Questions about AI preformance and bettering it...

[LordChance]

Turning down resources seems to be directly counter productive since most elite troops are alot more resource intense then money intense...and you can kill any amount of chaff with the right strat, elites are more interesting

Ye been looking at the stuff from semi random and implemented a number of builds that would set a human player at -points....seems to be working for some nation perhaps some are just lost causes

Well there seems to be alot to work through in the map commands and I got some new ideas

cheers

/LordChance

[LordChance]

Followup and new questions...

After running about a dozen test games (3 of them 3 player multi games with 20 AI for added mayham) with diffrent tweaks I found that most nations can be forced to produce alot better armies. As an example my brother did force restart one game after being seriously wiped by a knight/longbow army from Man. But it wont work in the long run due to the AI's build policies yet a potent boast in early game to midgame.

Yet it seems only the armies with one good higher gold then resource and one good high resource unit gets a really good mix as ulm and other nations just spam out their 10-15/20ish
units which while potent in mass in the first year or two kinda has no long time use. I assume they first buy their basic infantry say 20-40 and then choke on the resource cost...a fix that comes to me as I write is increasing the capitol resource cap by adding resources to capital sites must test that one.

So to the questions.
1: Is there anything told by the makers of the game how the AI prioritize unit buys ? (Im thinking of ways to encourage buying capitol only mages and heavy cavalry/sacreds for example..without to radical changes)

Would be interesting for example if it is biased towards building units in the capital if so giving it enought resources to build for most of the money the AI will recive would be a fix for units but not mages...

2: Poptypes for map commands, while nice and so are there anyway to affect what units are in a poptype ? (I guess not since BI takes the route it does with limiting units)

3: Since I asume the above is a no no, I've been replacing some of the units in diffrent poptypes with better working units removing units such as militia for longbowmen and heavier infantry and it works well but when I try to change commanders/priests/mounted commanders and such for units with any magic they do not recive it. Anyone know what I'm doing wrong generally I just run #selectmonster XXX (random independant commander) #copystats #copyspr
sometimes adding #magicskill X X for some custom magic.

4: Is there anyway to make a timed event ? Something that will happen turn X ? Never heard anything like it so I assume no but there has been something like it in every game I've modded so why not ask

/cheers Lord Chance...who intends to make some good AI opposition whatever the cost

[Endoperez]

Have you tried changing the resource setting from new game creation, or starting games in which RanDom has given the AI better-designed pretenders?

[LordChance]

Endoperez said:
Have you tried changing the resource setting from new game creation, or starting games in which RanDom has given the AI better-designed pretenders?

Yepp...and changes like it mostly affect the start of the game, only a few nations seems to gain a real strong long time bost, the rest produce chaff a bit better chaff then normal

Did a quick test during the moring with a map I'm working on an added a +500 resource site on the startsites and running only 50 on the resource setting...darn getting wiped up by multiple 120+ Saurmatian Catapracts antrophagi archers and sauromatian raiders before third years end felt very nice instead of facing hordes of barbarians Had the same effects that changing scales has ultimatly thought working widely diffrent with diffrent nations while all got stronger.



Perfect Gandalf going to run the debug and check if I can find any patterns to when and why the AI buys what


Yepp placing the mages where no problem well going to go through it tonight if tonights game doesnt drag on to long


About nationals and poptypes its all fooling around to see what makes an improvement currently I run tests with only 7 set nations on a heavily modded map so its a controlled enviroment. One experiment so far is making human expansion harder (limiting usefulness of awe/fear) since its where the human player really starts to dominate the AI's.



About the timed events...well somewhere around the 3-4+ yr in most of the mp games with AI's we have run the AI can no longer keep up why because SC and thugs eat up their armies if they are a hundred or seven thousand. It would be interesting if at around that time it was possible to give the AI some possible solutions to such problems in a moderate amount since it rarely produces any by itself.
And it would open up some RPish elements to creating maps which are always fun since the games I play are rarely about balance just a fun and challanging pasttime.


It did just strike me has anyone made some real tests between the diffrent AI models for singleplayer, aggressive seems the better/harder one since its more ready to push early but who knows...


By the way D3 mapmaking and modding is very much exciting because on one hand you seem to be able to do everything and on the other everything but what you really want I'll continue the mad experiments tonight


/Cheers LordChance

[Gandalf Parker]

LordChance said:
About nationals and poptypes its all fooling around to see what makes an improvement currently I run tests with only 7 set nations on a heavily modded map so its a controlled enviroment. One experiment so far is making human expansion harder (limiting usefulness of awe/fear) since its where the human player really starts to dominate the AI's.

