Dominions 2. AI. Suggestions, that how to fix it.

[Nerfix]

I doubt that. I have seen plenty AI priests.

[Gandalf Parker]

Yet another item on the list that "fixing" it for one nation would make it pretty stupid for another. Then someone would be saying that it made too many.

[HJ]

I've seen plenty of AI priests. Who else would build a temple in every province otherwise?

After some prolonged games some points seem to be ok as far as I was able to tell. There are enough mages, commanders have items and gems on them, when the AI has enough troops it sticks them on the border, etc.

However, there are some points that still stick out. The main one is the lack of forts beyond the starting one, which leads to the lack of national units. There are heavy units in those armies (a lot of militia too, but they are there), but they are indie heavy infantry and cavalry, and not national ones. It's kinda odd when you fight IF Ulm and its whole army is brown and not black in color, and not a single templar in sight. This leads to the lack of supply and starving armies as well, since they recruit a lot of cheap units that then require a lot of supply. So this is all interlinked - no forts->reliance on indie militia->cheap units->numerous armies->big supply requirements->starvation due to no supplies->no forts. And they even take cheap forts (usually mausoleums), so it's not like they have to have tons of money to build them. They certainly have no problems with building temples everywhere. Labs are there, but I haven't seen one in provinces that give access to indie mages yet.

Another thing is gem income, which seems to be rather low. Also, it does seem that the AI empowers its pretenders, but I don't know whether it does that to his mages as well, both to make them stronger in their starting paths and to diversify them. If they don't diversify or empower, they cannot do searching very well, and I don't know how the AI handles the site-searching spells and whether it uses them at all. This would also be linked to alchemy - I don't know how the AI handles conVersion of gems, and whether it does that at all, and this would help with empowerment and diversification. Those things are very hard to tell, especially in game that takes so long to get to later stages and hence get a good sample size for judgement, but low gem income is readily observable in score graphs.

[ November 22, 2003, 20:24: Message edited by: HJ ]

[Aristoteles]

I meant, that the AI wont use his priests in battles correctly, IE wont cast enough bless spells etc.

[ November 23, 2003, 09:15: Message edited by: Aristoteles ]

[Kristoffer O]

Originally posted by HJ:
I've seen plenty of AI priests. Who else would build a temple in every province otherwise?

After some prolonged games some points seem to be ok as far as I was able to tell. There are enough mages, commanders have items and gems on them, when the AI has enough troops it sticks them on the border, etc.

However, there are some points that still stick out. The main one is the lack of forts beyond the starting one, which leads to the lack of national units. There are heavy units in those armies (a lot of militia too, but they are there), but they are indie heavy infantry and cavalry, and not national ones. It's kinda odd when you fight IF Ulm and its whole army is brown and not black in color, and not a single templar in sight. This leads to the lack of supply and starving armies as well, since they recruit a lot of cheap units that then require a lot of supply. So this is all interlinked - no forts->reliance on indie militia->cheap units->numerous armies->big supply requirements->starvation due to no supplies->no forts. And they even take cheap forts (usually mausoleums), so it's not like they have to have tons of money to build them. They certainly have no problems with building temples everywhere. Labs are there, but I haven't seen one in provinces that give access to indie mages yet.

The AI gets forts sometimes (events, sites, conquests) and likes to have them, but I do not think it builds them. As you say it would probably do much good if the AI knew where to build forts and saved money to do it when not in war. I suspect that this is the change that would improve tha AI the most. I hope something can be done about it without too much complications. Currently the AI is somewhat based upon each individual commander. Together they work as an ant hill, but the ant hill itself has few goals. To add a queen to the hive mind will probably not be easy (and I suspect that a queen is needed to decide where to build a castle).

Another thing is gem income, which seems to be rather low. Also, it does seem that the AI empowers its pretenders, but I don't know whether it does that to his mages as well, both to make them stronger in their starting paths and to diversify them. If they don't diversify or empower, they cannot do searching very well, and I don't know how the AI handles the site-searching spells and whether it uses them at all. This would also be linked to alchemy - I don't know how the AI handles conVersion of gems, and whether it does that at all, and this would help with empowerment and diversification. Those things are very hard to tell, especially in game that takes so long to get to later stages and hence get a good sample size for judgement, but low gem income is readily observable in score graphs.

The AI empowers mages. I have come across a mystic with 7 astral that doomed my poor tartatian titans when I stormed the Arcoscephalian home.

I'm not sure to what extent the AI uses search spells. It generally prefers normal search i believe.

[Mortifer]

That list from Aristoteles is valid, but I do not agree about the priests. The AI is building many priests.

