national armies?

[PDF]

Originally posted by Gandalf Parker:

quote:

---

Originally posted by Zen:
I wouldn't vote for Satyr Hoplites (in their current form), they are useless. I'd say Centaur Warriors and Coral Guards for Atlantis.

But I dont think the AI does them that way. Isnt it more like lt troops, hvy troops, lt cavalry, hvy cavalry? Its just one AI. If the good thing for Ulm is hvy infantry over lt infantry then Pangaea and Atlantis get the same change dont they?

Of course if it goes to seperate AI's for each nation then all kinds of great strategies could be put in.

---

If you just change the order to Hvy Inf, Hvy Cav, Lt Inf, Lt Cav it'll be OK IMHO - From what I've seen in MP this is more or less what 90% of players do ...

[Arryn]

Originally posted by Gandalf Parker:
I think the best answer will be splitting the AI. And Id love to see them outside scriptable.

Which is what Paradox did with HoI, for example. IMO, the only thing that makes HoI remotely worth playing is the fan-made tweaks to the AI, since Paradox has no clue how to code a decent one. IW has made a good one, but it could be made far better if the almost unlimited amount of time that fans can devote to it were to be applied to externalized AI files. It would also free up IW's time to devote to other aspects of the game.

[ February 25, 2004, 21:25: Message edited by: Arryn ]

[mlepinski]

It seems like the single most significant thing that could be done to improve the AI is getting it to build some forts early in the game. (Since in the long run these forts lead to more resources, more income and more oppertunities to build national troops). I'm personally of the opinion that any simple heuristic (whether it be build a fort in "any high income province", "any province with many neighbors I control" or "every fifth province") would be better than no AI fort building. Getting the AI to build forts in truly strategic locations seems difficult but it seems very likely that even sub-optimal fortress building would still help the AI significantly.

- Matt L. :->

[Peter Ebbesen]

Originally posted by Arryn:
Which is what Paradox did with HoI, for example. IMO, the only thing that makes HoI remotely worth playing is the fan-made tweaks to the AI, since Paradox has no clue how to code a decent one.

The fact that Paradox provides the tools that allows dedicated modders with lots of time on their hands to build better AIs than the stock AI they ship games with through an open AI architecture, and that they add new AI commands in patches to support new ideas that have come up years after the release of their games, proves that Paradox has no clue on how to code a decent AI!?!?

No modder could have made HOI use fronts intelligently if the HOI AI programming did not have support for the concept of fronts.

Say rather that there is rather more time to write and test AIs amongst the dozens of modders and thousands of happy players, than the ~7 man development team and beta testers have during each 12 month development cycle.


...But I digress. I certainly quite agree that if AI variables were exported for players to tweak, there would most likely arise a more competent AI. It would be strange were it otherwise, as the amount of time that players can spend on actually PLAYING and TESTING always outweighs the time developers can manifold.

[ February 26, 2004, 07:19: Message edited by: Peter Ebbesen ]

[Arralen]

Finally I found out:

Originally posted by Pocus (in another thread):
add to this the preferences about castles, temples, etc. in AI budget. 2 patches out, and we still have AI building mostly light indeps troops, because there is never new castles build by the AI.

That is to the point that fighting Ulm, Man or Pythium is often the same thing, with a mass of rabble in front of you, with the occasional national unit or mage thrown into the lot.

AFAIK the problem is two- or morefold:

1) The provinces stats have been dumped (dumbed?) down. DOM2 prov are smaller, generate less income and ressources than those from DOM1

2) The AI doesn't build forts. Maybe because of 1) - could be it's still searching for the right place when the game is already over.

3) It has the annoying habit of building troops all over the place. Fits nice to the other bad habit of moving armies constantly around - they could pick up troops quite nicely this way. But results in a hodgepodge of indie troops with few nation-specific ones (mostly the handful the initial commander started out with from the province I think)
Thinking 'bout it - could it be the case that the AI build troops in the forts first, than in the provinces in DOM1, and that this has been reversed somehow in DOM2?
Would not so easy to test this, but doable !?

