The AI would be dangerous if ...

[Code Monkey]

... it spent more time attacking cities instead of roads.

Something seriously needs to be done about this fascination with destroying roads when no strategic purpose is served. It cause the AI to spread units out too much, waste resources, and generally be ineffective until I launch a coordinated attack and take out cities.

In a game I just won (first Monarch and first Domination ) me and the Aztecs were at war with Greece. I watched as Montezuma coordinated sending carriers loaded with bombers after roads that weren't of any importance whatsoever. I watched as he would spend several turns to send infantry out to destroy a road that, again, was of no importance whatsoever. Meanwhile, Grecian cities were only a few squares away - had he bothered to send those same units he was using to hamper travel for non-existent units against the weakly defended cities, the war would have been over a lot sooner.

It doesn't do much good to have the AI send hordes of units out if it's weighting algorithms make a non-important road into a more desirable target than a city. Taking cities wins wars and eventually games. Taking roads that aren't part of a crucial supply network does nothing.

The problem seems to lie in the AI using a one strat fits all approach to warfare. If your purpose isn't one of all out conquering, it can make sense to take out roads because you can send cities into unrest and cut off reinforcements from the larger empire. I've used this strategy when all I'm trying to do is hurt the AI enough to get them to sue for peace. But, when you hold the advantage and you're planning on conquering everything in sight, it's counterproductive to destroy roadways.

Between the AI's fascination with roads and workers it's a wonder it ever gets anything done during warfare.

[Zealot]

The AI doesn't conquer cities because it already knows that it's pointless to manage a city with such high corruption levels.

[jgflg]

I think the inability of the AI to use artillery offensively also seriously degrades the challenge.

[Golgo13]

I am also not very impressed with the war-time AI. One, they never attempt to use any kind of combined arms attack. I also notice that while you may have several armies destroying or capturing their cities, they send their units past their cities, past your armies and attempt to attack either your cities or your infrastructure. You end up hunting down huge stacks of often inferior troops (because they always prefer to build those cheap units) instead of actually worrying about a counterattack against your armies or a tough defense of their cities. I often feel like I'm fighting a retard and not a computer. I find it disappointing that the AI has such poor tactics given the simplicity of the combat system.

[Unregistered]

lol!

AI is a really bad choice of words. I've seen the computer opponent do some fairly 'intelligent' things as far as diplomacy goes, but their military tactics leave a lot to be desired. I have yet to have one of my cities even bombed let alone attacked.

"The computer opponent would perhaps be dangerous if it ever actually attacked one of my cities" is right on indeed.

Curiously enuff, it does sometimes, occasionally, attack other computer player cities & take them. Recently in my current game this has gotten fairly rare tho. China, America, Japan & the Aztecs all share a continent in my current game. In roughly 50 turns of free-for-all warfare between them (monarch, modern era, huge map, China, America & Japan all have about 50 cities, the Aztecs 25 - 30) perhaps 5 cities have changed hands between the four of them that I can tell.

Perhaps it is a modern era problem, because the computer players did manage to eliminate 6 of the 16 civs before we started getting into the modern age & they did it with no help from me. (none of the eliminated civs were in my neighborhood)

They certainly could use some lessons.

Or we could use the mp ability to try on real opponents. Even an email option would be good. Of couse, that will all be in civnet III (tm).

[barefootbadass]

Originally posted by Unregistered

Curiously enuff, it does sometimes, occasionally, attack other computer player cities & take them. Recently in my current game this has gotten fairly rare tho. China, America, Japan & the Aztecs all share a continent in my current game. In roughly 50 turns of free-for-all warfare between them (monarch, modern era, huge map, China, America & Japan all have about 50 cities, the Aztecs 25 - 30) perhaps 5 cities have changed hands between the four of them that I can tell.

It does attack and take cities, but usually only small cities with only a few units(for me). In my current game about 3 or 4 of them are ganged up against the Americans and they captured about about 10 of their cities in one turn, I thought the americans were out, but they must have had one defect to them somewhere.

[VetLegion]

this is very important to topic. please dont turn it into "I like - I dont like AI" thread.

we should point out to design weaknesses that are not really bugs, but are bad for AI. the more we point out the more there is a chance they will fix. after the last patch there is no chance of fixing anything.

for example:

AI flanking with transport is often not a good thing. I suggest algorithm is made to test: how many defenders player has and can built in target city. how many cities does he have and how much is he damaged in percentage of pop or production when/if AI captures and raizes his city. Only do flanking if it is secure to suceed and do much damage. In any other case dont.

AI building wonders in war is bad thing. I suggest AI counts the number of units it is attacked with - if total war then cease all build and only do military. if it is a small handleable number, then continue. This is to prevent player from feigning attack just to stop AI wonder building but also give AI chance to defend if it comes to life or death.

