Good game, but....the AI is very bad.

MStavros · #1 May 10th, 2004, 08:01 PM

I have some ideas.

The AI shouldn't build LI at all. The AI cannot make good moves with the LI. It shouldn't build LC after turn 10. This should be a core rule for all nations.
I posted my idea about the AI castle building, in a previous thread.

As I said before, the castle building problem must be fixed first of all. Maybe with more forts the AI wont build as many LI units. If it will, than code it, to dont build LI at all, because LI controlled by the AI is everything just not a good army.

Well I started to make my mod, so we will see what will happen without LI.
One problem could be here, the provincial militias. Those are weak LI units, and I bet that the AI will build them.
Perhaps I will change all provinical militias as well. They will be strong HI units.

Gandalf Parker · #2 May 10th, 2004, 08:14 PM

Well with your mod you will be able to test that. There is a --superhost command which will let you loop the game on automatic. You can add --scoredump so that each turn will spit out the scoreboard. That way you can keep an eye on things and decide when to kill the loop to jump in and see how things are.

MStavros · #3 May 10th, 2004, 08:20 PM

Ah thanks Gandalf, good infos! It will be a few days until Im ready with the mod. If it will work, I will make it downloadable.

NTJedi · #4 May 10th, 2004, 09:37 PM

Actually the AI is okay considering the size of the gaming company. Developing an intelligent AI is probably one of the most challenging tasks gaming companies have. Most gaming companies develop a single artificial intelligence which follows a series of x,y,z steps. Once gamers identify the set pattern it's easy to conquer the computer opponents.

I believe the best choice would be to begin developing improved AI for Dominions_III since the biggest and best changes most likely cannot be done for Dominions_II.

BEST IMPROVEMENTS FOR THE IMPROVING THE ARTIFICIAL INTELLIGENCE:

1) Randomly selected AI personalities at the start of each game. Thus only one or two personalities out of 10 would use the tactic of massing lots of low level units.

2) With the assistance of scripting gamers are able to create or improve existing AI personalities. This was done long ago with Warcraft_II and Age_of_Empires_II.... and even recently done with Neverwinter Nights. I see no reason why it could not be introduced for Dominions_III.

This would allow for a strong replay value for AI opponents.

PvK · #5 May 10th, 2004, 10:01 PM

I'd say the AI is remarkably good. Especially considering both the complexity of the game, and the level of AI in other strategy games, even ones with much less complexity.

My own win/loss record against the AI is something like 3 complete losses (I was wiped out after making some initial progress), 2 convincing wins (not yet completed), and 5 games with very long stalemates and very near defeats before eventually I'm starting to pull ahead (but also not completed).

I'd say that's a pretty awesome record for the AI, considering I'm a very smart strategy game player with decades of gaming experience, and that the game has insane amounts of complexity.

Of course, I'm not just duelling against a single AI in these games - I tend to play against multiple AI's with some or all of them given bonuses, and I don't go nuts trying to exploit techniques to the utmost, but rather I play to enjoy the game by trying out different strategies and techniques to see how they work, and so on. I don't recall every having a more rewarding experience against an AI player in a strategy game.

The "Light Infantry issue" isn't so much an AI issue as it is a game balance issue. The LI fight like HI and so become rather ineffective in most cases, and therefore they are generally a weak choice. I would say that should be addressed by making LI a more viable choice, by adjusting the tactical AI so LI don't often get caught in melee with superior enemy forces, and/or by reducing their gold cost in most cases.

PvK

MStavros · #6 May 10th, 2004, 10:42 PM

Hum, maybe "lot of" problem will be solved, with the removal of LI. We shall see.

Damn I cant wait to see, what will the AI do than!

[ May 10, 2004, 23:14: Message edited by: Gandalf Parker ]

proteus · #7 May 11th, 2004, 11:57 AM

Quote:

Originally posted by PvK:
I'd say the AI is remarkably good. Especially considering both the complexity of the game, and the level of AI in other strategy games, even ones with much less complexity.

My own win/loss record against the AI is something like 3 complete losses (I was wiped out after making some initial progress), 2 convincing wins (not yet completed), and 5 games with very long stalemates and very near defeats before eventually I'm starting to pull ahead (but also not completed).

I'd say that's a pretty awesome record for the AI, considering I'm a very smart strategy game player with decades of gaming experience, and that the game has insane amounts of complexity.

Of course, I'm not just duelling against a single AI in these games - I tend to play against multiple AI's with some or all of them given bonuses, and I don't go nuts trying to exploit techniques to the utmost, but rather I play to enjoy the game by trying out different strategies and techniques to see how they work, and so on. I don't recall every having a more rewarding experience against an AI player in a strategy game.

The "Light Infantry issue" isn't so much an AI issue as it is a game balance issue. The LI fight like HI and so become rather ineffective in most cases, and therefore they are generally a weak choice. I would say that should be addressed by making LI a more viable choice, by adjusting the tactical AI so LI don't often get caught in melee with superior enemy forces, and/or by reducing their gold cost in most cases.

PvK

Hummm, are you a veteren player?? I am not, and I can beat the AI all the time easily, and I am not using any SCs etc. at all.