2016 Winter Patch - Combat AI...?

My rule of thumb is 3 ranged units can kill 1 unit of equal strength entirely per turn. So you have 2 crossbows (say 45 strength with promotion firing at 48 strength knights) and 2 city/encampment ranged attacks.

On turn 1 let's say the first knight hits the encampment with the siege tower and then dies to the ranged attack. Knights #2 and #3 finish off the encampment and also kill crossbowman #1 which was in the encampment.
Turn 2 starts and knight #2 is roughly half-damaged by crossbowman #2 in the city center and the city center ranged strike. Knight #2 retreats and heal-promotes, assuming it's not yet reached level 1. Knights #3 and #4 attack the city and take to something well under half-health, not sure exactly, maybe 30-40%.
Turn 3 starts and knight #3 gets attacked down to half-health before attacking, but it doesn't matter. If Knight #4 doesn't end the city, #3 still has enough juice to end it, and knight #1 is probably still within range to attack that turn.

I'm pretty sure that's the way it plays, with a siege tower. Knight #4 might have to cross a river on turn 1 and then wait for the siege tower to show up in order to let all knights attack the city center. Siege mechanic doesn't really matter here, in this hypothetical with an overwhelming 4 knights and a siege tower. I'm not sure if 3 knights could win this scenario or not so I went conservative. :)

I understand the AI is terrible and Samurai are a bit unique still doing full-damage when hurt, but I was just trying to answer your question of how many units and how much time would it take to beat walls, an encamp, and 2 crossbows. My samurai example was actually performed on a human player, we were both kind of shocked at how fast Samurai took the city down. Cities themselves are very weak.
I guess that would work on Prince with no AI combat advantages and flat open paths from your position to both the city center and encampment. I didn't take into account that I'm thinking more about when the AI has significant combat bonuses. But I still have to question your attack plan. It relies heavily on open ground for approach and retreat. What if you can't attack the encampment on the first turn because of terrain? I don't understand Knight 3 and 4 attack city center and take it down to 30-40%. When did you take down the City Walls? Because a smart AI would definitely sacrafice a crossbow to capture your siege tower after attacking encampment. Even if you held onto your siege tower, how did it get from next to the encampment to next to the city in one turn? It's definitely possible but also unlikely. Why did the city center and crossbow both attack Knight 2? I would attack Knight 2 with a crossbow and Knight 3 with city center leaving you with only 1 non injured Knight and no upgrade for Knight 2 or 3. Since there are no siege mechanics being used the city is going to heal about as much as a single Knight did damage the previous turn. Is it possible to attack a city multiple times in one turn using a single siege tower? I didn't think it was possible.
I really appreciate your input for my question, not trying to beat your answer down. Just a little back and forth discussion. The samurai thing is just super unique because of the normal massive decline in attack strength for injured units. I think this is major factor why the AI is so hesitant to attack with injured units.
 
I agree it is super easy to take cities now because the AI is not really defending. My point is that when properly defended they can be almost impossible to take when terrain limits the number of units that can attack per turn or limit approach attack and retreat options.
The AI doesn't use garrisoned ranged units currently. If a city has walls, an encampment and two ranged units that's 4 ranged shots per turn. If it is not possible to make a single turn approach where you can attack the city that turn or the next turn that unit will be decimated before they do any damage.
If the ranged units are level 2 upgrades then the 4 ranged attacks are going to kill 1-2 units per turn as they approach.
Do you think if a human was defending a city that it would still be easy to take without overwhelming forces?
I see your point, and agree to some extent. A city with an encampment should be hard to take, and the default approach should not be "just rush the city". The default approach should be to take out the encampment before you approach the city. Otherwise you should be ready to get your behind handed to you. But certainly, it's a problem that the AI is clueless about how to use a ranged garrison, even more so because ranged units are generally OP and the +10 when sitting in a district is rather devastating under those circumstances (and that's a level 1 promotion!). Also, something I really liked from Civ5 community patch was how a garrison unit suffered collateral damage when city was attacked. Thus, you would gradually take out a garrison unit unless it used some of its turns to heal, which was a really nice balancing feature, and definitely something I feel should return.
 
