Force Organization

I think the ability to create a sequence of orders that can be preserved is fantastic, and in fact I have another wall of text like the original post with a similar system to that effect. However such a system makes far more sense if you already have established groups as conceptual building blocks before you start getting into the awesome things you can do with them. Like being able to issue a group a set of pre-planned orders, for example.

If you don't have that sort of conceptual structure already in place, then assigning units to places within the highly detailed pre-planned scheme becomes extremely tedious and complex. It is much easier to have a suite of operations available for a specific type of group, such as an army, which is assured to have the right units for the slots in that plan because the plan requires an inventory of units that matches the group's roster.

Doing things like formation-move becomes quite straightforward if you have a conceptual group already established. This same mechanism can be used for more complex maneuvers like splitting into groups which move independently while still being grouped together, such as scattering, increasing spacing, forming a battle line, encircling a target area at a specified distance, or what have you.

edit: I re-read my post and think more detail is necessary. Suppose you wanted a large army to march in a specified formation. Something simple. Battle tanks in front, anti-air in middle, artillery in the back. The easiest way to do this is to create a squad composed of one tank, one anti-air, and one artillery unit arranged to your satisfaction, and then create a company composed of an arbitrary number of these squads, and arrange them in a line. If you want to duplicate this pattern, order another company. These two forces can move about independently, but have the same governing logic. Modifying the pattern will modify both.

We then assign this "line formation" as a possible order to that type of army. There might be several such patterns that can be switched between by that force, all customized by the player.

Drag select a group of units, shift-right click first waypoint of plan, done. Doesn't seem too tedious or complex to me. Maybe hit ctrl-1 so you can easily select the same group again later, or not if the plan you've assigned them to is a simple 'go die in the foreverwar'.

I think we can break it down like this:

Grouping factories into big units - excellent idea. In fact, necessary. They should improve on Supcoms system. You should be able to select all factories at least on the planetary scale with one click.

Grouping units into big units - great idea. The reason Total War doesn't become micro hell with thousands of units is because you select them as groups, so you only have 20 to manage. PA absolutely needs something like that to make the scale bearable. I also like the idea of auto-reinforcing unit groups.

Forcing players to rely on the AI - bad idea. Hearts of Iron 3 is the nightmare example. The issue is that an RTS AI can never outsmart a half decent player, and probably not for the next 20 years. So then the players have the 'choice' of relying on the AI and dying in the game, or doing everything themselves and dying of exhaustion IRL. I like everything that streamlines the interface, but it gets risky when you bring the AI into it. I honestly doubt even Uber can design an AI that knows how to capture planets and so on. Instead of trying to teach the AI to micro, why not just design the game so that less micro is required?

My thinking is that we want to have an "AI" that is a product of the player creating rules and patterns one time, and then being able to copy those behaviors in other situations. So while the computer is issuing the orders, it is doing what you told it to.

Rather than needing to issue a lot of similar orders at different points in time, you set up that behavior once, and reapply it with relatively few clicks or keypresses.

Suppose I wanted to increase the spacing between units in my army. Rather than manually grabbing and splitting units every single time I wanted to do this, I create a formation that is split, and order the units to assume it with a single action.

This is exactly what I was going to say. This is how I want the game to play out. As I heard the game is about "executing your masterplan" and not about endless micromanagment. So it feels perfectly resonable that they AI can handle what they are supposed to do without being total retards if you are not in direct command.
For example if the west side of a base is attacked. The base would mobilize its defences to defend against the threat. Not like many other RTS where only units that can directly see an enemy attack them to defend.

That sounds perfect.

This kind of command and control is exactly what I'd want from the next step in RTS user interface.

As an extension of the production suggestions above, I would much prefer being able to issue production orders by opening a build menu from the UI and dragging a building or template of buildings onto the map, and letting the AI find a nearby construction unit to add the order to its build queue. Likewise, I love the idea of selecting a base and issuing construction orders for template-groups of units, and having the base work out which factories should build what. Micro-ing factories and construction units is not what I want to spend my time doing. By issuing orders from global menus or to bases (instead of individual structures or construction units) I can manage my pipeline of violent little robots with ease, and then gleefully throw them into battle.

Excellent posts. This is exactly why I made that monster post detailing how and why we need a powerful and customizable unit AI to work with.

