avoiding micro-management

In Zero-K with gadgets you can avoid many (at least most annoying) micro management things. Of course, in battle you still have to micro (though units usually choose their targets in a quite smart way), but in base-building, it's very good.

Most shining part: idle builders will help whatever is happening around. They will repair units, help building and other things. If a constructor unit is set to roam (follow enemy), it will help quite far around himself, and even automatically reclaim wrecks and map features (trees, stones etc.). In original TA, you could do this with patrolling command, but then builder isn't marked as idle.

there are other things I'm missing:
1. great concepts:
* area commands. I select a builder, click repair, and select area on the map. It get queued repair orders for all damaged units in the area. Same with reclaim, and by default they won't reclaim units, only wrecks.
Same with attack.
* micro-AI for units. If a constructor sees an enemy, it will start retreating to safe location (and get back when there's no more enemy. Moreover, it won't retreat if it executes direct command, e.g. it wil retreat only if idle or executing some macro-command). Small fast units will auto maneuver to avoid enemy shots (strafe, randomly change direction, etc.). Also, constructors will auto-reclaim any wrecks blocking their building sight (especially useful for re-building destroyed mexes).
* repeat mode. For both units and factories. Finished commands added to end of queue. Especially useful with macro commands.
* auto-retreat on low health. Mostly useful with big and costy units, to avoid granting free metal to an enemy. Retreat location must be specified explicitly, there can be many of them, and units will get to that location once their heath drop below threshold.

2. macro commands:
* One thing will be very useful in PA: command to build mexes in an area. I just click Ctrl+W, click and drag (select an area), and selected constructors will build mexes in the area using optimal path. If a spot is already taken, it's skipped. (though it's not built-in ZK feature, it's LUA widget for Spring Engine, and it works with all Spring games which have metal spots)
* formations. Right-click and drag a line, and units will get in the line. If required (there's not enough space), they'll make several rows. Want to recon and area? Form a line of scouts and send them in that line.
* Fight command. Units will move but start attacking any encountered enemy (this includes deviating from the route). This is must-have I think.



Without these things, I'm getting really sad and annoyed playing PA, and getting enough of this for one play.

I'm all for macro commands especially for mexes and reclamation, but auto-assisting may not be good if you are stalling or starving for metal.
Auto-retreat would kinda be annoying if you are the attacker and might make turtling even more viable - I think this would take a bit too much micro from the game.
Repeat mode is planned and so are formations.
Fight command is know as "roam". There is a key for it in the keybinder, but I'm not sure if it works yet.


AS for the pool. I voted Good, but only to some degree. Minimal micromanagement is something I would not like to see abolished completely, but we definitely dont need another StarCraft2 here.

It can be smart. It can stop assisting in case of stalling.

First, it must be configurable, as in ZK. There you set threshold for each unit, and ofc you shouldn't set threshold for units going in attack. Second, it can be smart also, for example, work only if a unit isn't selected (so it won't interfere with attacking).

Also, It will allow to attack two or three front simultaneously: on 2 fronts just throw units there (with high probability they'll stay alive even if won't penetrate defence), and on 1 front you doing micro.

Very good about repeat and formations. About roam – is it separate command? Sounds like Move options: hold position/maneuver/roam. It's very useful, but not the same as Fight command, which is move Move until enemy met.

Good. Micro won't go away, even Dungeon Keeper had some. Manual units control in unavoidable at current gaming tech levels, unless explicitly banned. There's just no way to explain units what you want to do beforehand, so control is a must. And units are quite stupid by themselves.


BTW, I'm from Realm also, but PlanetSide 2 mostly

I've never played Zero-K, but you guys make me want to try.

You might as well. Its free and comes with some of the best UI stuff I have ever seen.

I've noticed that placing a new metal extractor where a previous one was destroyed takes to much micro. First you have to remove the wreckage then you have to build your metal extractor. You cannot chain these commands with shift because it is on the same spot.

It would be nice if the reclaiming of the wreckage happend automatically.

ghost, I think that's getting fixed in the upcoming patch.

Just remember that too much automation can be annoying for the players. For example by default idle static builders help what is around them and mobile builders do not. The initial state for each unit type can be toggled with a setting and it can be set for each unit individually with a state. The same can be done for the automatic retreat state.

This
Some automated things might be really cool. Some others might be really annoying. PA currently definitely can use a lot more automation features, but it should not go too far.
We had the discussion already. Personally I don't want units to make any decisions on their own. They should be stupid robots that only move when told.

The wreck on my thing btw I think really is just a missing feature in general: You cant place buildings on top of wrecks at all. I am pretty sure that will change.

Automated elements which don't impact my economy (repairing is free?) are good, but otherwise need to be left to the player.

But moving beyond automation are things like retaking MEX points going to be simplified? The inability to queue a MEX on a spot with a wreck on it is bad. Either I should be able to queue the MEX and the reclamation is done as part of that, or a MEX shouldn't leave wreckage ala TA.

I'm not saying don't add this automation. Just that it needs to be easy to not use. The most invasive automation should be off by default otherwise people will lose their units as they decide to wander off and they decide what to do for themselves. Automation has to follow the desires of the player using it otherwise it's something they are likely to fight against.

