SP Improvement Mod Thoughts

[halfzware]

I'm a big fan of MP games in general, but a D2 MP gameis a too much of a commitment for me, so I stick with SP. Currently SP has a lot of problems, so I was thinking about putting together a mod to address some of those problems.

The biggest issue SP games face is the weak AI. The AI generally takes the attitude that more has to be better, and as such produces a ****load of troops. The problem is that most of those troops suck.

With your basic two map types (wrap around maps with zero impassable squares (inland, eye) or your more choke friendly maps (paragnos) - you are faced with two crappy options. Either a frustrating game where maintaining a border is impossible, and the AI just throws 30-50 armies of 60-300 crappy units at total random -- or you get a dull game where you put a couple good armies at critical points and the AI just throws endless useless troops at you with no chance of winning. Both these options suck.

Ideally one could play one of the more choke friendly maps against the AI and have the AI be roughly capable of putting up a fight. I think the only way to remedy this is to manipulate what the AI builds.

I messed around a lot with two great mods, the recruitable rebalance mod, and the conceptual balance mod (national). Both of these seek to rebalance basic troops, but they both take very different directions.

CB basically slashes recruitment costs across the board.
RR basically buffs recruitable units.

A great example of both is javellen troops. CB reduces the cost to ~5. RR increases the damage and precision. Both make javellen troops useful early game, although in very different ways.

Although your multiplayer milage may vary, for SP - the CB approach seems better. The AI can already output huge amounts of troops, so reducing gold cost is not a good idea here - making more or less anything recruitable useful is much friendlier to the AI, and makes for a more interesting game IMO. The real key to any AI tweeking mod would have to start with improving recruitable units.


Gold and Resources
I also took a serious look at the effects of different resource and gold settings on the AI. Greatly increasing available resources had a surprisingly good effect. The AI produced much larger quantities of much better troops.

My guess at why this happened is: the AI wants to spend all its gold every turn. It buys a bunch of troops from its one fort, then spends the rest of its often sizeable income on whatever else it can buy once it hits the resource limit in capital.

So the more you increase its resources, the more it spends its gold in national troops - often mixing in healthy doses of high end troops. This is exactly what the AI needs to do to put together better armies, so its a step in the right direction.

Another nice thing is the split between resources and gold. A large increase in resources, with only a minor increase in gold allows the production of many higher quality national troops without the player being overly able to spam mages.

Scales
The AI takes seemingly random scales. That means that the more punishing the scales are, the more the AI suffers. The biggest problem with regards scales - and with regards to the AIs tendency to field HUGE armies is supply.

The easy fix to this is to just wildly increase available supply, while decreasing the effect that negative scales have on supply. The problem with this is players can abuse the hell out of this.

Since my goal here is to make the AI as competitive as possible, increasing supply, while slightly reducing the supply penalty for bad scales (perhaps which increasing other negative effects) is probably the best approach. If players want to abuse this, its their time and their game.

Summoning Spells
The AI likes summoning a lot more in recent patches. One of the problems is that it keeps summoning weaker low level summons till very late in the game. My guess is that the AI takes a more balanced approach towards research than your average player.

Early Game Summons
Early game summons suck. Even with the CB spell changes, many early game summons are less than useful.

Gem cost alone isn't the problem. Part of the problem lies with the cost of having an excellent researcher spend tons of turns summoning units one at a time.

Its my opinion that the early game summons (conj 2 for instance) would be balanced even with very low gem costs (ie: 3-5 gems). For one thing, in the early game taking valuable researchers off of research is a hefty cost alone especially when most of the time you are going to need to spend a minimum of 10 turns summoning in order to have a useful set of units. Is it worth while to spend 10 turns with a ~200 gold mage and 30 earth gems (before turn 30) in order to produce 10 cave drakes? Yes. Is it worth it spending the same amount of gold, but 50 earth gems to produce 10 cave drakes? Not unless you're desperate. Otherwise its better to wait till you can at least summon the troll kings court, or mechanical men.

Another problem that the AI has in this regard

Mid/Late Game Summons
CB does a good job with some of this. Giving the seasonal summons a little protection is a smart move. On the other hand, CB also seriously buffs Vine Ogres (!?). If anything in the game needs a buff, vine ogres aren't it. 1 gem for 3 or 4 ogres is one of the deals of the century.

