DGB246

Join Date 04/2006

(Discussion) How to improve the strategic AI

I thought I'd start a new thread focusing on the strategic AI so that we can discuss how to make it more competitive. I have seen improvements here across the betas although it still feels like a pushover on anything except ridiculous (due to big AI bonuses).

If we start listing thoughts and observations here then this may help Brad with programming effective strategies for the AI.

So, I'll start by listing a few thoughts:

1) The AI tends to overexpand too rapidly, planting settlements everywhere. It can get away with planting villages everywhere because of the tame monsters but many villages presents lots of problems compared to 1 or 2 focused settlements mainly due to a ) diluted prestige growth b ) lots of maintenance buildings costs c ) larger armies (and costs) needed to defend the wider territory d ) can become vulnerable to opponents on many fronts e ) reduces effectiveness of city enchantments - I think AIs should avoid spamming cities and focus on developing and defending a smaller number of cities in the act 1 phase. Other resources can be claimed with outposts (which doesnt risk city population).

2) AI puts lots of city buff enchantments on cities which prevents the AI building big mana surpluses, personally I tend to cast meditation on small settlements and I leave the other city enchantments (like inspiration) for big settlements

3) AI doesnt use spells like pillar of fire at all - AI needs to use these against big stacks of troops

4) The faction power rating seems to be purely based on the number of units a faction has, it doesnt seem to consider things like the power level of the spellcasters and stored pool of mana and gildar or researched techs and how developed the cities are

I have some more thoughts on the strategic AI but I'll leave it at this for now

Locked Post

Advanced Search

Watch this post

Do not email me updates for this post Email me updates for this post

Successfully updated karma reason!

1) Over expanding is PERFECTLY FINE! Pioneers are super cheap afterall

2) Buildings should NEVER be built, unless they have no maintenance.

3) Buildings that require maintenance should ONLY be produced in the BEST city, and only if it will improve that best city.

4) The best city should choose either Training Yard or Garden ... ALL OTHER cities should choose market.

5) Start out with spamming weakest possible soldier type. Have a full army ready for each champion if possible.

6) Attain loot. Sell all loot. Except maybe keep a cool weapon for the Sovereign.

7) Run at 0% taxes for as long, and as often as possible. Always switch to 0% taxes after selling loot.

8) Once you get high into the tech tree, focus on Armored Spearmen and Archers

9) put + initiative traits on archers (including charge), and place charge on your Spearmen too.

10) get Impulsive on any caster. If a caster has impulsive, try to get Fire III (or whatever Water level Blizzard needs)

11) pick your best units, and make them the army for your best Champion.

12) If you have an Impulsive Champion with Fireball or Blizzard, this is your best champion

13) Else, your best Champion is the one with the highest attack/health combination. With a preference to the Air 1 spell (tiebreaker).

14) Later on, if you can spare it, start replacing the charge-Spearmen with charge-Knights, with the additional trait of "Scout"

15) armor is your friend for Melee and Cavalry.

16) Archers are best produced in mass (no armor), but if you end up in archer vs archer battles, it might be worth investing half of your archers in leather armor. Still, initiative is always the priority. (Monks robes better than leather cuirass, leather (mostly) better than chain)

17) Maximise Armor

-> This means choose Chain where no initiative penalty, and Plate where there would be one.

Successfully updated karma reason!

There are alot of exploits that the player is currently using, like making weak conscripts, that the AI needn't learn. The devs will stomp those soon enough. As a player who avoids obvious exploits or mods them out when possible, I have a very different AI experience. For me, and you next week, the AI's biggest flaw is not knowing the problems the current improvements present to balance. The maintenance costs in the the early game are out of control. I have posted on how I would balance them. Once that is fixed, the AI needs to learn to scout its boundaries and find the best spots for a city. It needs at least a total of 7 combined points of Grain and Materials to settle in the early game. Outposts should be used otherwise until the midgame is reached. The areas between cities need to be filled in, even if there are no resources. Territory control is something to value as much as resources. They should also be able to realize when buildings will take too long in a small city with no production power.

There are also several points where the AI does things that are stupid because they annoy the player into declaring war. That is fine if they are stronger than me, but If they don't want war, they need to learn not to do certain things. Putting 2 units in every city of no power for the sake of having them there will ruin your economy and get you killed in the endgame. Making outposts inbetween two of my cities will get you killed. Demanding tribute when our power is nearly even will cause war and get you killed. Cutting off my ability to travel will get you killed. Settling too close to resources I currently control will get you killed and your women taken. Questing in my territory will get you killed and your head placed on a pike outside the ashes of your capital.

Successfully updated karma reason!

The AI shouldn't build cities on weak tiles. A 2 in wheat is too low.

The AI has explorers but they aren't off on their own scouting. They get lumped in armies with other units.

Pioneers also get lumped in armies with other units. They are so cheap to build there is no point to this.

The AI needs to get armies built with a movement of 3 and used for leveling champions. Then if they luck into something that boosts champion movement to 4, units on horse with the scouting trait.

Successfully updated karma reason!

Strangely enough the Pioneer/Scout army problem is because the AI does not have enough units. If take some time and play each faction, they will have plenty of units to choose from and almost never use scouts or pioneers unless they are scouting or, well... pioneering.

Successfully updated karma reason!

Tasunke, you're correct that you can spam pioneers and build villages everywhere and then just build money raising and pop raising buildings - I've done that too but the AI doesnt do this - it builds everything plus garrison troops so overall the settlement is making huge losses - the AI then raises taxes and gets productivity penalties. When you overexpand you're forced to make a choice between thin/patchy defences or building loads of costly units. Selling loot works at the moment but I suspect this will be rebalanced in future betas.

