My proposition when the game came out was that ranged units should be allowed to stack on melee units. One ranged per one melee. If the melee unit is killed, then so is the ranged unit.
The AI might never use ranged units well, but at least they'd be defended if stacking were allowed.
That's been suggested by many people, and remains a bad idea. It would help the AI to use its ranged units, but it wouldn't help it to beat the human, because the human then has better-protected ranged units that the AI can't touch.
Some capabilities are certainly there in the existing AI, and I have the sense that different civs are programmed to prefer different types of units and different unit combinations, some of which are more effective than others. I've had AIs use combined missile and melee attacks to take down my units - then again, I've had situations where my only melee unit attacking a city is down to a couple of health, and if I move it away from the city for a turn (but still within range), the AI garrison will shoot a healthy archer instead.
In a recent game I had Babylon actively move to preserve its ranged units, moving them away from attackers, using melee troops where possible to block my attacks. While this game I massacred everything Wu Zetian threw at me without her getting the hint - she did pull a couple of archers back, but I think only because Isabella was attacking Beijing.
The best play I've seen recently from the AI, though, was Alexander's use of Companion Cavalry in a recent game, keeping them in the fog of war, moving to attack (and kill) my ranged units from behind and then retreating back out of sight ... and then the next turn moving them to give a melee unit the flanking bonus so that it could win an otherwise even combat with one of my melee units
So, I have no idea why the AI is so hit-and-miss, but it can certainly demonstrate good tactical play on occasion, so the routines must be in the AI's existing coding somewhere.
It'll also be really cool if they would "improve" their warfighting capability as time progresses
I've had the feeling that they do, since the AI seems to have been getting better and better in my recent games without any patches. But in my current game, although I lost a few Warriors, I've already forced surrenders from two civs who declared war decs and captured a key city and it's only 150 AD. The Ottomans didn't even bother producing ranged units so just ended up losing a lot of Warriors, while the Chinese did the classic 'lead with ranged' manoeuvre that I hadn't seen for a while (they did, however, combine melee and focused fire well enough to destroy most of the melee force I then possessed, while I think I only lost one melee unit to the Ottomans).
I've just played about 5 games so far (installed the game recently), but i do understand what the OP means. Yesterday while playing on the 4th difficulty (whatever that was) I stopped a massive invasion north of 30 units with just 3 crossbowman, no losses.
Prince. It's received wisdom around here that, like previous versions of Civ, the AI never gets better at combat at higher levels, it just gets bigger bonuses. However, my experiences have convinced me that the AI is programmed to give the player an easier time on lower difficulties by switching off some of its default behaviour at higher ones. So it will, for example, fire on your units with garrisoned ranged troops on King or Emperor when, in a recent multiplayer Prince game, AI garrisons never fired on my units. On Prince they won't generally escort settlers with melee units, but you'll never see a settler without an escort on higher difficulties. Ranged units will focus fire (i.e. multiple ranged units will fire on one of your units) more often at higher difficulties rather than just shooting at random. A damaged unit is programmed to fortify or retreat against a superior opponent - I've never had a case I can recall on Emperor where an Immortal will attack a nearly full strength Swordsmen on a hill, as happened in that Prince game (and resulted in the complete destruction of a nearly full strength Immortal).
I've found another thing that i find strange personally. It happened that same game. He attacked the first time and I stopped him cold, remained in war for about 100 turns with him just sacrificing a unit from time to time asking me for peace with me giving him everything i can (short of cities) even though i pummeled him. After getting my tribute low enough I sign peace with him and 20 turns later war again. After I crushed his army again (same 3 units, did not build any other the entire game) he again wanted me to give him stuff so I'm like it I'm gonna make him suffer than and went close to his cities. The moment i did that i receive offer from him for peace with everything he can spare but cities.
This is my big problem with the system, and it's because the AI is apparently programmed with too little sensitivity to understand the strategic situation. It is, after all, the same AI that calculates your odds in war and gives you military advisor suggestions like "We should sue for peace" when you've just captured three enemy cities, simply because they still have a larger army in score terms. There seem to be differences between civs in how far you can bargain them down, but I think a big problem is that the AI is not programmed to understand that cities other than the capital can be important. I can have Nebuchadnezzar demanding everything I have, regardless of how many cities I've taken, until the moment I have a unit inside Babylon's city borders, when he'll offer everything he has to make me go away.
Again, this seems to be one of the things that exists somewhere in the code, but is badly implemented in practice, because there are exceptions when the AI behaves more rationally. Sometimes wars will be fought over a particular city, and the AI will offer or accept peace based on its chances of achieving that objective, so then it knows to pay attention to a city that isn't the capital.
