Just wondering about the computer AI

Anything else to do with MoM IME
Post Reply
Posts: 108
Joined: Tue Nov 21, 2006 8:56 pm

Just wondering about the computer AI

Post by elliott20 » Tue Nov 11, 2008 6:23 am

If there's a thread on computer AI, just direct me there and close this thread. I was just wondering about how the AI in this game makes decision as I feel sometimes a lot of their actions are almost done on a purely random basis. diplomacy especially, but this goes to things like unit production, unit deployment, army movements, spell research, etc.

I understand that the AI personality trait dictates a large degree the direction of their efforts, but I'm having a hard time observing this as in the end, the only thing this seems to effect is their diplomacy.

Anyway, I'm just wondering how Implode has decided to handle the AI aspects.

Posts: 63
Joined: Sat Sep 23, 2006 3:54 pm

Re: Just wondering about the computer AI

Post by Virm » Tue Nov 11, 2008 7:40 am

As of the time of this post, no definate decisions have been announced by Implode regarding how he will handle AI. There's been a great deal of discussion about how (meaning implementation) and what (meaning the intent the AI should have) the AI should be in various threads, but if any decisions have been made, they have not yet been shared on the forums.

User avatar
Site Admin
Posts: 433
Joined: Fri Feb 24, 2006 3:35 am
Location: Newfoundland, Canada

Re: Just wondering about the computer AI

Post by Implode » Thu Nov 13, 2008 9:15 pm

No sorry, I've been to flat out busy. Any time I get to spend on MoM IME I'm spending trying to get 0.9.4 out, and not worry about the AI!

Its not dead (and never will be!! :) ) but I'm just finding less and less time to work on it these days :(

Posts: 108
Joined: Tue Nov 21, 2006 8:56 pm

Re: Just wondering about the computer AI

Post by elliott20 » Fri Nov 14, 2008 4:19 am

no worries dude. If anything, it just means the game would be best played in multiplayer. Right now, I'm content to just sit here and dream about the future.

I've been doing some reading on the subject myself and I do believe that a sophisticated AI is completely do-able. After all, most RTS games will have AIs that are horrifyingly efficient and will have no problem running you into the ground if you're not an efficient player. (And that's without cheating, mind you.) So I do think it is possible to do the same for the computer AI.

I came up with a simplistic model

A computer's strategy is basically the composite of several things: 1. personality, 2. assets, 3. priorities, and 4 situational circumstances

strategy also happens at several levels: global, resource, and tactical.

Global designates the direction in which the computer will take his/her empire and it's overall strategy to victory. That is, is the computer a builder, expansionist, or hybrid? It also dictates what the computer's winning strategy is based on: military conquest or spell of mastery?

resource management is, of course, a derivative of the global strategy at large, and an extension of priorities + assets + situational circumstances.

tactical purely focuses on the actual logistics of movement for things that can be moved. Most frequently, this would be referring to military forces in and out of combat.

The composition of personality, assets, priorities, and situational circumstances are still somewhat difficult for me to pin down, as I sometimes wonder if one is merely a derivative of another or vice versa.

Personality is probably the most concrete we can have a handle on, as in game mechanics already define this to a certain extent. (Though, I can't seem to find a list of all personalities available in MoM, so I'm kind of stumped as to how to model it any further)

more to come as I organize my thoughts.

Posts: 108
Joined: Tue Nov 21, 2006 8:56 pm

Re: Just wondering about the computer AI

Post by elliott20 » Mon Nov 17, 2008 3:45 am

Global Strategy

There are really only two ways to win the game, military conquest or SOM. And really, there are only 3 methods of expansion that you can run with: builder, focused, and expansion. The composition of those two aspects will for the most part define the global strategy.

Builder: This wizard will focus a great deal on climbing the tech tree and investing in his cities. Each city will be a zenith of production and prosperity as the wizard puts every effort to make said city more and more productive. These kinds of wizards will usually not have an aggressive expansion schedule as they would rather fortify their position first than branch out haphazardly. This is perhaps the most suitable one for going for an SOM victory.

Focused: This wizard has cities dedicated to a particular function and tries to specialize each city for it's own purpose. The cities sitting on borders will have military buildings as it's top priority while headquarter cities will probably get buildings that increases the wizard's resources. Fundamentally, this wizard will not dedicate as much time as the Builder to infrastructure building, but will spend at least enough to make the required units for battle. Flexible, but sometimes lacks focus.