[Gandalf Parker]

hey Kristoffer,
I know I can start a player as human then change it to be AI. Is there any way to change from AI to human? That would allow us to see what the AI is doing and verify if the list is generally vaoid or not.

[HJ]

Originally posted by Kristoffer O:
The AI gets forts sometimes (events, sites, conquests) and likes to have them, but I do not think it builds them. As you say it would probably do much good if the AI knew where to build forts and saved money to do it when not in war. I suspect that this is the change that would improve tha AI the most. I hope something can be done about it without too much complications. Currently the AI is somewhat based upon each individual commander. Together they work as an ant hill, but the ant hill itself has few goals. To add a queen to the hive mind will probably not be easy (and I suspect that a queen is needed to decide where to build a castle).

Thanks for the reply, Kristoffer.
I wouldn't know much about AI coding, but what formula you use for temples? Is it just "build temple in every province" or some other things also influence the decision? The thing is that I don't know what map characteristics the AI can recognize and take into account. JK said in another thread that it doesn't recognize victory points, so I wouldn' know what to suggest exactly since I don't know what parameters are there to draw upon.
For example, if it can recognize the indie troops recruitable in the province, it might be good to build the lab in the province where indie mages are available. Likewise, you can give it an algorythm to have a chance to build a fort if there is some high quality troop avalable (knights or amazons) and the province has high income/population at the same time, and is far enough apart from previous fort (taking supply into account). This would give them both access to more resources/money and higher quality troops, both indie and national ones. It might not be placed in the most strategic postition, and it might seem a bit arbitrary, but it would surely be better for the game. And it would also not be as common so as to build them everywhere all the time, and destroy the economy in the process. The random variable would help with that as well, so that the fort is not necessarily built even if requirements are there, but rather has a chance to be built so as to not overload the economy.
As for strategic placements, those are very context dependent. But maybe something with number of dryland neighbours in addition to some other parameters, to make it more likely to fortify choke points? It could result in something like a Maginot line, but then again, it would still provide supply and recruitment options at the same time.
As to hive mind vs. ant hill, again, the question would be what is the algorythm for temple building. Some of those independently thinking commanders are building temples, so the money for them either has to be saved, or is taken before troop recruitment. If there is a way to determine the sequence the money is spent, i.e. first on buildings and then on troops, maybe it would be a viable option. So the AI would first evaluate whether to build buildings, and then spend the rest on troops. If not enough money is available for a building, or there are no locations that match the requirements, then it would go on with recruiting troops. This could possibly circumvent the problem of scripting queen mind.
Also, the AI seems rather good when determining where and how much provincial defense to buy. It would seem that it's taking enemy presence into account. Maybe some parts of this algorythm could also be used in fort building decisions?

Originally posted by Kristoffer O:
The AI empowers mages. I have come across a mystic with 7 astral that doomed my poor tartatian titans when I stormed the Arcoscephalian home.

I'm not sure to what extent the AI uses search spells. It generally prefers normal search i believe.

This is very good to hear. The only thing I am not sure about is diversification and corresponding alchemy then. For example, will there be any Ulmish smiths with astral magic on them? If this is possible to script somehow, and again make it rare enough so that not all gems are wasted on empowerment and alchemy/diversification and that not every mage needs to get them, this could in theory greatly help both site searching ability, and magical threats that AI is able to field.
Building labs to recruit indie mages could also be an aid to this (this is the easiest way to diversify probably). Even though most of them would still have to be empowered, it would help to find the sites otherwise unavailable to the nation, and in fact ease the empowerment of both indie mages and the national ones.
Btw, I guess that the summons work ok, meaning that the AI will summon creatures if able to do so? Global enchantments don't seem to be a problem at all if they can cast them.

And Lastly, I just summoned my first Tartarian titan Last night. This game is great.

[ November 23, 2003, 20:28: Message edited by: HJ ]

[Mortifer]

Originally posted by Gandalf Parker:
hey Kristoffer,
I know I can start a player as human then change it to be AI. Is there any way to change from AI to human? That would allow us to see what the AI is doing and verify if the list is generally vaoid or not.

Someone already had this suggestion, as a 'special debug mode'.
This would help a LOT indeed.

[MStavros]

Originally posted by Mortifer:

quote:

---

Originally posted by Gandalf Parker:
hey Kristoffer,
I know I can start a player as human then change it to be AI. Is there any way to change from AI to human? That would allow us to see what the AI is doing and verify if the list is generally vaoid or not.

Someone already had this suggestion, as a 'special debug mode'.
This would help a LOT indeed.

---

Yup. I agree with this..and yup, someone had mentioned this already.