4) The AI understands the concept of troops types only partially. It's uses rider for flanking, javelin-equipped troops are kept apart from other so they can fire their weapons during approach (at least, most times), it even sometimes uses "target Archers" or "target largest monster" instead of nearest for it's archers.
But it doesn't now about blessable troops or those with special aurae, or medium/heavy troops.
So obviously it cannot know if there's some type lacking in numbers or effective orders.
------------------------------------

Than again, I had Vanheim throw armies at me which where 80% those human guys with javelins and prot 15, backed up by some Vans etc., in a SP game lately. 'Was the Karan map (which has few "small" provinces, in contrast to Illwinters maps), with ressource settings on default.

[ February 25, 2004, 12:13: Message edited by: Arralen ]

[Gandalf Parker]

Personally I would love the idea of player scripted AIs. There are some old threads here discussing possibilities. The defensive one, the diplomatic one, the barbarian horde, the reource collector, the magic researcher, the insane one which every 10 turns randomly selects from the list... The problem is that the game has to be written from the beginning with hooks (tags, variables) available for everything seperate from the variables that the game uses to process things. You cant write "if my_gold greater than 600" without a my_gold tag in the game. Such a huge rewrite would take alot of effort. Whether the Devs thought it was a good idea or not, its still not likely to happen soon.

Changes for a patch could be done if they are quick and simple. Try developing the rule you WISH the AI would use and test-play it. I disagree with the "even if it built a fort every" type of responses, but I dont know because I havent tested it.

Hmmmm I guess I COULD start a game with Ulm and some nation of my choice and try to play Ulm but the rules I made up. Writing an AAR here would be nice also. Force myself to ...
A) if commander has 10 light (any-type-of-troop) then build 1 heavy (same type of troop)
B) every 5 provinces build temple
C) every 10 provinces build castle
Does that sound about right? At turn 100 I could turn Ulm over to the AI for control and see if I can beat him.

[PvK]

Originally posted by mlepinski:
It seems like the single most significant thing that could be done to improve the AI is getting it to build some forts early in the game. (Since in the long run these forts lead to more resources, more income and more oppertunities to build national troops). I'm personally of the opinion that any simple heuristic (whether it be build a fort in "any high income province", "any province with many neighbors I control" or "every fifth province") would be better than no AI fort building. Getting the AI to build forts in truly strategic locations seems difficult but it seems very likely that even sub-optimal fortress building would still help the AI significantly.

- Matt L. :->

Doesn't the AI sometimes build forts? Or are the extra forts I see them with all captured, conjured, or discovered forts?

A heuristic should probably be smarter than every fifth province. Should probably take into account current economy and number of existing forts, at least.

Incidentally, I think I've only built a single fort in all my Dominions II play so far... and I think it was overrun during construction and I lost that game to the Marignon AI. In winning games I've captured capitols at close to the same rate I'd want more forts, anyway.

Just some thoughts,

PvK

[DarkSol]

Originally posted by Gandalf Parker:

Then you are saying that Pangaea should build minotaurs over satyrs, and Atlantis should build those big guys over their little ones?

Pangaea should build centaur warriors, mostly.

[ February 26, 2004, 19:14: Message edited by: DarkSol ]

[Gandalf Parker]

Originally posted by DarkSol:

quote:

---

Originally posted by Gandalf Parker:

Then you are saying that Pangaea should build minotaurs over satyrs, and Atlantis should build those big guys over their little ones?

Pangaea should build centaur warriors, mostly.

---

Probably true.

But Im not sure how the AI (in general terms) considers Centaurs. I would suspect that Centaurs are cavalry and therefore it would be the light cavalry vs heavy cavalry rule.

But if the people who feel the AI should be "fixed" to purchase Heavy Infantry over light infantry (for Ulm is their usual example) then I think it would end up causing Pangaea to purchase Minotaurs instead of Satyrs. The point being that as long as its just 1 AI and 1 rule then it needs to be one that either takes the middle road as far as pros and cons, or be the rule most likely to boost the weaker nations.

[ February 26, 2004, 19:37: Message edited by: Gandalf Parker ]

[Arryn]

Originally posted by Gandalf Parker:
The point being that as long as its just 1 AI and 1 rule then it needs to be one that either takes the middle road as far as pros and cons, or be the rule most likely to boost the weaker nations.

Yet another point in support of having nation-specific AIs, preferably with externalized files for fan-modding ...