Defining micro

That makes no sense to me. The depth of your decision making is precisely that control you have over your units by telling them to do specific movements. Letting an AI do these things takes away the depth of the whole thing, cause they AI does everything for you. Or at least it will try. To control a situation in-depth you need to totally focus on it and make sure that everything works exactly as you intent it. An AI that hopefully does what you want does not help at all.
Also to minimized the punishment you can just tell your ship to run away for a while, so you can go somewhere else, place a building and return. In fact thats what good players are doing. You will lose a bit of space, but thats about it. If you forget about giving that move command, you might as well forget to enabled the unit ai.

Like I wrote: The players time is a resource here, thats why it is called RTS in the first place.

Another bit of thought:
Lets consider a spam-fight on tech 1 in SupCom on a wide open map.
Aeon vs Cybran. The Aeon t1 tank is a unit that scales with good kiting. In fact if you control it perfectly there is not much the cybrans t1 tanks can do. Vs a player who is able to control the aeon tanks good enough you pretty much need to go t2. -or just build more tanks, but lets assume equally good builds here- The control of the t1 aeon tanks is so hard though, that only few people can do it good enough. It is a large matter of skill. If we were to add in an ai that is able to do the micro as good as a good player in most situations we would need to nerf the aeon t1 tank. Otherwise it would just kill cybran/uef/sera t1 tanks most of the times. However if we were to nerf the tanks they would totally rely on the good micro of the ai. AI's are never perfect, so it might fail in some more complex cases. Since the tanks are nerfed to be balanced in the normal kiting-fights they would die hard in case of any ai glitch.
The endresult of adding the ai therefore is:
-We took away a big matter of pure skill
-We replaced it by an ai that sometimes makes stupid mistakes that might be hard to predict, therefore introducing a simple matter of luck: "Will the ai be able to handle it?"

I cant see how adding a kite-ai would improve the gameplay in this scenario in any way.

While it is indeed an example of player skill to do such actions as kiting with a few tanks, that is a very low-level, small, and to be honest not very interesting application of player skill. Moreover, if you have hundreds of tanks scattered over several planets, whether or not you can micro them to increase their effectiveness becomes increasingly irrelevant.

The scale of the game changes the fundamentals of the design of the game. This should be obvious. On a huge map, with many small forces moving about independently, the player should be concerned about composition of armies, and where they are, rather than how effectively each one is fighting at that moment. Simply due to the sheer scale, the AI must be able to handle low-level combat independently. Even if the units can be microed to increase their effectiveness, the player cannot, will not, and should not be expected to micromanage every fight.

The player should make bigger decisions about the types of armies they build, and where those armies move on the planet map. Not managing a single squad's combat maneuvers. However, the player might specify how all squads of that kind fight, and that definition will automate all those squads' behavior.

You are using the scale of the game as an excuse to condemn any kind of unitcontrol to be done by an ai. However I dont see how the scale of the game will be that much bigger than what we already have seen in SupCom. And SupCom works very well without any kind of automatic unit-control. Kiting might be low-level control, but it has nothing to do with controlling small amounts of units. The units pathfinding should be good enough to make it possible to kite with an army of 100 units by telling them to move back. That has nothing to do with the kind "micro all units individually" that you are imagining.

If you compare PA-fundamentals to Starcraft-fundamentals: Yes.
If you compare PA-fundamentals to SupCom-fundamentals: No.

SupCom-fundamentals already dont require you to control each unit as an individual.
You usually zoom out and tell your big army: move back a bit. Move forward again, move back again. You don't tell that to single units, you tell it to your whole army of 50+

You are not expected to do this. Why do you think this? Just because your units dont have an "I do it all for you"-ai does not mean that they will need your attention the whole time. They do shoot at the opponents units without you telling them to do so.

When I said "depth of decision", I meant the long term impact of each decision, with some value attributed to how trivial the decision is. When I say "micro", I refer to all decisions below a certain threshold of impact.
- By trivial I mean how easy the problem is to 'solve'. If there is a clear optimal solution in every circumstance, I call that trivial. There is no need to think, just do the optimal thing every time.