I know you can also spam cheap units and rush the AI but (as Sean mentioned) this strategy will definitely be nerfed soon.

So in my last game I deliberately avoided these strategies and just focused on building up and developing one city and just building a few troops to defend the city and assist my channeler with killing some of the tougher monsters - I only started building more cities when the 1st city was reaching its food cap - taking this approach, I was still able to massively out tech/economy the AI on hard due to maintenance cost savings, maximised city enchantment benefits, and being able to raise taxes to normal/high without losing productivity due to prisons/calm etc - you can quickly get the top tier buildings this way too. I might try this approach with ridiculous too. Out of curiousity, I checked the shops of other civs and they are selling basic things like leather armour whilst I have got plate armour, masterwork weapons, and enchanted cloaks - they have been left in the dust technologically.

So the point I'm making is that the AI should decide early on whether to go city spam (while minimising costs) or focused city but it needs to consider potential threats to its cities and whether it can defend them - neighbouring civs should always be considered a potential threat (monsters should be a threat too and hopefully by release they will be). Pioneers may be almost free but there are costs as well as benefits to creating new cities and the AI should weigh these up rather than just spamming them everywhere. By building a city you are choosing to funnel population growth away from your developed cities to new undeveloped cities, which means you lose population if those cities are captured or destroyed.

Successfully updated karma reason!

The thing to look forward to is that the AI will have different personalities. Some factions will try very hard to expand. Some will only want one massive city. This means the difference between only using outposts and mostly using cities as it currently does. But the AI is still crude. It hasn't been shaped for each faction yet, so we only deal with the crazy expansionists.

On a side note, in the game I just finished, the AI did very well because I removed the maintenance costs for early game techs. It was able to build more troops and have better research power for some reason. They all presented a pretty good challenge, except for a few that had bad starting locations. The ones I moved from challenging to hard did much better than the others. I like to put one Kingdom and one Empire on hard to get some powerful factions in the endgame. This, combined with all the other balances I made to items and the gold from shops and the prevention of zombie armies, made the AI very competent. The only thing it can't do is use Pillar of Fire, which is a must for defense. I don't use spells the AI can't though.

Successfully updated karma reason!

Quoting seanw3,

reply 6

because I removed the maintenance costs for early game techs.

Do you mean early game buildings?

I have no idea techs could have maintenance (very scared)

Successfully updated karma reason!

Early game improvements*

Successfully updated karma reason!

Sean, are you going to make a mod for the game? I would definitely be interested in trying it out.

I don't like the economy model but I guess we are stuck with it now and the AI needs to be aware of maintenance costs crippling its performance - that applies to gildar and mana. The AI needs to be wary of these costs - personally I avoid building loads of units because I want to keep tax as low as possible to maximise research - the AI should aim to keep research high to tech up to the better equipment. Currently, it will cripple itself through maintenance costs. Mass spearmen are useful but very often I can solo multiple armies of these with my channeler so they aren't that great.

The rebalancing for the next beta will probably mix things up a bit but I'm hoping that the AI won't keep focusing on the short term (mass armies) while neglecting things that help the long term (research rates and low maintanance) - we now know the shop is going to get a big nerf so maintenance costs will become more important for the next beta

Successfully updated karma reason!

I am working on a few really nice mods. But right now the modders are focused on making mods that don't change any core game concepts. I want to make sure the core game is great so that I can add mods like Ivory Towers that add some sweet content to the game. Modders are generally trying to be modular, so that my mod will function well with Heavenfall's which will function well with RavenX's.

There is alot of dev work going on that should greatly increase the fun level of the strategic aspect of the game. They are already working on balancing the early buildings to be viable in the early game. That means no maintenance costs for Lumbermill, Workshop, Barracks, and many other early options. They will also likely change the unit wages to have equipment cost less. Those two things will greatly improve the AI's ability to play the game.

But in general, if the game is released with merely par buildings and citylevelup bonuses, you can expect a mod from me that will adding content to buildings. I really want buildings to unlock shop items and trained unit traits.

Successfully updated karma reason!

I think you shouldn't be able to buy/build equipment of the "Champion" variety unless the city has an Adventurer's Guild.

(ie the +1 per level type of armor)

-> Equally, I think equipment you would be able to train normally (with units) shouldn't be all that expensive ...

---> Instead the Shop could have add-ons via buildings that, with City level, give a variety of rare loot to choose from, perhaps with a very limited stock.

This "rare stock" would be the type of thing you find on quest pick ups (the really good kind) and should be the ONLY sort of equipment that is very expensive (in the thousands perhaps).

Expensive equipment should only be expensive if it is rare, which means you can't build it normally with units.

Successfully updated karma reason!

Good ideas.

The major problem I have with endgame heroes is that they are easily surpassed by endgame trained units. I can make some units that are essentially a company of heroes. That is fun and all, but a hero should still be better than any one unit. This is where the Paths need come in. There should be a few new Paths every 4 levels. They should specialize a hero into a unique and powerful being. The lack of this creates a gap in the endgame. Heroes are only godlike when they find a piece of equipment that is overpowered. The mace that halves a unit's attack, for instance is an instant win in any game. There are lots of items that need balance and lots of content needed in the leveling system.

This is something though, that should be the final aspect of game development. Just want to make sure it is being taken into consideration when we talk about items.

Successfully updated karma reason!