength values of each army unit and summarizes them, and it will usually only DOW if it has a huge relative lead (3x as much for example).
In the case of a GDR, that's 130 army value, whereas deity ai standard starting bonus is 5x warriors at 20 combat strength each.
Add this together and the difference, from the ai perspective is 130 vs 100 army value, and so the difference is not big enough for the ai to DOW the other, even though there is no scenario ever where a GDR has to be afraid of warriors.
Even if there were 20 warriors (400 army value), they would get absolutely flattened by a GDR, but due to shoddy coding the ai doesnt understand this.
Good players can abuse this quite heavily to stay safe or trick the ai into declaring first.

How? How do you fail like High School Level math this hard?

 

[Oberinspektor Derrick]

How? How do you fail like High School Level math this hard?

Excuse me, what on earth is warranting you to speak to me like that?

Moderator Action: Post edited to remove inappropriate language. Please report posts if you think you are being trolled or flamed in future. leif

 

[aieeegrunt]

Excuse me, what on earth is warranting you to speak to me like that?

Sorry that was directed at the person who wrote the AI code not you

Moderator Action: Removed mod text. leif

 

[Linklite]

To be fair, the AI wasn't made with the concept of having a GDR while in the Ancient Era. Is that even possible without "cheating"?

Given the rather reasonable assumption that everyone will be in roughly the same part of the tech tree, it's actually (theoretically, I'm sure the poor AI in other areas let's it down) an alright system. It breaks down a little with advanced units, but they should only be +10 or so CS ahead, and that would trigger the need for extra units, and the numbers should allow you to run around the units. Your one powerful unit can only be one place after all.

Once you get to GDR v Warrior, yeah, it breaks down pretty bad. But then, should we expect the AI to cope with such absurd matchups when they shouldn't be occuring?

 

[UWHabs]

To be fair, the AI wasn't made with the concept of having a GDR while in the Ancient Era. Is that even possible without "cheating"?

Given the rather reasonable assumption that everyone will be in roughly the same part of the tech tree, it's actually (theoretically, I'm sure the poor AI in other areas let's it down) an alright system. It breaks down a little with advanced units, but they should only be +10 or so CS ahead, and that would trigger the need for extra units, and the numbers should allow you to run around the units. Your one powerful unit can only be one place after all.

Once you get to GDR v Warrior, yeah, it breaks down pretty bad. But then, should we expect the AI to cope with such absurd matchups when they shouldn't be occuring?

It can lead to some cases when they fail to upgrade units, and have like a million Heavy Chariots running around ready to suicide into a bunch of line infantry or tanks. Yeah, a GDR in the ancient era from World Builder isn't anything that I legit expect the AI to handle properly. Plus, obviously, just a linear addition doesn't make sense once you get to corps and armies. 2 Line Infantry have a lot higher combined strength than an Infantry Army, but the Army will slice through those line infantry like a soft cheese.

The other part the AI always fails to realize (and is perhaps because it's too easy anyways) is factoring in the production/gold/faith capacity of the human/target. So often I can have virtually no units, but if I'm banking 1k gold per turn and have a ton of production available, I can crank out an army really fast. Like in my current game I ended up getting "dragged" into a war because Norway declared against an Ally of mine. I only had a couple crappy units nearby to them, but just took the 4-5 closest cities and had them crank out units, plus a few purchases, and suddenly I have a couple infantry and tank armies at the ready, and they can't do anything about it since even their cities are barely denting my units.

 

[Abaxial]

The other part the AI always fails to realize (and is perhaps because it's too easy anyways) is factoring in the production/gold/faith capacity of the human/target. So often I can have virtually no units, but if I'm banking 1k gold per turn and have a ton of production available, I can crank out an army really fast.

Something I do regularly!

To be fair to Marbozir, the GDR experiment was something of a joke.

 

[Oberinspektor Derrick]

breaks down a little with advanced units, but they should only be +10 or so CS ahead, and that would trigger the need for extra units, and the numbers should allow you to run around the units. Your one powerful unit can only be one place after all.

Once you get to GDR v Warrior, yeah, it breaks down pretty bad. But then, should we expect the AI to cope with such absurd matchups when they shouldn't be occuring?