For example, I would consider telling units to take ten steps back "micro" unless part of some larger gambit (very often not), while deciding that a (known) area cloaker building *isn't* hiding a bunch of enemy units (the actual decision is to ignore it) is not micro and is the level of decision the player should be implementing.
Telling units to go stand on a nearby hill is somewhere among the lowest level of decisions I see as for the player to make.

As an extreme example, I presume you know about mass fabricators in Supreme Commander yes? When they were still useful in pre-FA, building mass fabricators meant that you have to watch your energy consumption to avoid stalling (stalling means you lose radar and shields = BAD).
BA in spring has a bunch of widgets along the lines of Improved Metal Maker. They automatically turn on/off your MMs to balance your energy (assuming that it is even possible of course)
So, while it has lots of impact, the decision of "should I turn this MM off/on" is trivially easy to solve (balance E at user defined level of storage).
I'm curious, what's your opinion on this? One could turn BA into a game of who can click their MM and press X to toggle, since MMs are a pretty big part of say... 8v8 DSD (vets please don't kill me)


The difference between giving a move command and having a Unit AI toggle (literally, there is a toggle for Unit AI on/off in ZK) is that one is active, the other is passive. If you have a unit AI toggle, you pre set it into ON and know you can leave the unit lone for a bit. Sometimes, even 1 second difference can spell the end of a unit after a short chase or a successful escape.


By implementing unit AI to automatically manage this, you make trivial decisions and low level decisions scale much much better than they normally would.

One of the APM challenges I try sometimes is 1v1AI on the map Melange in ZK. Mostly flat humongous map, regularly broken by spiky rock mountains and sand dunes.
Radar coverage is hard to get (radar in spring is blocked by terrain), los even harder. Most land units trying to get places takes quite a lot of time and even air units take significant time to cross the map.
Playing against the AI from opposite corners becomes an APM hell of placing four expansions, light defenses, intercepting raids from six separate advances, managing your own raids, aircraft and bloody everything at the same time.

This would have been impossible to play without unit AI.

This sort of scale is what I expected from a multiworld PA map.

Mass fabricators were nerfed hugely in Forged Alliance in order to stop the repetitive, Sim City-style play that they encouraged. In FA the only valid time to build mass fabs is when all your mexes are T3 and you have a surfeit of energy.

And even in that rare situation, you click one mass fab once, press Ctrl-Z (select all) then Ctrl-1 (assign group). Then any time your energy starts falling, press 1 and click pause.

Just to clarify my position here, I'm not concerned that PA will play like Starcraft. It's fairly clear what kind of RTS the PA devs want to build, and I expect it to be a lot closer to a TA/FA style implementation on a lower level.

I also don't have that big of a problem with Starcraft as a game, it's just not the type of game I consider to be truly strategic. As such, I tend to use it as an extreme example, since it is a very micro focused game.

That said, while Supcom was designed to be fairly low in micro, there were certain aspects that benefited from manual control, such as artillery usage. But it's also important to remember that PA is being developed to play more like TA than Supcom (if I understand the devs correctly), so I'm assuming there will be a bit more unit variety and more specific unit roles than there were in Supcom. This means that PA may end up requiring more attention to be paid in close up unit engagements.

While unit variety will make for a much more fun and dynamic game, it means there's a potential that unit micro will become more prevalent. I've already explained why I'm against the concept of micro in an RTS, but I do think that a tactical level for decision making is just fine for this type of game. For the most part, that's as close to micro that Supcom got to, with the exception of commander usage and certain, specific units.

One of the better solutions I see to the potential micro issue in PA is the whole concept of advanced attack-move, and Zero K is a good proof of concept for how it would be implemented, which is why I strongly suggested using it as an example. Whether this solution is adopted for PA or not doesn't matter, as long as the end result is minimal micro in PA, so the player can focus on the more important aspects of the fight.

The problem here is that our views on what is trivial differs. I dont think the micro you are talking about is trivial. In the end everything could be seen as trivial, so why not let an ai play the game for us? The question is where do we set this "trivial" threshold.

There is an ui-mod for Supreme Commander that is doing this automation. I never used it, since I didnt played that much Eco-games in SupCom and FA doesnt need mass conversion anymore.