I'm perfectly fine with killing mandatory micro. A flexible unit AI will always be inferior to a human, but a game of this scope needs as many ways to minimize micro as much as possible. Humans are not natural multitaskers and research has shown that even those who say they're good at multitasking will perform both tasks considerably worse than expected. A strategic game like this that involved lots of micro would overwhelm almost everybody eventually, and being overloaded will end up degrading the game into just giant blobs of homogenous units, simply because it's the easiest way of handling things.

Making sure your artillery is spaced out properly and not advancing into the range of defenses like idiots, making sure the AA units are covering the artillery properly, but are moving around a little to protect the units that *are* under air attack, and making sure those units aren't just standing still getting chewed up is all a lot of work. It becomes exponentially more work when you have this crap happening on two different maps. If this is the case, and we can't trust the AI to attempt to act intelligently for us, then people will probably just end up building a lot of T2 tanks and shoving them at the enemy and saying 'screw it' because they have more important things to worry about.

AI programming is incredibly hard, yes, but I think seeing some advanced features and semi-intelligent behavior on a per-unit, per-squad, and per-group basis would befit a game made in this day and age, especially one set on this scale. This game shouldn't be trying to be the RTS it's not, and I see too many people making arguments for micro, as if this game is going to be the next Starcraft-killer and having huge skill ceilings is necessary for it to succeed as an e-sport.

I completely agree with the last post. You aren't fighting one battle, highly zoomed in, where this micro is possible. No, you might be fighting a prolonged campaign on one world, holding a choke point on another, and setting up a moon base as fast as you can in the meantime.

A great RTS, to me, is a strategy game; I want to draw up plans and see all the pieces fall in place. That, however, is not what most RTS'es are, which is too bad. I like RTS'es, don't get me wrong, but this idea is something 'new', and it'd be sad if this game, so full of potentional, still isn't a 'true' strategy game, but just a normal RTS with a massively increased scale.

I completely agree with the spirit of this thread. The focus should be on grand strategy, but all the details should be your own design. The computer doesn't decide for you; it is there to execute your orders and free you from tedium so you can think strategy. All the while letting you micro any part of it if you really wanted to.

It is a very tricky thing to make such a UI both powerful and smooth, so I don't have any suggestions for that right now. I do want to point out that it ought to integrate multiplayer seamlessly. That is, it should support scenarios where some or all of those "AI sub-commanders" are human players.

There are (imo, acceptable) downsides to this approach: micro-focused players will become annoyed when the endgame armies become so large that no amount of ubermicro will succeed. Some may see this as intended but, this being a game, the ultimate goal should be about providing fun to everyone.

Not everyone. Target audience is extremely important, especially since they're going to push this out the door extremely quickly. The initial version will have no campaign and only one faction, for example -- this'll turn some people off, but containing scope-creep is extremely important for getting something complete on-time and in-budget.

To avoid derailing the thread, let's talk about strategic UX. My first thought was to take advantage of the close-in-space relationship of units at each level, but I'm not sure how well that would work at the highest scales.

It should almost certainly be an aspect of "strategic zoom", though. At the highest zoom level, you see individual unit models, and control is given at that scale.

As you zoom out, the individual models for a platoon merge into one symbol to represent the group -- perhaps at the center of gravity of the group itself, scaled for clicking.

Zoom out further, and a company of platoons gets one icon, and so forth. Probably only need three levels (again, tactical, operational, and strategic), which would simplify many things.

Commands given to icons at each level could differ: a the tactical level are commands that are particular to an individual unit. Special abilities, etc.

The operational level shows supply lines, zones of control for your armies (as an abstraction), and has commands like attack, move, defend, etc. Units should get resupplied automatically unless the player basically turns of the auto cast (at operational level, perhaps).

At the strategic level are commands that set objectives and constraints: take this continent, but avoid this island, spies report heavy AA in that region. I see this being somewhat like Battlefield's commander mode, or the analogous mode in Tribes.

In this way, you should be able to smoothly zoom through multiple levels of commands, extending the metaphor of strategic zoom and making it much more powerful in the UX sense. I don't envy them the task of extending it to extra degrees of freedom (for spherical planets), though...