I disagree here.
Even a difference in + 10 CS (one tech level) is pretty significant, and if you dont play accordingly you can very easily lose that war.
At +10 you will take roughly about twice the damage that you deal, meaning you need (very roughly) about a twice as big army (assuming AI standard tactics of just suiciding) to stay even.
Worse still, just a single high tech unit on the enemy side will dramatically boost the CS of every defender city, and considering how much the ai struggles already with cities, that one unit can very well be the difference of the attacker taking a few cities, and suiciding himself to death (especially when we factor walls).
Hell, even for the player, attacking an AI that suddenly pops his first higher tech unit, suddenly makes the war that much more difficult.
Example: You warrior rush the AI on his 20 CS cities, and the city suddenly turns into a 30 bastion - 35 or more if the unit is also garrisoned there.

All in all, +10 or 15 doesnt sound like much, but they have very clear implications to both the success rate of the war, and the tactics that need to be employed.
As long as the AI ignores the relative strength of units, he will keep making dumb decisions.
And as we all know, the ai cant employ superior tactics like the human can, in order to circumvent that disadvantage.

 

[UWHabs]

I disagree here.
Even a difference in + 10 CS (one tech level) is pretty significant, and if you dont play accordingly you can very easily lose that war.
At +10 you will take roughly about twice the damage that you deal, meaning you need (very roughly) about a twice as big army (assuming AI standard tactics of just suiciding) to stay even.
Worse still, just a single high tech unit on the enemy side will dramatically boost the CS of every defender city, and considering how much the ai struggles already with cities, that one unit can very well be the difference of the attacker taking a few cities, and suiciding himself to death (especially when we factor walls).
Hell, even for the player, attacking an AI that suddenly pops his first higher tech unit, suddenly makes the war that much more difficult.
Example: You warrior rush the AI on his 20 CS cities, and the city suddenly turns into a 30 bastion - 35 or more if the unit is also garrisoned there.

All in all, +10 or 15 doesnt sound like much, but they have very clear implications to both the success rate of the war, and the tactics that need to be employed.
As long as the AI ignores the relative strength of units, he will keep making dumb decisions.
And as we all know, the ai cant employ superior tactics like the human can, in order to circumvent that disadvantage.

As a very rough estimate, it's not terrible. But yeah, the deeper you go, the more it breaks down. Especially once you start dealing with units that differ by 20+ points, suddenly the lower value units are basically worth nothing. ie. One tank vs 5 warriors, even everyone attacking the same turn, I will kill every unit and take maybe 6 points of damage. Even 5 warriors vs one Musketman, I could probably still survive all 5 attacks in the same turn.

And yeah, city strength definitely can play too. I know I had started a recent game where Canada was very nearby, so I had the bright idea to try a rush on them. While I captured their initial city without a lot of trouble with a couple archers plus a mix of warriors, Heavy Chariots, and an Eagle Warrior from a barb camp, when I got to their capital suddenly they got that first Swordsman, and then that was the end. Even bringing along a battering ram, suddenly none of my units could do enough damage to really make a dent in the city, and they took so much in return that they basically immediately died from a counter. I almost could survive by drawing their swordsman out to kill off one of my units, but because they were across a river I didn't have enough to kill him, and he could retreat back and heal, and in the end, I just quit the game as I was too far back.

I probably technically had a 150 combat strength edge to their 35 or 70 (when they got a second sword). But because of positioning, garrison and the exponential nature of war, there's nothing you can do. You see it the same when the AI pops out that first Crossbowman when you're attacking, and suddenly the crossbow plus city shot can start picking off your sieging units. I can have a huge edge in raw numbers but they're still dealing me more damage than I'm dealing them.

 

[Oberinspektor Derrick]

I probably technically had a 150 combat strength edge to their 35 or 70 (when they got a second sword). But because of positioning, garrison and the exponential nature of war, there's nothing you can do. You see it the same when the AI pops out that first Crossbowman when you're attacking, and suddenly the crossbow plus city shot can start picking off your sieging units. I can have a huge edge in raw numbers but they're still dealing me more damage than I'm dealing them.