Example of my most recent game, which has to this point largely revolved around civs competing for the originally Ottoman city of Edirne, which is remarkably well-placed (in fact in the exact spot I'd originally intended to send a settler), with access to plenty of silk and a coastal location that makes it a good choice for a commerce city.
At first, the Ottomans declared war on me (Persia), the usual early war dec by someone with a big army and an attack route to an enemy capital. As the war was going on, China started moving forces into the area - whether initially to capture Edirne or to see who ended up with the weaker military and attack them I don't know, but the AI is usually inflexible in its objectives once they're set, so I suspect the goal had been to attack the Ottomans. Which they did when the Ottomans ran out of Warriors fighting me, and Persian forces were less well-positioned to attack Edirne themselves having fought a defensive war to that point. China succeeded in taking the city (I reduced it to 1 HP in an attack I'd expected to capture it, leaving it wide open). China immediately declared peace with the Ottomans. Interestingly they'd already taken Ankara, so they appeared to have a specific objective in mind - they declared peace the turn after they took Edirne, without any attempt to attack Istanbul.
China next declared war on me - I was between their main empire and Edirne. My friend Isabella decided to pitch in at that point, while the Ottomans apparently saw a chance to reclaim their lost territory. On this occasion I was the one in a position to capture Edirne, and the other civs' response to this was interesting. The Ottomans only had a couple of Warriors nearby; they'd retaken Ankara but, again, only seemed particularly interested in Edirne. There was, however, no way the one Warrior they could hit the city with could capture it before I would if they attacked. The AI declared peace with China.
That seemed to give me a free run at the city until Isabella's horsemen showed up over the hill. Since they were quick enough to grab the city if I reduced its defences enough, and even a half-strength Horseman is stronger than a Warrior against cities, I responded by moving my units closer to the city, making sure that the horses would have to take the long route round (which also happened to be through forest). So I would have time to reduce the defences and then capture the city without being beaten to it. Isabella's response? To declare peace with China. This was in her turn immediately after I moved my units to block the horsemen's approach, so it looked like a direct reaction to losing her chance to capture Edirne.
Finally, once I captured the city I also offered peace with China - they wouldn't give me any luxuries or gold, but they accepted a straight no-strings-attached peace deal, even though I'd never been anywhere near Beijing and in my experience civs will rarely offer or accept peace without conditions if they lose a 'minor' city and the AI's still telling them they have a larger army (it might even have been a 'they could wipe us off the planet' at that point). Again that's suggestive of an AI war objective "capture/keep Edirne" and an AI that can factor in whether that objective is achievable/has been achieved in making peace deals.
I play on Prince (I'm a noob) and the war AI is incredibley stupid. They never use a navy, they bring their embarked units near my units on land, etc.
The embark/disembarking rules, and naval combat/invasions generally, are the AI's weakest point - it has more problem doing that effectively than it does using ranged units.
EDIT: I think the core problem with the AI in combat is not any of its combat mechanics, it's its inflexibility - as I mentioned above, if the AI sets an objective it struggles to change it. This is true of its 'set destination' orders for units as well. I've noticed before that the AI will ignore a poorly-defended city to move to a more distant target (such as a capital), allowing its units to be shot/attacked en route. I just got a similar case in my current game, with the latest round of warfare against Suleiman. He was moving a spearman (with GG) and several archers to attack my newest city. I had Swordsmen and Archers coming up to attack. Instead of consolidating and defending (all his units were in forest, and the swordsmen were at half health), the archers that could have helped even the score continued moving out of range on their original path, leaving the spearmen and GG stranded while they were forced to fortify to heal.
Apparently the AI struggled to know what to do with the archers once they got to their destination since the spearmen and general weren't where they were supposed to be, but the spearmen were still alive so it seemed unable to plan what to do without them, so the archers just sat just out of range of the city doing nothing defaulting to "wait for other guys to catch up" (until my German allies appeared and started killing them). This may be the general problem with the AI's ranged issues - less that it doesn't know how to use ranged, but that if it's trying to coordinate them with melee that ends up in the wrong place due to damage (or death), the ranged can only follow its previous orders until the melee catches up. If their orders weren't to run away, they won't run when they should, for instance. Indeed, the spearmen and GG just got a promotion from their defence, and of course healed. The swordsmen were at that point nearly dead (having just failed to kill the spearmen in their last attack) - but the spearmen just carried on towards the original target rather than killing the swordsmen.
