Does the AI have a rudimentary memory of tactics?

dante alighieri

Warlord
Joined
Oct 31, 2006
Messages
215
Location
Oberammergau, Germany
Does the AI have rudimentary memory of player tactics during a game? As Asoka, I had played a peaceful game up until the modern age. Sharing my continent with the greeks, the egyptians and the byzantines. Hatty voluntarily became my vassal because she's fearing invasion by Justinian. Pericles and I have gotten along thus far because I flooded his cities with hinduism early on and cut off his expansion way back in the bronze age.

Then Hannibal decides to be a jerk and declare war. (which dragged Mansa into it since he is Hannibal's vassal....not that it mattered, since Mansa's contribution was one galleon that got sunk) Hannibal sent over two stacks with 3 destroyers and six transports each. I used bombers to weaken them and then my own ships sank most of them. The few troops that survived and landed were quickly dispatched. Then he sent another wave over. This time they stopped their movement just out of bomber range so that on the next turn they could land and disembark the troops. I took care of the troops, but I was slightly impressed, since it looked like the AI knew I'd bomb the snot out of his ships and then wipe up who was left. This way his navy had full strength when my ships engaged them. He sank a lot of them, too. I was getting ready for the next wave but Justinian decided to DoW on me, probably because my attention would be divided into two fronts. Luckily i made peace with Hannibal the next turn. Now I'm going to show Justinian why Pericles was smart. I'm not going to capture his cities, I'm going to raze them. :mad:
 
It may be as simple as knowing about the bombers on the 2nd trip. On the first voyage he may not have known you had them until he'd gotten too close, and then it was too late. On the 2nd trip, keeping distance to make a one-turn push could be default behavior when facing a known threat of bombers.
 
Eh, I'm getting he just got lucky the second time. I've never experienced anything that would seem to indicate the AI "learning"
 
I don't know why he does what he does, but we are talking about Hannibal, after all. Here are a couple of my experiences with him doing what other A.I.s don't.

I was blockading Hannibal with pairs of promoted privateers. He used airships to knock them down 30%, then sent a handful of caravels out of at least half of his ports the same turn to try to finish off the privateers. Hannibals survivors returned to port to heal. He could turn out caravels faster than I could re-enforce my privateers, and they couldnt heal on station under aerial assault. I shifted privateers from city to city as best I could. In a few turns I abandoned the blockade completely.

The A.I. didn't attack too soon, and it didn't wait too long. It attacked with a mixed force when it had enough to break the blockade. That seemed smart enough to me.


I still remember my first warlords game with Blake's better A.I. ...

Me and my vassal had one continent, Peter and Hannibal had another. Hannibal was going for a space race victory, so I went to war on their continent to disrupt him.
I had oil, he didn't, Peter did. Hannibal got Peter's help. My tanks were busy with Peter's.

I normally do some things to run my civ the way I think it should be run (rather than what's the best way to use resources to win ) such as station a screen of frigates/ destroyers beyond my coasts to alert me of an invasion and possibly stop it, even though the A.I. couldnt conduct a proper amphibious assault in Warlords, or so I thought.

I saw Hannibal's task force coming towards my continent, and easily shifted adjacent destroyers to finish off his frigates, etc. Not too bright, right?

It was a diversion. Next turn two unescorted galleons of his cavalry went through the square where my my destroyer had been and went straight to my only oil. He managed to to pillage it and the nearby railroads before I could eradicate all of the cavalry. It was a major setback for me.
 
I don't think the AI "learns," although that would be REALLY neat...it could be a whole new difficulty level where nobody has production bonuses or extra units or any of that stuff...and tying the learned behavior in with user profiles would allow for different AI profiles. REALLY would be REALLY neat...not sure how possible, though...
 
I think the AI "remembers" what forces you had waiting last time, so the next time is more carefull aproaching.

I have seen something similar, when the first wave stoped outside my city and got destroyed by mass catapults. The next wave bypassed the city, out of range of catapults and went for the next city!
 
What I heard from Blake is that each turn the AI decides what to do with the data on its hand right at the moment.

With a few exceptions, the AI in each turn decides what is best to do each turn.
If I understood it completely(chances are I didn't), so in the event of a try for cultural victory, I asked myself, "how then do they plan for it?", and the answer that fits with it would be; One turn the AI checked that the chances of a cultural victory by themselves are good, rolls the "dice", decided to go for it and shut down its research. Next turn they check again, see themselves with research shut down and doing loads of culture, and decides to go for cultural victory then. And so on next turn and next.

I am probably completely wrong, but ah well.

Maybe he saw that his navy got owned by bombers last turn(after all, this is a data that they have) and therefore assumed that your cities had bombers?
 
I hope the AI didn't assume that all of my cities had bombers. I'd like ti think that it does the same thing I do when preparing to invade....check to see whats in the city as far as defenders go.
 
What I heard from Blake is that each turn the AI decides what to do with the data on its hand right at the moment.

With a few exceptions, the AI in each turn decides what is best to do each turn.
If I understood it completely(chances are I didn't), so in the event of a try for cultural victory, I asked myself, "how then do they plan for it?", and the answer that fits with it would be; One turn the AI checked that the chances of a cultural victory by themselves are good, rolls the "dice", decided to go for it and shut down its research. Next turn they check again, see themselves with research shut down and doing loads of culture, and decides to go for cultural victory then. And so on next turn and next.

I am probably completely wrong, but ah well.

Maybe he saw that his navy got owned by bombers last turn(after all, this is a data that they have) and therefore assumed that your cities had bombers?

On second thought, I probably don't really want to know how the A.I. "thinks'', because I'd only be tempted to out think it and trap it.

The anecdotes I mentioned made for fun games, because I didn't expect what Hannbal was going to do, and couldn't anticipate his every move.



Farewell, thread.:)
 
What I heard from Blake is that each turn the AI decides what to do with the data on its hand right at the moment.

With a few exceptions, the AI in each turn decides what is best to do each turn.
If I understood it completely(chances are I didn't), so in the event of a try for cultural victory, I asked myself, "how then do they plan for it?", and the answer that fits with it would be; One turn the AI checked that the chances of a cultural victory by themselves are good, rolls the "dice", decided to go for it and shut down its research. Next turn they check again, see themselves with research shut down and doing loads of culture, and decides to go for cultural victory then. And so on next turn and next.

I am probably completely wrong, but ah well.

Maybe he saw that his navy got owned by bombers last turn(after all, this is a data that they have) and therefore assumed that your cities had bombers?

I heard that the AI only remembers if it's planning to DoW on someone, and what victory type it seeks, from turn to turn.
 
I heard that the AI only remembers if it's planning to DoW on someone, and what victory type it seeks, from turn to turn.

I know there is exceptions(see my post), but I don't know quite which ones. What you said make perfect sense though.
 
I don't know whether the AI remembers previous defeats (would be great, but definitely a hassle to implement, so I don't really expect it), but another possible explanation for the situation you witnessed is that the AI had LoS to your city in the second attack, but not in the first. Although even then the decision to stay out of bomber range would be quite sophisticated for an AI. :)
 
Maybe your copy of CIV 4 has become sentient? Better unplug your PC quick before it takes over the world :crazyeye:
 
Top Bottom