This right here is why a relative unit strength difference is so huge, especially for the defender. Personally I think its flawed design, but that's beside the point. The main problem here is that the AI doesnt recognize these things beforehand, when it declares to DOW (or not).
I mean, as a human player it can be hard to tell as well whether you should declare or not, because the game might show 200 vs 200 army score, claiming that its even.
But, as a human you can tell roughly what kinds of units you will be facing, by looking at the combat strength of their cities.
Are you storming warriors against cities with 35+ or more combat strength?
Expect a part (or all) of that ai 200 army value to consist of swordsmen/horsemen or better, so you might wanna think twice before that DOW.
The AI should at the very least take tech levels into account by looking at city CS, and generally be more aggressive in cases where the enemy has an equal or lower military tech level.
That would by no means completely solve AI behaviour, but be one of many small steps the developers could take to improve AI behaviour.

 

[Oberinspektor Derrick]

Sorry that was directed at the person who wrote the AI code not you

Moderator Action: Removed mod text. leif

My bad there, and I hope you didnt read my initial heated reply.

 

[aieeegrunt]

My bad there, and I hope you didnt read my initial heated reply.

lol no worries, misunderstandings happen, especially in this medium

The way combat works in Civ the relative strengths of units are far more important than their absolute total

It’s fracking astounding if the way Civ AI works is by merely adding up totals

One thirty strength unit is not the equivalant of three ten strength units, its probably more like 6

Whoever came up with that has no idea how either math or combat in Civ works

 

[reddishrecue]

You can make armies and larger density units in civ 6 which makes them more powerful. You still have to be careful for more advanced units that can still abuse the primitive corps or armies though.

 

[iammaxhailme]

I've been playing civ 5 a lot the last few months and I recently went back to 6, and once again, I'm reminded how rarely the AI actually dominates others in 6 after the beginning phases of the game. In 5, on emperor difficulty, I regularly see an AI wipe out another one in like, the industrial or modern era. I don't think I've ever seen that in 6 on emperor except maybe with Eleanor.

 

[Bleys]

Here's the thing, in some earlier versions of Civ, the AI would conquer other AIs, but the resulting double-sized AI often ended up becoming a super-powerhouse. In fact, in 3 or 4, it was fairly important to try to keep the other AIs at peace with each other for fear one would conquer another and end up with a massive amount of land and cities.

Im sure that has played a part in how they decided to code this.

 

[Oberinspektor Derrick]

Here's the thing, in some earlier versions of Civ, the AI would conquer other AIs, but the resulting double-sized AI often ended up becoming a super-powerhouse. In fact, in 3 or 4, it was fairly important to try to keep the other AIs at peace with each other for fear one would conquer another and end up with a massive amount of land and cities.

Im sure that has played a part in how they decided to code this.

I hope not, because civ 5 had this imo perfectly balanced. Most civs didnt snowball out of control, but you'd still in a majority of cases see civs late game becoming a regional powerhouse and thus a threat, especially when left to their own on a different continent.
I liked that part, discovering the other continent and meeting both minor states and the one or two local hegemons (even some killed off).

 

[reddishrecue]

I hope not, because civ 5 had this imo perfectly balanced. Most civs didnt snowball out of control, but you'd still in a majority of cases see civs late game becoming a regional powerhouse and thus a threat, especially when left to their own on a different continent.
I liked that part, discovering the other continent and meeting both minor states and the one or two local hegemons (even some killed off).

Yeah that does happen a lot in civ 5. I recall seeing a powerhouse in another continent with 2 or 3 other capitals of civilizations that didn't exist anymore since the powerhouse wiped the other civilizations out and kept the capitals. You can tell that when you try to dominate the other continent and see the other capitals that you have to capture as well.

 

[Naokaukodem]

Civ5 could really make AIs monsters, just not to be attacked by them was requiring some fair amount of AI experience. I recall a game by Marbozir where he struggled to take Mongol's capital, he succedeed but barely, the capital being enclaved in Mongols' territory. (would have rebel in 6) Then the Mongols grew out of control militarily, I said myself "now how will Marbz deal with that" (Mongols had modern tanks and GDRs all over the place) and Marbozir was just like "it's fine, I don't think they will attack me" And they didn't !!! I still wonder why.

 

[Pistol90]

AI were able to conquer each other in previous Civs because of walls not being super-weapon like in Civ6.

 

[James_Champagne]

Interestingly enough, the game I'm playing now, the AI Mongols wiped out the AI Aztecs early on. One factor I think is that I have a mod on where capitals are always loyal, so if the AI wants a capital, it actually has to go and try to fight for it (rather than doing the out-loyalty-it-out waiting game).