From my experience with Gal Civ II, Stardock is great at providing appropriate tools to prevent annoying micromanagement. I'm sure that you will be able to automate army and city building when it gets too tedious to be fun.

Its interesting, but the problem would probably be that this would be frustrating and you would spend more time trying to manage/influende/give orders to your generals than you would have done just moving the forces manually.

It smells a bit too much like Master of Orion III

Direct control is certainly preferable, until it isn't. At some point, given the massive maps that Stardock plans on having, turns will take hours. Unless there are suitable systems in place that allow you to automate movement and production, allowing you to focus on your grand strategy instead of whether your 57 scouts move N, W, E or S. Think about the civilization manager in GCII. How cool was it that in 5 seconds, you could change the production of each planet in your empire. I would certainly love having a similar system in place for Elemental.

Goodmorning all,

If you could give addiquately detailed instructions, and see explicately in list form what your general is going to do to achive the goals you give them, AND be able to edit those goals and means to achive the goals.  This wouldn't suck.

if any of those steps are missing you're inevitably going to have 'The black AI box of damn why is it doing that!?!?!?' which all AI goveners in all games i've played so far suffer from. 

I actually posted a fairly detailed suggestion on how to get arround this a long time ago

 

On AI; Transparant goals/means, on MMI (man(player) mashine interaction)... https://forums.elementalgame.com/357530

The suggestion pertained to a global AI, but easily any subbranch of it could be used without the rest to create a 'general'. 

The key issue is that to run a good AI anything the AI must make priorities, define goals, and set out steps on how to achive those goals. All those steps must be transparent to users to make a good general.   So you order the general to do something, or take control of something, then you can look at how it plans to achive that goal and can look at the 5 - 10 - 15 turn plan and make corrections.

If the system is any less user friendly and transparent high compition users will still perfer to do it all manually simply because nomatter how good the pre-defined AI it'll be a black box that does undesirable stuff.

Robbie

 

 

yea ... I NEVER use automation (on soldiers). To me war is at least as important to me as my cities. To tell me to automate some generals, you might as well tell me to set cities to "automate production". And the only thing under my control would be tech research, spell research, and my sovereign (and troops in his stack).

Now, mind you, such a game would be relatively interesting from an RPG perspective (only controlling your leader), but in a game such as this? No thank you. I will take my full direct control, and I will do so until I have already won the game, where I will simply switch all cities to producing wealth and buy more mercenaries as my army begins to wear thin (if at all) and levelling up my super stack of doom as its commanded by my sovereign and stomps down the survivors.

Most likely destroying newly captured cities by this point, unless it has a global effects wonder that is worth my time.

Oh ... well I like this sort of Idea. If only for areas where your victory is All but Assured

So yea ... if your just going to steam-roll without much effort, its a really good system. I'd say it would reduce alot of the "boring" from end-game, and give a player who is Micro-Surviving a chance to .. well ... survive.

er ... figuring out which stacks will defeat other stacks (in fall from heaven) without a combat calculator is, well, FUN. I don't know wether its fun cause Im good at it or fun cause I just like to do it. (I still use combat calculator for skirmishes and assasins). So yea ... I suppose if you had 10 cities producing units, and you had a wide front with 5 armies, one army spaced every 10 squares or so ... then I could see you automating new recruits to join certain armies, and having an automator to say "X unit from Y city goes to army A" ... I find this is oft not the case, but to have such a system would be nice.

However, I usually find the the cast to be Cities 1-4, build units and converge on waypoint. X. From there meet army A. Where army A is the one large grouping of soldiers and is either softening up the enemy or is already in a steamroll (otherwise Army A would still be at location X).

<!-- @page { margin: 0.79in } P { margin-bottom: 0.08in } -->

Goodmorning all

sorry: re the long delay school and this post got eatted twice by the forum beast

I don't think this would be overly difficult to implement. Consider that the AI will have to make plans, make decisions, and prioritize goals, it won't be able to compete if it can't. In order to make a good general / automation AI all you need to do is outline the plan, and the goals and priorities. Now the general can give tell you explicitly what he or she is going to do, and if you don't like it, you retain 100% control to tell the general to change their priorities or plan.

With a system such as this you can make a 10 -20 turn plan, either by building it from scratch, or telling the AI what you want to do, looking at the steps it suggests taking, then editing them to match what you really want to do. Then turn the plan over to a general to execute. Of course the general would keep you updated with daily reports about any changes to the plan as discussed.

In terms of coding all this would involve is creating an XML like text which gets added to at every step of the AI's planning out how to achieve the goal. Which is then made more pretty presented to the user to be changed, and then fed back into the AI for execution over the following turns.

This system would eliminate the primary problem of black box automation AI's; they take control away from the user, and do stupid stupid stuff (normally based off short term algorithms that look only at the present turn, if even that.) The method I'm suggesting would leave the user 100% in the drivers seat, but get the most micromanagement for far fewer mouse clicks. Additionally one can focus on a goal, and logic out several turns of actions all at once, and review it latter reminding you what you were thinking and what you were planning.


Hopefully this is more clear by way of suggestion. (and hopefully it won't get eatted by the forum beast again.)

Robbie Price