I have a mini-map I use for tests like this if you dont have one already. It makes it easy to do different terrains and nations, but have them close enough to test against each other in one or two turns. Its here as a "Battle Simulator"
http://www.dom3minions.com/downloads.htm

About the timed events...well somewhere around the 3-4+ yr in most of the mp games with AI's we have run the AI can no longer keep up why because SC and thugs eat up their armies if they are a hundred or seven thousand. It would be interesting if at around that time it was possible to give the AI some possible solutions to such problems in a moderate amount since it rarely produces any by itself.
And it would open up some RPish elements to creating maps which are always fun since the games I play are rarely about balance just a fun and challanging pasttime.

Check out this thread...
http://www.shrapnelcommunity.com/thr...6&o=31&fpart=1



It did just strike me has anyone made some real tests between the diffrent AI models for singleplayer, aggressive seems the better/harder one since its more ready to push early but who knows...


By the way D3 mapmaking and modding is very much exciting because on one hand you seem to be able to do everything and on the other everything but what you really want I'll continue the mad experiments tonight


/Cheers LordChance

[/quote]

[LordChance]

Poked around in your ideas for the scenario earlier noticed that you asked about AI using custom spells it does...
By chance I made a really mean and cheap spell (was intended as a ritual not a combat spell ) and in the following test that AI killed everyone (including me) was a summon spell and likely there is a reason that most in combat spells are resonably bad or cost gems.


Making a themed scenario with events along a mod/change each turn sounds like an interesting idea for a long time game...
Must think of how to apply it in a quicker game where turns take a couple of minutes...perhaps a yearly change...

Perhaps a small side program darn...so many things to try...but I guess I'll go for a early/mid balance before poking around in that after all the AI's reaching end game is a first priorit And I do more then enought programing at uni...

/Cheers LordChance out...

[Gandalf Parker]

LordChance said:
Followup and new questions...
a fix that comes to me as I write is increasing the capitol resource cap by adding resources to capital sites must test that one.

That was why I recommended turning down the resources. I know it seems reversed but since castles draw resources from neighboring provinces it means that only castles would tend to have the resources needed for large armies. Therefore most of the armies would come form locations like home capitals.

1: Is there anything told by the makers of the game how the AI prioritize unit buys ?

Lots of interesting stuff can be learned by using the -d commandline switch to turn on debugging to a log. Like the order things are done in the game, how it handles stealth, or selection of mage spells, or ordering troops. Such as, a snippet from the start of a game would be like this...
Code:

---

host: buildarmies

  new com Astrologer in Arcoscephale (3736)

  new unit Hypaspist in Arcoscephale (45071) for player 27

  new unit Hypaspist in Arcoscephale (45072) for player 27

  new unit Hypaspist in Arcoscephale (45073) for player 27

  new unit Hypaspist in Arcoscephale (45074) for player 27

  new com Slayer in Abysia (3737)

  new unit Abysian Infantry in Abysia (45076) for player 8

  new unit Abysian Infantry in Abysia (45077) for player 8

  new unit Abysian Infantry in Abysia (45078) for player 8

  new com Manflayer in Sauromatia (3738)

  new unit Sauromatian Amazon in Sauromatia (45287) for player 4

---

Would be interesting for example if it is biased towards building units in the capital if so giving it enought resources to build for most of the money the AI will recive would be a fix for units but not mages...

It would fix some and break others. Some nations are meant to rely on indepts for basic troops. But it would make a different game and possibly an overall better game against AIs

2: Poptypes for map commands, while nice and so are there anyway to affect what units are in a poptype ? (I guess not since BI takes the route it does with limiting units)

Only by using MODs. And you have to be careful since some indept populations overlap units with some nationals.

3: Since I asume the above is a no no, I've been replacing some of the units in diffrent poptypes with better working units removing units such as militia for longbowmen and heavier infantry and it works well but when I try to change commanders/priests/mounted commanders and such for units with any magic they do not recive it.

Dont know the answer to your magic question. But map commands can give provinces mages. At least with the initial defenders. Check out "Poke in the Eye" map for some extremes,
http://www.dom3minions.com/downloads.htm
or the SemiRandom program for some more logical additions to maps.
http://www.dom3minions.com/~semirand/

4: Is there anyway to make a timed event ? Something that will happen turn X ? Never heard anything like it so I assume no but there has been something like it in every game I've modded so why not ask

It depends on the timed event you have in mind. Many could be simulated with some creative modding. But it is untested ground.