In the concept above, I figure that experienced players will spend most of their time in the operational level, with a bunch of HUD information: zones of control for their units, lines of attack, lines of supply, etc. (I envision it like they usually show these things in video: http://www.youtube.com/watch?v=Lcn4zsewnHQ#t=0m23)

Example of play:

I start the game with my commander unit (who appears as his own icon in any level of detail), and [operational] begin constructing my base. I zoom out, [strategic] getting an idea of the system I'll be dealing with this game -- how many planets, where the asteroid field is, what kind of planets are available. I don't know where my enemy is, though, so that's a problem.

I zoom back in as my land factory finishes, and [operational] begin creating a scout force. [strategic] I tell it to explore the planet, but stay out of sight: don't engage enemy forces if possible.

Unfortunately, one of my scouts is spotted by the enemy (I probably get an alert for this, as the scout's objective has been compromised). I zoom in to investigate [tactical] and see that the enemy has begin constructing tanks out of a small base on the opposite end of the planet from me. I hot key back to my main base and [operational] construct some defensive units while I get my commander working on a launch platform so I can get better intel.

Fast-forward: my units meet his in battle. I want to [strategic] force him to give up some territory he took closer to my base, so I [operational] split my units into two armies to bypass the front lines and cut off his resupply. I need a pinning force, so he doesn't out-flank me, but I don't have enough units. I take a small group and zoom into tactical view to hopefully make some maneuvers that will make it seem much larger, focusing my opponent's attention.

----

That's how I see this going: your attention will most often be operational. You must zoom out to strategic in order to have multiple large campaigns going at once, an extremely useful skill. You must zoom down to tactical to do anything out of the ordinary, also a useful skill. Someone who spends all their time at one level will have mediocre lieutenants below them, and nobody above, but players with little experience could potentially play almost the entire game in strategic against each other and still have fun.

To extend this to teamed multiplayer, you could totally just implement shared unit control and rely on the social contract to keep people with particular units.

No offense hackerblinks but I fear your idea would make the game wayyy too complicated. It also seems to overestimate the ability of current AI to make good decisions. Which is an issue with several ideas in this thread: Some of you guys don't seem to understand just how terrible RTS AI is. If you rely on an AI to make tactical decisions there is no way they will do it as well as a competent player. And there's the problem (one of many) with Hearts of Iron 3: delegating decisions to the AI, which was supposed to make the game simpler, has in fact made it that much more of a headache, because in order to play optimally you have to override all of the AI's decisions, which the game was not designed for, and thus you're stuck with more micro than before!

In my opinion, this thread has some excellent ideas and some flawed ones. The best ones are unit groups and plans. That is, grouping units (and factories) into mega-units, issuing them orders as one; and grouping command sequences into plans and saving them for later, 'Rainbow Six' style, then executing them with one click. Those are pure gold and should be put in the design doc immediately! However when it comes to AI commands for complex actions: "Explore", "Invade", etc, sorry to say I'm doubtful. It would of course be awesome if they could pull it off, but... Then again, maybe I misunderstood your proposal, sorry if this is the case.

I agree that it is unreasonable to expect an independently intelligent AI to do what the player wants. However such an AI is unnecessary. The only thing the AI needs to be able to do is remember orders, and duplicate them when asked.

From a strategic standpoint, let's suppose I want to take an island chain WWII style. I give the order "construct an airbase here, on that island."

The independent AI approach would be to have the AI have a pre-programmed algorithm for that type of command, and intelligently figure out the best way to do that. This is bad because it is a lot of work for the developer, and the AI typically doesn't do what the player wants anyway.

The correct way to do this is to have the player define what an airbase is. So I define an airbase as having 3 runways, a fusion, two air factories, a land factory, defenses X, Y, and Z, and is stocked with 40 fighters and 40 bombers, etc. etc. All the AI has to do is exactly match the pattern I already defined.

I might even pre-specify the relations these things would have to one another. I am ordering the creation of a group at the target location, which can have internal logic I specified as well as units and structures. So my factories could have build orders to fill the base's garrison, any fighters in that garrison might automatically attack any air units detected on radar, and so on.

The AI I suggest is actually really, really stupid. Its only new trick is that it can remember commands, and recall them when ordered to do so.

Exactly. I agree entirely. It just seems like some posters agree with your suggestion without fully understanding what you were saying.