Expansion: Expansion wizards function mostly on repetitive land grabs and military presence. For them, the best defense is a good offense and they will hammer away at their opponents borders until they wear them down. Most of the wizard's cities will be producing military units and military buildings. Spell casting efforts will be funneled towards strengthening the military. (i.e. casting summoning spells or casting spells like crusade)

However, not all combinations will work out very well and may produce very weak opponents.

i.e. SOM + expansionist means the wizard will aggressively build a bunch of cities but not bother build a lot of mana structures because of the resource investment. Unless the wizard in question is playing dark-elves, such an approach just won't work.

Posts: 46
Joined: Wed Oct 17, 2007 8:57 pm

Re: Just wondering about the computer AI

Post by Iluvalar » Mon Nov 17, 2008 3:41 pm

I don't think SOM is a strategy.
I mean a way or another, you'll need a good reasearch, a good skill and plenty of mana. Anyway, you'll need to increase your manas production. So it's not a big issue until at least SOM appear in your research list. Unless you play at easy or normal, you should end up in a good position (lot of territory, good armies, good spells) to acheive that. Most of the time, you can continu whatever strategy worked for you until then and continu your gradual conquest of the world. I don't say SOM is not in some case a quicker way to win, but for me it's never a goal straight at the begining of the game. AI should consider the spell as any other.

Posts: 108
Joined: Tue Nov 21, 2006 8:56 pm

Re: Just wondering about the computer AI

Post by elliott20 » Tue Nov 18, 2008 2:24 am

I agree, the more I think about this, more I'm convinced that the SOM victory is really nothing more than a bonus point thing than an actual viable method of winning. This game, because of it's tech-less design, is a lot more narrow in terms of strategy as a lot of the progression is built in.

In fact, I dare say that the expansionist builder would NOT do nearly as well in MoM as they would than say, Civ or Alpha Centauri for the simple reason that technology doesn't translate from one base to another. For any city to be effective, it NEEDS infrastructural development. (or else you'd be pumping out spearman and that's pretty much it) The question is how and how much build up should you do precisely.

there are several kinds of build ups with cities:

low-tech city, fast but ineffective troops: you plop a city down, you maybe put down a smithy and a stable but that's it. The rest of the troops effectiveness will either have to come from the heroes, retorts, or magic itself.

high tech city, powerful but expensive troops: you put down a city, and build the living crap out of it until you've reached your apex unit. expensive way to go but pretty much essential to remain competitive at later stage of the game. (that is, unless you cheat like crazy)

summoning circle cities: this is probably the only exception where you can still make an outpost city an effective spot. Once the city goes down, you drop a summoning circle in and summon your special unit. This is still reliant on huge magical investment though so it probably won't work well in tandem with the low-tech city approach. At least, not world wide.

specialized purpose cities: some cities are for research and mana generation, some cities are for money farming, and some cities are for troops building. This is probably the hardest one to do program wise as you pretty much need to create analysis tools that the computer would use to assess when to use what city for what. Besides, as the situation of the empires change, so will the role of a given city. This brings in another important wrinkle into the strategy outline - changes.

People's strategy do change over time. Nobody is going to play quick expansionist game forever nor would anyone think taking builder approach to it's extreme is a good idea. (else you'll end up with a bunch of crappy, undeveloped cities or a single super city but nothing else) Even builders will eventually go expansionist and vice versa.

In the theoretical world, the one thing that brings about a shift in strategy is situational circumstances. A wizard at war is going to react very differently to a wizard who has no enemies to a wizard who is at war with someone who is not on the same plane of existence.

Posts: 108
Joined: Tue Nov 21, 2006 8:56 pm

Re: Just wondering about the computer AI

Post by elliott20 » Fri Nov 21, 2008 3:10 am

so, what would the decision process for a fellow wizard be?

human beings make decisions basically like this

have a basic game plan
move towards setting up game plan
assess situation
assess own assets and assets needed
make changes to plan based on assets and situation

there's a permutation process. Unfortunately, you can't really program a computer to think this way. This is the reason why even though computers can think of every possible move in chess about 10-15 moves out, it still routinely loses to human chess masters. You can train a computer to think strategy and to do analysis, but fundamentally, the first step "having a game plan" will be it's undoing as it will always be playing the same game.

There is nothing wrong with that per se, as this is just a little game and it's not supposed to be the break through AI project. All it really means is that the pre-determined goal/game plan that a computer comes up with could very well be the only one he sticks with. Changing his game plan would be something a computer not bother with at all.

Post Reply