I really liked the "-1 : We know what you are doing"  In galciv2

In this case it could be " -1: You are a bit too cheesy "

Why can't the AI cope with this cheese?  Seems as though it could, so long as it took the threat as a priority.

I'm taking it for granted that we identify this situation, early tuns, low player city count, etc.

I'm also taking it for granted that the AI being caught out here has made a moderately risky start, and certainly hasn't decided to turtle.

I'm going to give to responses, one is my go at a counter AI and expects the AI to do a lot of work, the other is done through a game play mechanic with simpler AI strategies.  I'll do the second one in a different post.

AI - Counter Attack

The counter attack strategy should work here, so long as the AI hasn't been to super risky in it's growth and has no force.  This strategy implies that we consider the threat above all others.

There are 2 types of counter forces that the AI might use, the power counter stack, and the speed counter stack.  The power counter stack works by being big enough to discourage/defeat enemy attack, the speed counter stack works by being very fast and/or numerous to avoid attack.

Feel free to amend, I don't really know the game that well, and TBH this is not something I've done before.  Each Strategy should be able to (I think) run in parallel, although they do spawn each other.

STRATEGY 1 - Counter Stack Machine

1. Identify out next city along his attack vector.  This is our frontier city.

1. All small cities begin building scout units, larger cities build shock troops (Move adds?  Poison?) - These are heading for the frontier city.

2. 50% of our forces (in power - atk * hp, with a preference for movement adds and casters) need to link up outside the frontier city immediately - The the half closest to him (cap force depending on Power v ETA) - We cannot wait long.  This is a power counter stack

3. Remaining 50% of forces need to move towards and start shoring up into the frontier city.

4. If we sight the player at our city borders before our counter stack is ready, the counter needs to add to the city, and we create a few new counter stacks with the fastest movement (this may be a single unit, or a combination of units if they have army move adds), this is the fast counter stack, follow Strategy 3

5. If not, power stack follows Strategy 2

6. Loop step 1, increasing the power stack in size each time, or until threat has diminished.

STRATEGY 2 (power counter stack)

1. Power stack moves directly towards the enemy city, if we encounter the enemy along the way (not in city) we check its power.

1.1. If it's < -20% or move than us, we attack

1.2. If its between +10% and -20% of us, we split 1 to 2 (max 15% or stack power) units off our army as separate fast counter stacks to follow Strategy 3, and the original stack pitches a tent and attacks if it attempts to pass. Loop to step 1

1.3. If its +10% - +30% of us we pitch a tent and attacks if it attempts to pass. Loop to step 1

1.4. If it's more than that, then we split 2 units off our army as separate fast counter stacks to follow Strategy 3, the original stack heads back to the city.  (this step in effect waits until strategy 1 produces a bigger force).

2. Once within spot range of city

2.1 Attack (combined) if enemy < -20% power than all units within spot range.

2.2 If its more than that, we split 1 to 2 (max 15% or stack power) units off our army as separate fast counter stacks to follow Strategy 3, the original stack pitches a tent and attacks if it attempts to pass towards city. Loop to step 2.

STRATEGY 3 (fast counter stack)

1. Move in an perpendicular arc (with additional units spread equally over the opposite perpendiculars) away from any enemy stack and towards the  nearest enemy city that is along the vector to the center of the enemy territory.

2. If unit is chased and cannot outrun, unit heads directly along the perpendicular away from the frontier city and the enemy unit.  Loop to step 1 if he his distance between himself and the chaser becomes greater instead of smaller.

3. Once within spot range of city (or move to the edge if not already)

3.1 Attack (combined) if enemy < -20% power than all units within spot range (combined forces).

3.2 If enemy between -20 and +200% combined forces we have pitch a tent on ouskirts. Loop to step 2.

3.3 If enemy > +200% combined forces re target FURTHER along the vector and loop to step 1.

yea, when I work more on it.  Doing the AI is my night job at Stardock. The weekends I get to do whatever I want which, this weekend, was helping with sound and music.

My suggestion here:

Cities get +1 militia per city level.