[Zanzin]

I've found the AI can be very dangerous in the ancient and middle ages - basically, when it's stronger than you are. I've been beaten twice on smaller maps where the AI dominates me early.

Later in the game though, industrial and modern ages, where combined arms become really useful, the AI simply cannot win, n account of the fact it can't use combined arms (say cavalry and artillery) together effectively. The cavalry will never wait for the artillery (in terms of moves). Maybe stacked movement would have changed all this. Long and the short of it is, the human player is too well set up for an AI invasion to be possible.

And, when I invade them with 60 plus troops, instead of using their attacking troops to fend off mine, they send 50+ cavalry to my borders, who then get completely owned by my mobile defense force of artilery and infantry! Why doesn't the AI attack my invading force, destroying the threat?

There is some serious logic problems with the AI, let me tell you!

[notyoueither]

Originally posted by VetLegion
this is very important to topic. please dont turn it into "I like - I dont like AI" thread.

Good idea.

I've seen the AI be vicious b*stards when they have someone down. In my experience it is very common for them to go after a civ till that target civ is gone.

I think the problem is in the decision making process of when this blood lust is triggered. It usually happens when 3 or 4 civs war on a single civ, but not always. I have seen it when 1 civ went after 1 other and wiped it out (large India went genocidal against small Aztecs). If the balance of forces is the trigger, I am sure it is something that Soren would be agonizing over. Set it too low and AI civs will be too ready to seek Pyrrhic victories. Set it too high (as it prolly is now) and the behavior would not be triggered in many cases where it could win them the game.

Also, the trigger would be variable with difficulty level. Can't have AI sharpening carving knives for the player on Chieftain level.

I think it is obvious that where the game needs the most improvement is in the late game. Coincidentally, the late game is the hardest to get right (most variables). It might have been a conscious decision of Firaxis to sell the game, then work with player feedback to improve the late game (actually, I think it's kind of obvious). I don't mind, I would gladly have given them payment for the game in advance to become a play-tester. On the other hand I understand some people being upset about this state of affairs. I just wish they wouldn't be so vehement until Firaxis says they're finished, ie *no more patches*.

A suggestion for game designers. It might be possible to sell a beta to the public. Many fans would buy it and happily contribute to further development gratis. Others wouldn't, they would wait for the finished product. It would avoid the situation Firaxis finds itself in on these boards. It's kind of obvious that this game wasn't done when it went *gold*.

Salve

[Redstar]

What many people say here is true.
But i also have found the AI will come right up to your city and then just stop until next turn.

You then have the opportunity to bring in every thing you have
(since there is 0 movement cost --unless you have a water barrier).

You can wipe out their entire offense because you can bring the max. fire power effortlessly combined with them allowing you to strike 1st.

Only in one situation have i actually seen the AI turn this around and use all its units.
I landed 8 units from sea. On their turn they sent in 24+ units (or as many as needed to wipe out tanks/mechs).

My point here is that i would hate to see this in multiplayer! The defender has a HUGE advantage with respect to its 0 movement cost (modern age). The only thing working against the defender is the patience needed to move one's entire continental army.

(this might be why the AI does bomb the roads 1st

[jgflg]

Originally posted by Zanzin
I've found the AI can be very dangerous in the ancient and middle ages - basically, when it's stronger than you are. I've been beaten twice on smaller maps where the AI dominates me early.

Later in the game though, industrial and modern ages, where combined arms become really useful, the AI simply cannot win, n account of the fact it can't use combined arms (say cavalry and artillery) together effectively. The cavalry will never wait for the artillery (in terms of moves). Maybe stacked movement would have changed all this. Long and the short of it is, the human player is too well set up for an AI invasion to be possible.

And, when I invade them with 60 plus troops, instead of using their attacking troops to fend off mine, they send 50+ cavalry to my borders, who then get completely owned by my mobile defense force of artilery and infantry! Why doesn't the AI attack my invading force, destroying the threat?

There is some serious logic problems with the AI, let me tell you!

Your mention of 50+ cavalry is a good point . I'd be willing to bet that if they forced the computer to constantly upgrade units or stop building obsolete ones, you'd automatically face a much stronger challenge. I almost always check the AI's army and see that they have at least 15-30 cavalry, 15-30 swordsmen, 10-15 longbowman and so on (standard map) well into the Modern Age. And they keep churning out those units. Imagine if those were tanks and mech. infantry instead.

I've also noticed that the computer will counterattack with a ton of units if you take a few of its cities, but unfortunately it's always with cavalry never tanks. I'd lose a lot more conquered cities to counteroffensives if the AI would simply use 8-10 tanks instead of 8-10 cavalry against my 2-3 mech infantry.

So, in summary, I think it would be a good idea to either make obsolete units (from a previous era) disappear from the queue list or else force the computer to put a priority on upgrading units over almost anything else.