I am not sure what to think anymore. An AI that is as good as a good player is definitely not what I want. The need to control your units well is an important factor of the game. However an AI that just helps out a bit while the player is doing something else might be not too bad. It should be pretty limited in what it does, though. I DONT want my units to run away a mile from the position I have told them to move to. I do have plans for the units at that position and I dont want the ai to interfere with it in any way.

Against kiting, could be convinced of jinking, do not want reactionary dodging. Automated eco micro is totally fine, as eco is boring to micro and finally want unit roles.

Thinking about mass fabs, I wouldn't be against an auto-manager in principle. The only argument against it would be one of arbitrariness: if mass fabs turn off to save energy, why not radar?

Assuming an auto mass fab manager was included, I think that the least arbitrary and most logical implementation would have it apply to all economy buildings. In practise this would just mean that mexes were also be turned off if energy storage reached zero. As long as the priority was set to turning off mass fabs first, and to only turn off mexes if energy was still a problem then this would work nicely.

That's simple, if I can save some time from kiting, I get more time to think about what kind of defense structures I should build.

I don't know why you are talking about nerfing, if the game has automation since the begining, its balance would be built around it since the begining, how powerful the units would be partly rely on how effective their AI would be, so more mistakes their AI would make, more powerful they would be.
And in what cases auto-kiting would make mistakes? If there are enemy units behind your auto kiting ranged units, which is a rare case, then you would only need a move command to change the kiting direction, automation makes micro simpler even if it can't remove it completely.

Everything could be seen as not trivial too, like the strafing of the gunships.

Thats part of your strategy and general game knowledge. If you think about what kind of defence is best in the current situation you are doing it wrong. You are ought to know it beforehand. That is one of the things that make a good player good.

The strafing gunships are just a good example of an ai-feature I would totally not miss at all.
You just have to work against it so often cause they end up in the range of aa or move into a position from which they cant hit their targets anymore.
I am basically just worried that the other ai-feature will be just as bad.
Kiting and correct unit-control is just not trivial enough to make an ai that is really good for it. The AI will definitely fail sometimes and in such moments you might lose cause of an error some ai made. Thats not a situation I want to be in.

Why do you assume that AI control is mandatory? That would be awful, personally I set all my units to hold position because I don't want them moving when I don't tell them to.

If a good player can find out what is the best choice in a really short time, that just means the game lacks depth in strategies, some board games like the Go might need the player to take hours to decide what to do in the next step.

I guess the most players would rather having them making mistakes than have to manual control them to dodge some slow projectiles.

Regarding gunship strafing, I've never interpreted it as an attempt to dodge. If that's the intention then it patently doesn't work, I've never seen AA miss a strafing gunship, ever.

I've always assumed that it's solely for aesthetic and “realism” reasons.

absolutely. it had nothing to do with dodging, and everything to do with the aesthetics of a hovering unit.

It might be useless in SupCom, but if the gunships didn't strafe in Total Annihilation, even the rocket luanchers could hit them.

Thats a very bad comparison. Games like Go or Chess are turn based and actually have a very limited number of things that can happen. An AI can also easily beat any human in these games. And they are TURN-BASED. PA is an RTS. You have to be fast playing it. If you want no APM pressure at all you might be better of with some turn based game. Also the requirement to give a command in a very short amount of time has nothing to do with the strategic depth of the game. You can play chess with a timelimit of 5 seconds per turn. Sure humans will end up doing stupid mistakes, but it doesnt change anything about the strategic depth of the game. An AI could easily still give perfect moves within 5seconds and really good human players will still think about lots of strategic points within these 5 seconds before doing their move.

love all your posts CC but i'll think you'll find there hasn't been a Go AI that can beat competent humans.

Really? Wow, I have to admit all I know about Go is: "Thats some board game that looks not that hard for computers" After a bit of wikipedia it turns out I was wrong and it is too complex even for supercomputers so far.
While this is interesting it doesnt change that much: If you were to play Go with a limit of 5sec per turn it would still be a very complex game.

There are some non-mainstream 30s rules for Go, but they usually give the players some additional chances to wait one or two minutes, Go with 5s hard limit might become quite boring.
It doesn't matter if a game is complex or not, it won't be very strategic as long as the players don't have enough time for planning, you can imagine supcom with dune 2 like UI and automation to understand this.