AI Artillery usage

[King Alexander]

Waiting too long could also result in more human units in this city.

But your right, many attacks are just rediculous. But when waiting to get enough units to beat the 20 units you maybe stacked in this city, the ai may never come to the conclusion to attack. And then people cry about the passive ai

And there is again the problem to give the ai a good enough ai to actually come with a plan to hit where you don't expect this.

Agreed: the AI "should" have a plan(if we want it harder): build a SoD and attack a player's city that CAN be captured with those units in SoD. Take 2-3 cities and then defend it all the way(with infanties/artilleries), like a player would have done. And always aim to get those cities with luxs/resources, so the human player is crippled.

I think the above can be done. The developers could "copy" and "paste" the best tactical moves made by human players and respond the same Also, the AI could "save" ALL it's elite units and attack with them red-lined player's units, to get Leaders easy.

Now, I know that the human brain can adjust and overcome all the possible AI plans, but certain AI moves would made the player's life, hell...

 

[yoshi74]

Looking forward to civ4. More experience, better machines, maybe new techniques. Still i believe we won't see a ai with can pass a civ-turing* test during the next two/three sequels.
(*civ-turing test: having a ai which plays so well that its not possible to say whether this civ is played by a human or the ai)

 

[Turner]

In one of the many AI discussions we've had on this board, I remember someone saying that Firaxis originally had a much harder AI. This one was so 'smart' that the humans rarely won on the higher levels, because the AI adapted too much. The choice was, an 'easy' AI with bonus', or a 'hard' one no one could beat.

Unfortunately, I don't remember who said it, or when it was said.

 

[yoshi74]

Don't know the post, but i seriosly doubt that. As long as the ai recieves bonus production and other favors, thats a clear sign that there was no other way to make the ai challeging at the higher levels. When it would be possible i think they used a good ai at deity instead of huge production boni. At least at sid it would be appropriate, but instead they gave just more production boni and addional starting units/settlers. A clear sign that there was no other option.
When it would be possible they would have implement it, if only to prove that they were capable to make such a smart ai.

 

[Turner]

Well, they didn't have the bonus' with the more robust AI. The extra units were compensation for dumbing down the AI.

It was a while ago..... I keep bringing it up hoping someone says "Yeah, it was me who posted that, and this is what they said...."

 

[sw99]

In one of the many AI discussions we've had on this board, I remember someone saying that Firaxis originally had a much harder AI. This one was so 'smart' that the humans rarely won on the higher levels, because the AI adapted too much. The choice was, an 'easy' AI with bonus', or a 'hard' one no one could beat.

Unfortunately, I don't remember who said it, or when it was said.

I don't remember seeing that post but I think you're reading too much into it. On levels above emporer it really wouldn't be that hard for the AI to crush the human if they were allowed to. Purely due to the fact that the AI has a huge numerical superiority in the early BC's. However it would be no fun to always get declared on in 2500 BC, which if the AI had true intelligence, it would do. So the AI was programmed to just make a puny 20G demand or somesuch. This isn't isolated to cases where the AI has inflated production either, if I played a regent game, OCC style with just a warrior defending, next to a civ with 30 knights, it's not that rediculous to expect that they won't walk all over me. The ai is restricted in how brutal they can act, purely in the interests of fun gameplay.

 

[yoshi74]

Thats certainly is a difference between smartness and meaniness. Actually i don't believe the ai is really programmed to win the game at all cost. Otherwise a OCC or a low military game would not be possible. Also you would get nuked every game when your close to space victory.
The ai rather plays like normal real world nations. They want to get big and mighty, but don't look for any opportunity to throw someout out of the race or stopping someone who is in a winning situation. And thats o.k. because the game would be no fun when being warriorrushed at 3000BC.
Through, this has not much to do with ai cleverness. Its merely a game design decision to make the game fun to play.

 

[dgfred]

Great post yoshi74, You are right on the mark. The advantages the AIs get
somewhat balance out the "stupid" things they do. If you consider the AIs
see everything and know all a much tougher AI would be able to slaughter
the players on many levels.

 

[Tomoyo]

You can make it so that the AI get smarter with the levels, not recieving any bonuses.

 

[sabo]

I saw the AI use arty ONCE in 3 years in an offensive attack, which makes me think it was an accident, he probalby didn't have anything better to shoot at so he shot my stack and then attacked it, I wish I would have saved it now that I know how rare that is

 

[silver 2039]

I had the AI attack me with arty also I was shocked.

 

[troytheface]

i read the post that someone mentioned and it's contention was that programmers could easily make a smarter ai... the real trick programmers faced was making an ai that was beatable yet competitive. (I believe the author of the post was a computer programmer) Since i have played -years now...i have managed to reach emperor with a 50-50 rate of success (huge map) But i too would like a smarter Ai when it comes to tactics...not when it comes to building or tech race....it seems that the ai is fine with the latter, but lame on the former....but i also suspect that when the ai does come up with a good, unexpected assault, many players will reload....a temptation that should be somehow noted in a players stats..(# of reloads ...just to keep ya honest- or at least to keep self deception in view...)

 

[binyo66]

Agree the game was designed for "fun and playable." The only problem I see is, IMO, after human reach the same balance (or have eliminated the AI bonus), the game is start to be boring (many of my games on monarch/emperor level) been abandoned after I knew I would win the game.
From my style of game
Ancient is a critical time,
Middle ages is a fun wonder time ,
industrial ages is where we knew win or lost.
Modern ages is boring
Basically I'm agree with bonus on the beginning for higher level, but after middle/modern ages, from that ages, AI should be programmed for smarter and smarter AI, since by that time basically AI has no bonus at all.

