I am with Pitor! I would love to see how they make this AI tick! It would be even better to get the privilege of working on it. Too bad those guys are on the other side of the world from me or I'd be beating their door down for a job . Its tactics are perfect, and its overall strategy are descent, and with the addition of transports, we may yet learn more from it by playing single player.

It makes sense Pitor, what this AI really needs for it to be the best it can is a good assessment of the current state of the war (heuristic function). Its not enough to weigh just units... you need to weigh in terrain layout, your positioning, your enemy's positioning, the possibilities of where the enemy may have secret allies, etc etc.

Now, I think these problems can be divided into two or more categories, and thereby be manageable. (I think this point has been brought up before, but can't remember who said it).

Category 1: Unit tactics - How to make the units involved in a battle manuver most effectively to bring the most firepower to bear, yet done in a way to minimize the enemy's ability to strike back (protecting weaker units, hitting the more powerful enemy units when possible, etc). This problem will become vastly more complex with transports, as all sorts of new movement possibilities are added. I'm quite impressed that they have even made it possible for the AI to use tranports, thats a nasty task.

Category 2: Non combat movement - Trying to get your units to the battle fronts as quickly as possible. Again, this will be affected by the addition of tranports. Also, this category can cover the strategic decisions of where the units are needed. If there is a nearby neutral country, for example, the AI needs to make a decision whether it wants to build up forces ready to invade it.

Category 3: Unit Choice - This is perhaps the most difficult category of AI decisions. Choices of what units to buy are heavily impacted by many factors. The AI needs to chose where its fronts are, what its goal is in each of those fronts. Couple examples: On one front the AI may be just trying to delay the enemy, and so only deploys the minimum number of units to do that. The AI may want to quickly take another territory, so it will focus on mobile units. On a third front, it may have plenty of time before the enemy can send reinforcements, so the AI would try to buy possibly slow but more cost effective units.

Hey, if you have any messenging services, I'd love to discuss this with you, Pitor! I want to learn all I can about this subject.

and post the log of discussion here

Maelstrom, you may be interested in an article I have written on my website (link in profile) on the subject of Zoning, an AI concept I came up with for RTS, but equally applicable to turn based gaming.

Give it a whirl.

2 Maelstrom : your ideas are really nice
I agree that it's necessary to divide the main problem into different interrelated subproblems/categories. I'll be interested to discuss all of this "in live". You can contact me on MSN or by mail with the following adress : "pseneus@yahoo.com".

Added! Don't block Dave S!

Just sitting here waiting for you to come online....

Very interesting! And quite applicable to MA, but not nearly on as complex a level . The Zones could be easily translated into country borders (and battle fronts). And this in turn could be used in the Unit Placement and Non-combat portions of the MA AI.

As per Mrakobe's request. here's a log of a discussion between Pitor and I about the AI. I don't have time to edit out all the fluff, its kinda long, but here it is if you're interested in looking. More to come I'm sure.

Interesting stuff, MAelstrom...

Oh.. If a butterfly flaps its wings in Beijeing, it'll rain in Times Square.

Chaos Theory is truly interesting stuff, it's not as complex (heh!) as it seems, on the theory side, the math side is an absolute monster, though. . .

Me.

Yes, AI uses multiple scales of thinking. They can be roughly divided into 2 parts - strategical and tactical, but they are not 100% distinct. Orders issued by top-level strategical part get more and more exact untill they are transformed into unit commands.

On the top-most level it estimates country control situation, concentrations of troops and their mobility. It has some prediction of the development of situation.

Now it uses transports for non-combat operations... The transports will be included into tactical groups with their passangers to make combat transportations(still in development).

Danmed Traitor!
Sky Keeper! Stop giving away our trade secrets!

"Danmed Traitor!
Sky Keeper! Stop giving away our trade secrets! "

What are the laws about treason over on your side of the planet?

Me.

The laws are very strict and severe for the treason - 10 games of MA against Tiger on New Paradize

that's too cruel even for a severe a crime as that. Nothing deserves that.

/low whistle...

And here I thought execution was bad, how we do it here...

Let us have a moment of silence...

Me.

Well, Belarus, like Russia has not yet signed that European Capital Publishment Maratorium agreement yet...

They need to talk a little better I quite reguarly see the AI place guerilla troops, then the next turn move them all one space... why didn't it just place them there?

The order sounds like "surround invaders with LAVs" So it does this very thing. Then they are moved by tactics for some reason. Precalculating tactics while placing units gives a burst of variants and should be done only if a good target is available (a kind of special opps).

One more thing: the last version of AI currently available is dated Oct 2003. And it was put aside.

Yeah, but as long as it's the current AI, and the people who designed it are still the ones who will give us new AI, we should point out its flaws. If the original AI is flawed, there's every chance that the same people might make the same mistake again... only one way to make sure that doesn't happen

2 Vic : Now that i've all the tricks of the AI motor, I'm rich as i can sell them to others (Im joking )

2 Skykeeper : thank you for these informations By having multiple scales of thinking, I guess you can save a lot of computation by discarding a lot of possibilities that otherwise (if you look only at the low level), you have to check. I remember when I started to play how the game and my choices were differents when I discovered the zoom function that allowed me to see the game on a global scale

I'm also impatient to see the new amelioration for the transports. Thank you guys for continuing to develop the game and keep up the good work