[Badtz Maru]

The AI tends to prefer using fast-striking units against weakened units in the field over attacking cities. This is kinda smart, but I think this preference is overdone somewhat. Slower units with good offenses tend to get used against cities more - in Ancient times I frequently see the AI stacking up it's Swordsmen, Legions, or Immortals against my cities while it's Horsemen sneak around to hit my reinforcements on their way to the besieged city. The problem is that in the later game there are no good slow offensive units, which I believe slows down the AI in regards to attacking cities. I modified the game to give Riflemen and Infantry better offenses and flagged that the AI use them both offensively and defensively, and the AI now frequently attacks my cities with them.

[Green Giant]

My biggest problem isnt really with the AI invasions, its with the AI's ability to defend. It simply cannot put together a coordinated defense. Get a second civ to ally against an enemy and it doesnt seem to know what to do. The AI needs to know when to go on the offensive and when to sit back and hold territory

Half the problem is the lack of modernized units. The AI just won't spend money on upgrading units. The worst part is if I have 20 tanks, and another civ has 40 knights, it calculates that it has a stronger military. I don't know where it spends all its money because the AI is always broke.


The initial invasions are impressive though, I like that. Its just that after the first wave is stopped theres nothing left.

Overall though I am having a good time, its just a lot of little things.

[Kc7mxo]

The problem isn't just that the AI won't spend money on upgrading units, its that it can't. In all my games the AI has been DREADFULLY poor. I'll pay the persians 500 bucks for a tech, and two turns down hte line they're below 50 again. They have no concept of a "nest egg."

Which is an interesting change from civ2 where the ai would sometimes rack up tens of thousands of bucks if you let him.

[Arrian]

The AI definitely has some weaknesses. I don't know how feasible it would be to "fix" these, but hopefully Firaxis can work on it.

The more complex warfare gets, the worse the AI is, as it does not use the units various capabilities as well as a human.

1) The AI does not defend its borders. It does not fortify good defensive units on good defensive terrain near the border. Thus, invading is easy - just shoot straight for those unoccupied mountains and hills. Any counterattack the AI may mount will be blunted by the fact that your troops are on mountains. Also, the AI's "patrolling" of its attack units hurts, because your invading army will often catch a number of those units out in the open and kill them on turn 1 of the war.
2) The AI doesn't upgrade its units, or disband old ones that cannot be upgraded (frigates, ironclads, cavalry, swordsmen, etc.)
3) The AI is ok on the attack, particularly early in the game. On the attack, it seems to understand the value of concentrated force. On defense, however, the AI will rarely, if ever, hit your invasion stack(s) with all they've got.
4) The AI still can't mount a serious invasion across water. When it does land troops, they last no more than 1 turn.
5) The obsession with capturing workers and hitting a damaged, obselete unit has got to end. Instead of concentrating their efforts on units that are right next to their city, and have the power to take that city, the AI will charge out and capture workers.
6) I totally agree about the AI wasting time and effort on breaking roads. Unless by breaking the road, the AI denies a resource to its opponent, or cuts reinforcements (in the rare case of a road that is only 1 square wide) to its target city. Most of the time, it just seems like the AI blasts every terrain improvement in sight as a matter of course.

Overall, I like the AI, but I think it could be better. Here's hoping that Firaxis is able to pull that off. I also agree that implementing fixes for these weaknesses should be restricted to the higher levels of play, or made an option in the startup menu.

-Arrian

[Golgo13]

Yes, the AI does not value upgrading its units. I think this is mainly because it doesn't take into account unit quality. Regarding the poor defense tactics...I belief this is largely due to the combat system. It seems that it is almost always better to be on the offensive than the defensive. However, I notice that the AI will avoid large stacks of troops even if they are closing in on their weakly guarded cities. They build attack units almost exclusively but then are afraid to use them against your armies even as they are about to lose a city. The AI needs to recognize when its territory is threatened and should respond by attacking your troops even if they are outnumbered.

[Redstar]

but it does develop modern units.

a wee disclaimer first..i have only played a few games all the way into the modern era. that said, in my experience the AI does a good job at upgrading/building the right units.

for example by 2030:

eg..
the english have...

9) settlers
22) workers
15) Pikeman
1)Longbowmen
10)knights
1) rifleman
21)inf

55) tanks
79) mechs
21) modern tanks
8) cruise missles

a stack of sea units and a handful of bombers

2 armies.

the germans have...

5) settlers
25) workers
1) warrior
4) spearman
1) swordsman
8) longbow
2)riflemen
1)cav
3)inf

68) mech
27) armor
13) artillary
31) panzers
18) bombers / 9) jet fighters / 3) helios
1 army


each of these empires is slightly more than 1/4 my size. They only produce modern units now.

to compare what i have at this point then...

