I would like to ask the developpers if it is in the realm of possibility, with our current level of software developpement, that we could not only build an AI that reacts to our actions (because, let's face it, the AI is currently very good to REACT, 400% more than most game I played), but to establish a general strategy, thinking on a chain of actions as a whole, rather than going for the targets of opportunity or usual "fleet + transport".
I mean, look at my actual game. The Altarians simply send some fleet in my territory, trying to go for the unescorted starbases. They don't seem to have a strategy in mind. Then, once in a while, I see 4-5 transport fleet going to 3-4 different targets, with disseminated fleets..
I have been at war with the Altarian for about 2 year (maybe more, I don't really keep track), and at first, I was concerned that they may attack the 3-4 planets near their borders that I grabbed from Minor Species or other major civs. These planets were under-defended, and I already planned to "leave them behind". Fight to the last ship, making as much damage as possible, but don't sacrifice my strategic position on it..
And they sent about 20 transports... at the Terrans, who were like 10 sectors away! I had all the time to intercept them, destroy every invasion fleet. Not ONCE the planets were more treathened by more than 1-2 fleets + 2 transports. Eventually, I used those planets as range-extender to make raids in their territory..
If the AI had focused on these target, really tried to beat me planet by planet, they would have seized these planets.
Also, WHEN it happened the AI sent transports against me, it always failed to conquer the planet. I would understand it if I mis-calculated the # of troop required to take a planet, but the AI should know better. Also, they don't try to cripple me by targeting the planets with:
- The highest production
- Highest cash-income
- Highest Research
So, in short:
- The AI should have a masterplan at war. It should begin as soon as they decide "I shall conquer Race X", until Race X finally dies
- Masterplan should be spread in those steps:
-- Build up transports + attack ships, and identify the best planets
-- Try to make friends with Race X's friends
-- Position fleets within striking distance to take at least 3-4 planets in at least 5 turns. The target-planets should be prioritized as followed:
1) Target with Strategic position (if Race X has planets within the attacker's core worlds, or right at its border, or even isolated planets which offer strategic advantages overall (maybe against future ennemy))
2) Offers the highest short-term advantage to conquerors and cripples the ennemy. I mean: planet with high productions (Manufacturing Capital, I'm looking at you!)
3) Offers the highest long-term advantage to conquerors. I mean: tech-focused planets
4) Highest population centers/economy/culture powerhouses.
The AI should not focus on the planet the least defended, but the planets which offers the more to conquer. There is a REASON why you defend more or less some planets.
-- The AI make is move, focusing on no more than 3 planets at the time. He shall not spread his transport fleets, and he should concentrate his battle fleet in power-packs, to gang-up powerful ennemies.
-- The AI shall accept a relapse (peace treaty) if he doesn't have any more transports but have the field advantage. Again, he shall not declare way except: he has to strike to defend or: he can finish the specie within 10 turns.
The AI has an incredible advantage over us: he can calculate much much more quicker than us, feeble humans. He should be able to plan how long an invasion is supposed to last, how many transports he need, etc..
- We need also a new rating: Soldering. I would like to know easily how well-trained are my troops compared to race X.
- If the AI is defending (which means, he didn't had the time to prepare his masterplan, but is forced into war), he should not sit and wait for the invaders to come to him. He should be aggressive, simply sending packs of fleet to stop conquering fleet, using his # advantage as much as possible.
I don't know WHAT is possible in all of this. I know AI programming is already tough ennough, but that's the main AI's weakness i've seen so far. If you do that, you will change opposing AI forever.