Bell Tower produces larger militia (groups of 5 instead of 3)?

Training Yard produces stronger militia (loses conscript trait perhaps if it has it, maybe gains one random positive trait)

Right now I select neither of those city upgrades, usually pick Mason or Gallows, Apothecary in a high prod city I don't plan on touching.

I'd second it. If for no other reason than Impulse has been bought/renamed as Gamestop iirc, and I'm not sure where we will be getting our "free Fallen Enchantress" from

FB has mentioned in several threads that he was testing in small maps.  Don't recall him ever saying he was testing large.

But maybe things would look different on large.  Early game rush might be less of an issue with more space between factions, filled with more tough monsters.

Just curious, how many people play on which sizes?  And what sizes do the other devs mostly test on.  (fwiw, I've only played large so far)

Why do I imagine now Frogboy singing the main-theme for a special edition of FE?

For some reason I get the impression of Frogboy singing Annie in the bathroom on the weekend.......Very disturbing

Forums aren't real-time.

I read through people's feedback and take notes. But I don't necessarily implement them in real-time. I  have to sit back and think about how I would go about doing it and talk to others on the team about the best way to implement.

I would assume that most people posting are posting because they enjoy posting. I would hope so anyway.  

I don't think it is stealing other's ideas if you making the mods for the love of the game and you are getting no financial reward from them.  And I mean, how many ideas are totally original, very first person who expressed them in the history of the world, anyway?

Yep, I did enjoy it, much more than I enjoy my job.  I'm your typical geek needing an outlet.

This only works if you give a crap what other players think of you. There's no real significant economic benefit to caring in ninety nine percent of cases, and if I've already won a rush I am about twice as strong as any other player on the map right off the bat. 

So, diplo relations penalties are irrelevant. You can't just tell me I'm supposed to care about which side of the line I'm on, you need to give me significant rewards for it, or penalties for ignoring it. You also need to give me some way to rationalize my border with an AI or I'm going to invade them anyway out of sheer annoyance. 

Based on Civ4 MP player reactions to rushes, I'd advise having one to three scouts out and patrolling the most likely invasion path for the nearest empires whether or not they appear strong or hostile. This is just good policy. I'd also teach the AI to rush. Rushing means giving up the ability to defend yourself from a third party's rush. If there's the potential I'll be attacked myself, I'll have to play slightly more defensively.

Every player should start the game with leather armor and basic weapons unlocked. These two item classes are so critical to early warfare that not immediately researching both is a suboptimal strategy. Period. No arguments. If you don't have both and you get rushed, you will lose. It's so blatantly required that it shouldn't be an active element of the early game because it screws up balance. 

The last thing you absolutely need to do is add mana cost scaling per model affected by aoe spells. My heroes faceroll through your unit heavy AI armies because I can just fireball or dirge them once or twice for irrelevant amounts of mana and they all die. With level, oh, ten or so magic sov focused on cost reduction and using that special staff, I can get the cost of fireball down to three mana. So you need to make aoe more expensive, because right now it utterly dominates single target damage spells (other than soul burn or whatever) on pretty much all counts. 

So it doesn't matter that your AI outbuilds me on troops, because I nuke their armorless minions into the ground. And if they actually attack my cities, a few leather equipped spearmen are more than sufficient to kill them all.

tl;dr

90% of your problem comes from poor unit construction prioritization (armorless units), lack of easy upgrades for units (units should have a big UPGRADE ME button you can hit to change them into something else), and aoe damage spells ripping them to shreds. Fix those balance problems and teach your AI to rush and you'll be there. 

If your territory is twice that of an opponent, they should find an ally and attack you. Unless you have the appropriate diplomacy techs to keep them happy. It should be seen as an imbalance of power and be thwarted. This is how it has always been. 

Other historical factors would be borders touching. To properly create the checkerboard allies we see with real empires, there needs to be some serious penalties to touching another's boarders. Once again, this can be offset by researching diplomacy.

Peace should take alot of work. It is so much more complicated than war.