I'm not quite sure what to do with the mid game summons. It seems to me that Legion of Wraiths is a little overrated, especially if you have Well up and running. Being able to produce 15 wraiths every turn for 30 gems is a no brainer every time.


Anyway -- this has gotten way too long. Anyone have any thoughts or comments?

[shovah]

make a god you want the ai to use and set it as human then turn it ai on turn 1, also try making a mod that alters a specific nation to make its unit more useful and then set the ai as that nation.

[Saber Cherry]

I had been working on a magic mod, but I never finished it. I did use it in a single test game though. Unfortunately I have not had much experience with the late-game "Omega" spells. For example I've never actually experienced Murdering Winter, Haunted Forests, Armies of Gold, or Astral Corruption... so I can only analyze them theoretically (from reading threads, mainly), which makes them very hard to balance accurately!

I think it does quite a good job on the early-game summons - stronger, cheaper, less hungry, and so forth - but I never wanted to post it until I thought it did a good job on all magic spells. Too bad, because it gives a neat new descriptions to each of the foul spawns

One way that the AIs in Space Empires 4 and Total Annihilation were tweaked was to create AI-only technologies and units that compensate for concepts the AI cannot understand. In Total Annihilation, for example, the AI would never build nukes, even though it built silos - so it could never launch nukes. The solution was to make an "AI Nuke Silo" that did not need nukes built before firing; it just automatically reloaded itself every few minutes.

In Dominions II, every nation (not themes, unfortunately) could be given a commander type that generates surplus food. If the commander is cheap, the AI will automatically build them, and keep them with armies... the player simply has to NOT build them (or disable them for his own nation). Or there could be one mod per nation giving all the national leaders surplus food, leaving the player to enable the mod for each AI nation that will be in the game. Hmmm.... yeah, I like that version. Unfortunately, it will make besieging the AI much harder if they never starve.

Well, I'll think about it a bit. I'm attaching my magic mod to this post, if you want to look at it... it does NOT make vine ogres stronger =)

p.s. I should note that it contains every spell in the game and (IIRC) most of the summoned creatures in the game, with complete (usually modded) stats.

[shovah]

sweet dude, this sounds like alot of fun

[quantum_mechani]

halfzware said:

I'm a big fan of MP games in general, but a D2 MP gameis a too much of a commitment for me, so I stick with SP.

Why don't you play blitz games? That exact issue is why I often prefer them to other dominions games.

[halfzware]

Wow, so I tried MP, and it was unbelievable. Such a great game. I always really liked D2, but multiplayer completely changes the game. There is just no comparison. Needless to say, any thoughts on making an SP mod have gone out the window while I spend my free time trying to learn all the tricks in MP.

BTW - thanks for answering huge numbers of stupid questions QM, you're like my mentor

[PDF]

Hi,
I'd be very happy to have a SP mod improving AI - I didn't really play SP anymore for more than a year due to the AI shortfalls ..
Before that I did some experiments myself some time ago with AI-improving mods (it was before the CB mods).

I totally agree on your conclusions regarding gold and supply.

It seems that with current CB and its very cheap militias the AI has even more ****loads of it ... OTOH, I had some success improving the AI builds by just "removing" the crappy troops (militias/non ranged LI mostly)- modding them to cost 999 gold. Then the AI was just forced to build heavies, so couldn't waste all its gold in basic troops due to resources constraints, so did build more mages ... a real virtuous circle !
The only real problems with that approach are that some troops are totally removed from the game, plus you have to make some tweaks so as *not* to get thru events free militia costing 66 each in upkeep !! So it's not very practical...

Regarding summons, I suppose it's the same : as the AI waste his resources on the easiest thing it can get, making the bad low level summons "non castable" should make the trick (sure it'll reduce some of the variety for the player, but these summons are usually really unnecessary).

[NTJedi]

Personally I've been changing the .map file to provide the computer with greater advantages... such as giving it a magic site with blood income, one or two starting blood spells, and a high level immobile blood mage able to cast those spells. Last game the AI had demon knights in his armies which don't suffer from starving and give quite a kick in battles.

Also giving the AI a few other immobile casters which increase its research rate as well.

HOPEFULLY the AI for Dominions_3 will be improved so it doesn't buy tons of low-level crappy units which cause serious starving and disease.