 

[alamo]

The thread-starter hasn't posted in days, but here is my 2 cents.

Building a big arty stack isn't always the most efficient way to advace your position or even just wage war. Arguably it is better to rush for airpower, since bombers are much more flexible than 1-step artys. There are alternative ways to wage war, also.

What if you build a bunch of cav or just a bunch of explorers? You could pillage and loot workers and have a crippling effect faster than arty can take down a city.

Also, the AI tends to attack invaders with everything available, so send in a bunch of infantry and wait for the AI to bleed itself.

 

[Berrern]

1. The thread-starter hasn't posted in days, but here is my 2 cents.

2. Building a big arty stack isn't always the most efficient way to advace your position or even just wage war. Arguably it is better to rush for airpower, since bombers are much more flexible than 1-step artys. There are alternative ways to wage war, also.

3. What if you build a bunch of cav or just a bunch of explorers? You could pillage and loot workers and have a crippling effect faster than arty can take down a city.

4. Also, the AI tends to attack invaders with everything available, so send in a bunch of infantry and wait for the AI to bleed itself.

I divided your post up in 4 points.

1. Don't worry, I keep my eyes peeled on this thread It's just that I haven't had much to add yet, so I've refrained from posting. I normally only post when I have some useful input, or when a thread is dying in which a question still remains unanswered.

2. Well, most players wants the ToE as soon as possible in the Industrial Ages, meaning they'll have to research Electricity first. When Electricity is discovered, I'd say most people go for the Replaceable Parts because of the speed increase for workers, and the powerful Infantry unit. As you know, you also get Artillery with this tech.
Now, depending on a lot of factors, rushing for Flight ain't always a good idea. First of all, depending on whether you've researched Industrialization or not, you have 5-6 techs left to research before you can build the Bombers. 5-6 techs = 20-24 turns assuming that you discover new techs in 4 turns. During all those turns you can build up a massive stack of Artillery! Also, many players have aged artillery units like the cannon, which can be upgraded.

I guess it all depends on 4 things:
- Your first target tech in the Industrial Age
- Your research speed/pace
- Number of aged bombardment units (that can be upgraded)
- Number of cities, and their production values

So, in some cases it might be best to go directly for Flight, but in others it's best to spit out as many Artillery units as possible.

3. The main reason I build Artillery units is to avoid losing units when attacking cities (by bombing them to size 6 and redlining all defenders), and building Cavalry or Explorers wont help with that.

4. If I place a good defensive units on Hills or Mountains (which I always try to, for best defense), the AI - in most cases - just walzes right past my mountain units.


I still haven't seen a good reply to one of my main questions - quoted below - have anyone tried any experiments on this?

[...]isn't it possible to mod the AI to produce more Artillery in the editor? Or will this only result in them keeping enormous stacks inside their cities?

 

[silver 2039]

I tried to make the AI build more artillery but they never used them offensively so it made no diffrience.

 

[Nihilist]

Korea, for example, is programmed by default to build more artillery. In my last game against them, though, they had about 7 artillery pieces in their entire empire, as opposed to other empires having perhaps 5.

I guess there is a point to the AI not building many artillery units, though, seeing as the human player can easily exploit this and capture them all, as has been posted earlier. That's pretty sad. I'm wondering whether they'll find a solution to this in Civ4. I certainly hope so.

 

[alamo]

4. If I place a good defensive units on Hills or Mountains (which I always try to, for best defense), the AI - in most cases - just walzes right past my mountain units.

Yeah, the AI is smarter about avoiding pointless attacks, but stupid in avoiding a SOD.

Try putting them out in the open, in stacks of 3-4. Pillage while you're there.

 

[Greebley]

The easiest change that I wish they would make is that if any AI artillery has movement left and has a enemy unit in range, then it will fire. You avoid the problem of huge artillery stacks because artillery is specifically moved to attack. It does mean if the AI has 4 Artillery in a city and the human SOD is approaching, it will fire them. Right now it doesn't even do that.

I think the biggest drawback of the AI is that it thinks locally and not globally.

Examples:
A worker can't look at the "global" picture of a city and ask the city what it needs. It instead looks at the terrain and decides.

This one may have changed:
When a unit moves it has no idea where any other AI units are (calculating all the units that can hit a given target is actually non-trivial). It can only look and see what enemies it can hit. In conquests, the AI seemed better at grouping units.

I don't think the AI builds units based on global need. It might know the defenders in the town, but it doesn't say "Our civ is short on bombers, lets build some". I think it will choose what to build based on probabilities biased by the advantages a building confers.

One thing I know works like the above is tech. The Tech preference table is a very good case in point on what I mean. It doesn't globally look ahead at techs or drive to Cavalry for its attack. Instead it just chooses a tech based on some parameters and random chance (My guess is building/unit selection is the same).

Global things are harder program and require a lot more interdependence, but can give better results if done right.

One example of global workers would be that every city "bids" for a worker based on need ( so if a city has plenty of worked tiles, it will bid low). This is modified by the distance the town is from the worker; the amount of corruption (more corrupt = lower bid), etc. I think worker actions would be much better if this was the way it worked. I am hoping they make the civ 4 AI more of a global thinker.

Caveat: All these examples are guesses (except the tech). There are how things appear to me. I may be completely wrong.