I think the best situation for automation is for the engine to provide the framework without actually doing very much which is visible to users. Automation is best left to UI scripts. The scripts just need to be powerful to do and see everything a player could. Of course many of the initial scripts would be created by Uber to ensure that PA initially has a reasonable UI. But I think the powerful scripting system should come first and the automation later.

I say this from my experience with the history of widgets in Spring. With fairly invasive automation it is hard to know how to implement the idea such that it is easy to toggle and doesn't take too many stupid actions. It is likely that the first pass on any automation such as this would be unusable and disabled by most people. If it's hardcoded into the engine that is the end of it. If it is a script hen many people can fiddle with until someone works out how to make it work or abandons the idea entirely. There is also the possibility for any work put into hardcoded automation to be wasted as a better script based implementation is developed. Spring has a formation system hardcoded into it but nobody uses it because the drag-line-move is much better. Finally there are too many automation ideas flying around for Uber to try out all of them so it is much more efficient if the community does it for them.

I think there's a huge gray area for micro, and I'm not talking about specific actions given to units.

The enemy AI in FA made by Sorian is designed to learn and adapt right? It's supposed to be aggressive when you're turtling, and defensive when you're the one on the hunt. I've seen posts here about other AIs in one of the C&C games where it learned by adapting the player's strategies and beating him with it. (Something about sell-rushing the command center with engineers in a Subterranean APC, and the AI doing it to them the very next game.)

The AI having a learning curve also appears in 2D fighters where it watches and tracks all of the player's movements and their speed of keypressing and uses that to try and match an enemy AI to give the player more of an even fight that could sort of give a feel of actual intelligence.

I think the micro AI in PA should do the exact same thing. It should be able to not only micro everything you want it to, but naturally learn what you do so often and take care of it for you. It should also learn what you want to do manually. I don't mean building structures or whatnot, but I do think any miniscule micro like repairing defensive structures when the enemy is attacking if your eco isn't stalling (or if repair doesn't cost anything), and such.

A great example of this would be the AI learning your priority decisions when an enemy force comes rolling into range. If a player tends to use arty to take down the AA and then send bombers to disintegrate whatever's left, the arty should take that role if the player has taken a notice and is taking action against the incoming force. Otherwise it should have a natural priority checklist if the player isn't watching, likely any large swarms of Levelers first, then either any remaining levelers or the large ball of Ants depending on the defenses and situation.

A micro AI should have much more use than simply building structures and units for you. It should do the things you worry about but don't have time to take care of.

What? Where did you get that idea? His FA AI is great, much better than stock, but it was still limited by what he could mod on the AI, he did create multiple difficulties and an add on for the lobby that gave players more control over the AIs Cheating options.

Mike

Am I the only one that actually read the descriptions written for the AI? The very definitions the AI description gives when you actually mouse over the different types of AI tells you this. This was before I got FAF and one of the semi older generations of AI and AIx. One was made for naval, one for air, another for turtling, yet another for rushing, and the adaptive.

You cannot tell me that's impossible to do, since the AI for PA right now is already learning. It avoids defenses, pulls back, waits for more units, then attacks with a force large enough to break. I don't think it's impossible for it to do what I suggested.

I wonder if that's true. Any post from sorian about it?

I don't know of any post directly from Sorian about it, but the update notes in-game and on the site state the AI was improved, but that thread has several posts on the last couple of pages about the AI improvements, and even one about my run-in.

If you don't believe me, just go into a game with a few AI and just turtle behind T3 defenses and walls while leaving a large hole. You can easily watch what the AI does with the advanced orbital radar. I can always try to find my last replay so you can spectate my clear prowess in APM (read: severe lack), though I don't actually know where they're at.

This was my post:

And here's the post about C&C Tiberian Sun:

Automation should be player defying and customizable. The user should have the final say on that.

Yes, there were different types but that is not the same as 'learning'. Even the 'Adaptive' didn't learn, all it did was pick the best Type for whatever the map was.

Mike

I've had dozens of experiences where the AI, whatever one it was, has made about-faces and went completely defensive instead of offensive. The Adaptive simply picking a map AI is probably true, but the AI itself still changed tactics during the game depending on what I did (or didn't do) to it, and I would consider that as learning as you can get for a modded AI that didn't function exactly as the modder wanted.

Irrelevant of the SupCom AI, the ability for AI in general to 'learn' is not impossible, and I'd like to know what you think of what I've stated since you've commented on nothing else.

Just to get it out of the way, I am aware that AI are only based on core programming, and cannot be asked to think for themselves as a sentient being would. Nothing I suggested is defined by real learning from actual intelligence.

Completely agree with this. The user should always have final say, which is why I'd like the AI to not only define what the user does often, but also define what the user wants to do manually and/or simply things the user just doesn't want the AI to do at all, so there can be a better flow of unit to player interaction as well as the opposite instead of only a player to unit interaction which is what we have in RTS now.

You and I have different definitions of 'learning' then, to me everything you describe isn't learning, it's simply reacting to what you're doing. Learning to me is when you give the AI the ability to make judgment calls and look through it's prior match history to guess what you'll be.

Mike