My guess is that they are working on the AI but that doing so created other bugs which they r squashing. Everything always takes longer than you think esp re computer programming
 
Yeah, I think that if they literally just programmed the AI to always make every possible attack, no matter what, that would result in a significantly more effective AI. Sure, the AI would waste units, and sure, human players could easily bait it into traps. But at least that AI would actually be able to do something now and then. It would always take a city if its units surrounded it. That would be far better.

Ideally, obviously, you'd have an AI that shows good judgement about when to attack and when to hold back. But I think an AI that was hyper-aggressive would be much more functional than one that too terrified of taking any damage to attack. And it wouldn't be very hard to program.
 
I guess that would work on Prince with no AI combat advantages and flat open paths from your position to both the city center and encampment

You're right, this is a critical difference...we were playing on Prince and I'm used to the other human not having any undue combat advantages. I'm not even sure what the Deity bonus is, perhaps it's +5 or +6 and also applies to the city and changes substantively how much damage is done.

On some of the questions - the siege tower lets units completely bypass walls as if they are not there (i.e. you don't have to take walls down) and also allows more than one unit do so (so yes multiple attacks one turn). You would think only units next to the siege tower would benefit but one time I swore my unit on the opposite side of the city was benefiting...not sure on that point. I was assuming the encampment is only 2 tiles away from the city center and that there might be one river and some hills in the way but wasn't assuming this was like worst-case nightmare scenario where you can only get 1 or 2 units next to the city (in which case you are waiting for Cavalry Corps or Artillery). Knights occupy the same tile as siege tower and thus defend it...crossbowman can't just jump in and capture it. Attacking knight #2 on turn 2 is arbitrary and it could be knight #3. I also can potentially see your argument for splitting up the ranged attacks. I tend to focus fire and prioritize taking a unit off the field so that it can't escape and then go heal somewhere and come back to cause me pain another day. All that said if I'm wrong it can't be by a ton. Maybe it takes an extra knight and an extra turn (5 knights, 4 turns) on Deity.

Maybe the AI is trying to be a little too fancy or recalculating what it should do too often. The basic human logic/evaluation applied to the question of "how much military power do they have locally, how much do I have, and can I win here?" is not complex. Then put some firing sequence and positioning logic in there (ranged first, focus fire, melee next to city with support unit) and tell the AI to commit to that plan for 5 turns (i.e. don't recalculate and suddenly run away after 2 turns)....and even if it's not winning every time, at least it's scaring somebody.
 
Yeah, I think that if they literally just programmed the AI to always make every possible attack, no matter what, that would result in a significantly more effective AI. Sure, the AI would waste units, and sure, human players could easily bait it into traps. But at least that AI would actually be able to do something now and then. It would always take a city if its units surrounded it. That would be far better.

Ideally, obviously, you'd have an AI that shows good judgement about when to attack and when to hold back. But I think an AI that was hyper-aggressive would be much more functional than one that too terrified of taking any damage to attack. And it wouldn't be very hard to program.
This is more or less how the AI behaved in Civ3. Upon war declaration, they would throw their stack at you and this way, become a real threat on higher levels. To me, this was really fun and exciting as you could lose the game. Once the AI had access to bombers, they wouldn't hesitate to bomb the hell out of you. You would also get runaway Civs, usually the agricultural types with fast city growth.
I'd prefer such an insane AI plenty times over the useless AI we face now in civ6.
 
I'd prefer such an insane AI plenty times over the useless AI we face now in civ6.
At a minimum, there should be a "raging AI combat" game option you could select to do this. Maybe the game would even speed up a bit, because who knows how many useless calculations the AI is making for those useless movements.
 
Here is my evidence of post patch AI intelligence, at war with France, havent seen a military unit but they move an unescorted settler next to my city....
 

Attachments

  • stoopid.JPG
    stoopid.JPG
    461.1 KB · Views: 219
Here is my evidence of post patch AI intelligence, at war with France, havent seen a military unit but they move an unescorted settler next to my city....

On my latest playthrough I've not only seen two unescorted barbarian builders - apparently captured from Arabia - but they actively moved in range to be captured by my archer or warrior. I've seen multiple unescorted settlers - and was unable to take them only because my adjacent unit was a scout.
 
This is more or less how the AI behaved in Civ3. Upon war declaration, they would throw their stack at you and this way, become a real threat on higher levels. To me, this was really fun and exciting as you could lose the game. Once the AI had access to bombers, they wouldn't hesitate to bomb the hell out of you. You would also get runaway Civs, usually the agricultural types with fast city growth.
I'd prefer such an insane AI plenty times over the useless AI we face now in civ6.

