AI's cost/benefit analysis needs improvement, BADLY!
from 
GalCiv2 Forums
GalCiv2 Forums
For the time being, I have lost my interest in this game, and I don't think I'll come back to playing before the AI is significantly improved. As so many others, I find the crippling level too easy, and masochistic too difficult, because of the insane bonuses the AI gets. However, this morning I had an idea how the AI could be improved to pose more of a challenge even without bonuses to it's economy. I am not a computer programmer, so I don't know if this is easy to do, or even possible, but I am SURE it's worth a try.
The idea came to me after playing a game on crippling, large galaxy, 9 opponents, everything set to abundant, normal tech rate development. I was playing as the humans. The Iconians were in the upper left corner of the map, and they were incredibly powerful. They had really massive fleets, stronger then any of my other neighbors. I was happy that they liked me.
Then they invaded the Thalans, and began beating the crap out of them. Since I set everything on abundant, there was a pretty normal Thalan star system with four habitable planets close to Iconian territory. There were two class 5 planets, one class 11 and one class 24. The Iconians captured two small planets pretty quickly, but they didn't even try to take the bigger planets with larger population, even though they completely controlled the "air space" with their massive fleets and they had a couple of transports ready to invade (and they could have easily produced more, if they wanted to). What they did instead was to make peace with the Thalans after which they withdrew their "armada".
So I started to think - why the hell did they withdraw? Why don't they continue with their "blitzkrieg"? And then I realized that the AI's cost/benefit analysis is flawed. It's not only that the AI doesn't have a game plan, the AI also doesn't take long term cost and benefit in its calculations. The AI obviously thought like this:
"These planets are heavily populated and I will lose too many transports capturing them, which will make my empire weaker after I invaded them then before I did it."
IMO, the AI should be programmed to think like this:
"Yes, these are heavily populated planets, and I will lose many transports capturing them, BUT NEVER MIND, these are good planets, and in a long run, this will make my empire more powerfull, so I'll do it."
How difficult should it be to make this adjustment in the program? The chess programs are able to sacrifice pieces to achieve a strategic advantage, so why shouldn't GalCiv2's AI be able to do it?
The fact that the AI is such a coward is leading to the more and more ridiculous developments as the game progresses. Here's what happened meanwhile with me, the Drengin and the Korx. First there was a revolution in the Korx Dominion, and they changed their alignment to good, after which the Drengin invaded them. The Drengin were vastly more powerful, and began taking system after system. This was in the same time when the Iconians invaded the Thalans on the other side of my territory, so, for a while, I thought that the Drengin and the Iconians would build vast empires that would completely surround me, so I decided to answer the Korx demands for help and I invaded the Drengin, thus creating "a second front". Here's where the AI's stupidity came into play.
Since I had better ships then the Drengin, I successfully destroyed their "homeguard" small fleets in the systems close to my border, after which I successfully invaded them and took some 5 planets including the empire capital, Drengi.
Now, the logical course of action for the Drengin should have been to make peace with the Korx (a deal that they would be happy to accept), after which they should have moved their fleet of respectable size to defend their core systems from my invasion. HOW DIFFICULT IS IT TO PROGRAM THE AI TO DO SOMETHING SO OBVIOUS?
Once again, we come to see the AI's flawed cost/benefit analysis. I am sure that you have seen the following in your games as well. The AI masses several fleets in one single sector on the map, and then it doesn't use them at all. Why? Well, I try to explain it with my cost/benefit theory. The AI knows that my ships are better then his and that his counterstrike would cost him more then it would cost me (in number of ships). But it's also true that its army is larger then mine, and the AI could have slowed me considerably. The Drengin could have defended their worlds. So, it seems to me that the AI thought like this:
"His ships are stronger then mine, so it is not reasonable to attack him because I would suffer heavy losses."
Instead, the AI should think like this:
"His fleets are invading my really important planets, and I need to defend them at all costs. Yes, I will suffer heavy losses, but I can slow him down and in the long run defending those planets is more important then protecting my fleet."
How difficult would it be to make these adjustments which would make the game far more enjoyable?
The idea came to me after playing a game on crippling, large galaxy, 9 opponents, everything set to abundant, normal tech rate development. I was playing as the humans. The Iconians were in the upper left corner of the map, and they were incredibly powerful. They had really massive fleets, stronger then any of my other neighbors. I was happy that they liked me.
Then they invaded the Thalans, and began beating the crap out of them. Since I set everything on abundant, there was a pretty normal Thalan star system with four habitable planets close to Iconian territory. There were two class 5 planets, one class 11 and one class 24. The Iconians captured two small planets pretty quickly, but they didn't even try to take the bigger planets with larger population, even though they completely controlled the "air space" with their massive fleets and they had a couple of transports ready to invade (and they could have easily produced more, if they wanted to). What they did instead was to make peace with the Thalans after which they withdrew their "armada".
So I started to think - why the hell did they withdraw? Why don't they continue with their "blitzkrieg"? And then I realized that the AI's cost/benefit analysis is flawed. It's not only that the AI doesn't have a game plan, the AI also doesn't take long term cost and benefit in its calculations. The AI obviously thought like this:
"These planets are heavily populated and I will lose too many transports capturing them, which will make my empire weaker after I invaded them then before I did it."
IMO, the AI should be programmed to think like this:
"Yes, these are heavily populated planets, and I will lose many transports capturing them, BUT NEVER MIND, these are good planets, and in a long run, this will make my empire more powerfull, so I'll do it."
How difficult should it be to make this adjustment in the program? The chess programs are able to sacrifice pieces to achieve a strategic advantage, so why shouldn't GalCiv2's AI be able to do it?
The fact that the AI is such a coward is leading to the more and more ridiculous developments as the game progresses. Here's what happened meanwhile with me, the Drengin and the Korx. First there was a revolution in the Korx Dominion, and they changed their alignment to good, after which the Drengin invaded them. The Drengin were vastly more powerful, and began taking system after system. This was in the same time when the Iconians invaded the Thalans on the other side of my territory, so, for a while, I thought that the Drengin and the Iconians would build vast empires that would completely surround me, so I decided to answer the Korx demands for help and I invaded the Drengin, thus creating "a second front". Here's where the AI's stupidity came into play.
Since I had better ships then the Drengin, I successfully destroyed their "homeguard" small fleets in the systems close to my border, after which I successfully invaded them and took some 5 planets including the empire capital, Drengi.
Now, the logical course of action for the Drengin should have been to make peace with the Korx (a deal that they would be happy to accept), after which they should have moved their fleet of respectable size to defend their core systems from my invasion. HOW DIFFICULT IS IT TO PROGRAM THE AI TO DO SOMETHING SO OBVIOUS?
Once again, we come to see the AI's flawed cost/benefit analysis. I am sure that you have seen the following in your games as well. The AI masses several fleets in one single sector on the map, and then it doesn't use them at all. Why? Well, I try to explain it with my cost/benefit theory. The AI knows that my ships are better then his and that his counterstrike would cost him more then it would cost me (in number of ships). But it's also true that its army is larger then mine, and the AI could have slowed me considerably. The Drengin could have defended their worlds. So, it seems to me that the AI thought like this:
"His ships are stronger then mine, so it is not reasonable to attack him because I would suffer heavy losses."
Instead, the AI should think like this:
"His fleets are invading my really important planets, and I need to defend them at all costs. Yes, I will suffer heavy losses, but I can slow him down and in the long run defending those planets is more important then protecting my fleet."
How difficult would it be to make these adjustments which would make the game far more enjoyable?