2 warriors
1 horse
1 musket
7 knight
4 rifle
15 cav
16 inf

21 tank
78 mech
228 modern tank
5 cruise
4 ICBM

9 cannon /97 artillary

18 bombers/22 stealth bombers and some fighters
minimal naval units.

SO they are producing almost exactly like me. Like me, they must find it tedious to track down the obsolete units and upgrade them. But over time, the upgrade processes approaches completion.

In all modern conflicts i have had so far...they develop the right units..they just do it slower.

[G.A]

AI will build modern units in one case:

It previously had a huge war, and lost most of his units of certain type, and at the same time, it has a required resource. Then, as it`ll start building units of that type, and these will be modern ones. E.g, if it lost most of it`s cavalry, and has oil, it`ll build a lot of tanks, cause it needs mobile unit, and tank is the only option.

As for upgrading, it really can`t do that, cause it never has enough money, but... In a couple of games, when we reached modern age on scientific parity (I wasn`t the strongest ), I have seen the AIs accumulating the fairly large (2,5 thousands the most) sums of money. It still wouldn`t upgrade units.

[Jodo Kast]

The AI is sufficient on Diety difficulty, but it cheats!

[Redstar]

does not mean they are broke.
Often, all of their income will be tied up in trades or something fishy.

I say this because i routinely make deals with ai's that have no money and yet can still offer me 240 o a turn on top of their other deals.

ya, i think it sounds fishy too. AI cheater. hehe

[Xentropy]

I think unit prices should be raised as the game goes from age to age.. This opinion comes from two facts:

1) The AI appears to place the total value of a nation or city's units in the total shield value (not attack value or anything else) of those units, and

2) With no stacked movement, there are WAY too many modern units running around, making late game tedious.

My change should have the following effects:

1) The AI will build more modern units because the AI's idea of their value will go up with their shield cost,

2) The AI will more accurately estimate varying power levels, and

3) There will be less OF the modern units running around, since at a higher cost, even the high output cities will be able to produce modern units at approximately the same speed as older age units were produced, rather than faster, helping to stave off late game tedium.

This is just an idea, not playtested at all. Has anyone tried making a mod like this? No other changes, no more attack rating on modern units or anything, JUST higher costs as the game progresses.

It may work, it may not, would require playtesting :>

[Redstar]

because with empires of 150+ cities that each require 4 units to make people more happy (depend on gov), you can just imagine the overhead needed before one even think about attacking.

Besides the ai is perfectly aware of the unit value (how could it not be --its such a basic thing). The ai just needs to speed up its production depending on the level of difficulty.

[cassembler]

The very firt line of text in this thread, as posted by the Code Monkey-meister:

[The AI would be more dangerous if]... it spent more time attacking cities instead of roads.

I fight wars all the time that I have little to no interrest in AI cities. I'll send in troops for the sole purpose of pillaging and probably dying in the process...

Warfare doesn't have to be about expansion. It could be a brief tango over a strategic resource, or even a measure of diplomacy to just kinda "nudge" the opponent into a weaker spot...

I have yet to play a game of Civ3 that DIDN'T require both good diplomacy and good warfare... just hasn't happened yet.

[Code Monkey]

Originally posted by Xentropy
I think unit prices should be raised as the game goes from age to age..

This is just an idea, not playtested at all. Has anyone tried making a mod like this? No other changes, no more attack rating on modern units or anything, JUST higher costs as the game progresses.

It may work, it may not, would require playtesting :>

I'm playtesing a mod I made over the past 3 days (still in ancient era so no idea how well it works) sort of similar to what you mention.

I tweaked unit values and costs so that with minor exceptions they maintain their relative strengths within a given age but are significantly stronger than older counterparts. Accordingly, cost scales with a unit's strength. Some older units got cheaper, modern units got a lot more costly (260 shields for modern armor as an example but the stats are 46/24/3).

It's not exactly like your idea, I was trying to see if it were possible to make a more "realistic" representation of unit advances but balance the costs as well. A tank no longer goes down to a cavalry charge, but you can build more than 3 units of cav for the cost of one tank. I'm hoping the AI will upgrade and advance more aggressively as well as weigh modern units more importantly in military strength - we'll see.

[Xentropy]

Originally posted by Code Monkey
I'm playtesing a mod I made over the past 3 days (still in ancient era so no idea how well it works) sort of similar to what you mention.

I tweaked unit values and costs so that with minor exceptions they maintain their relative strengths within a given age but are significantly stronger than older counterparts. Accordingly, cost scales with a unit's strength. Some older units got cheaper, modern units got a lot more costly (260 shields for modern armor as an example but the stats are 46/24/3).

Sounds cool, CM! Please let us know how the AI plays under this mod (not to mention what other ways the game is affected) :>