It's also essentially the Civ V approach that people complained was too easy because it was exploitable and the AI never attempted to preserve its units. Civ VI appears to be the reaction to those complaints. But it did produce better results than Civ VI's AI - and especially as cities are less able to defend themselves may work better here than it did in Civ V.
 
This is more or less how the AI behaved in Civ3. Upon war declaration, they would throw their stack at you and this way, become a real threat on higher levels. To me, this was really fun and exciting as you could lose the game. Once the AI had access to bombers, they wouldn't hesitate to bomb the hell out of you. You would also get runaway Civs, usually the agricultural types with fast city growth.
I'd prefer such an insane AI plenty times over the useless AI we face now in civ6.

Speaking of Civ3 anyone remember on launch in regards to air units? Vanilla Civ3 fighters intercept ability was broken and you were unable to defend against air unit. The poorly play tested and buggy launch features of post civ2 world is becoming quite the constant theme!
 
I see your point, and agree to some extent. A city with an encampment should be hard to take, and the default approach should not be "just rush the city". The default approach should be to take out the encampment before you approach the city. Otherwise you should be ready to get your behind handed to you. But certainly, it's a problem that the AI is clueless about how to use a ranged garrison, even more so because ranged units are generally OP and the +10 when sitting in a district is rather devastating under those circumstances (and that's a level 1 promotion!). Also, something I really liked from Civ5 community patch was how a garrison unit suffered collateral damage when city was attacked. Thus, you would gradually take out a garrison unit unless it used some of its turns to heal, which was a really nice balancing feature, and definitely something I feel should return.
Definitely used that mod on 5, and the need for it illustrates my point. It was difficult for the AI to take a city in 5, and now in 6 you can have two garrisoned units and the additional encampment ranged fire to deal with. And as you said, after a couple of shots you have a level 1 ranged unit that is OP +10 ( +15 for Teddy with his home continent bonus). So basically if you always keep enough gold in your treasury that you can buy 2 ranged units for the city about to be attacked (Spoiler: it's the city with several units from another leader milling about your border) then the AI stands little chance of taking the city even with a substantial force.
It's actually a bit of an AI exploit that I used a couple times. Remove all units from a border city with an encampment, the AI will see the city as weak, they will declare war and attack your city (or at least approach your city with a few units and have them randomly attack and dance around a bit). I buy a couple of ranged units and fire away. By the end you will have a couple of upgraded ranged units and everything left in the opponents treasury when they wave the white flag and beg for peace.
 
Havent noticed any improvement on the combat AI, it does seem to be better at the builder aspect of the game though: I've seen it hit Modern very far ahead of me, though that was Deity and the AI beelines like crazy. Combat improvements may be for a later patch, if not it at least actually builds some units now.
 
From what I've been reading since the game was released, the primary root causes of AI incompetency in combat are:

- AI currently overvalues unit self-preservation. e.g. even if they have a strong unit, they are reluctant to bring it near a walled city because god forbid their HP goes down from 100 to 96. If they do decide to bring it, as soon as it takes a hit, it wants to retreat and heal.
- AI units don't behave as a "pack", it's more each-for-his-own kind of mentality. So they don't "sacrifice" or "risk" any units because every life is precious
- AI is indecisive. e.g. one turn a unit gets orders to attack a city, next turn the orders might change to "chill out"

@Siesta Guru explains all this a lot more eloquently in the AI+ mod threads. https://forums.civfanatics.com/resources/ai.25439/

I understand some of this he is able to tweak but most of the behavior is in the SDK that only the devs have access to today
 
Here is my evidence of post patch AI intelligence, at war with France, havent seen a military unit but they move an unescorted settler next to my city....
I just had this happen literally 2 hours ago. To be fair, I didn't have a unit in the city when it moved there... but I sure did as soon as I saw it. AI has to respect ability to buy units.
 
On my latest playthrough I've not only seen two unescorted barbarian builders - apparently captured from Arabia - but they actively moved in range to be captured by my archer or warrior. I've seen multiple unescorted settlers - and was unable to take them only because my adjacent unit was a scout.

you might be interested to know that scouts can take settlers too
 
Back
Top Bottom