What if micro pigs... are just shaved guinea pigs?

Yeah but what your describing is a full 'mico AI' system that can:

A: Pick a target
B: Auto manoeuvre the unit in relation to that target
C: Prioritize when there are more than 1 possible target in the area.

That's a bit more than a simple tool that 'automates clicking left and right'....

Still not so difficult. Simple optimization by multiple goals. You know where the shots are going to hit the moment they are fired, so you can penalize being in a certain spot at a certain future time.

Striving for the location of minimal enemy DPS (kiting) is also simple, not different from what PA does itself with the terrain cost map and a simple single source, single destination path search.

A days worth of work, at most.

It's even better than if a human did it, because a human tends to do it chaotic. Works still well for individual units, but with larger amounts of units you need to time the evasion so you can mislead the projectiles first, otherwise the chaotic results could make it impossible to evade.

Image recognition makes it difficult, see my previous post for a rambling explanation. I'll try and bullet-point it better here:

• Determining what a dox is is hard, unless you have strat icons visible always
• Always visible strat icons are also needed to determine what is yours and what is theirs
• Determining that that dot is a bullet is tricky, there's no strat icons to help
  
• Determining trajectory requires at least two screen captures
  • Multiple projectiles in close proximity complicates things
  • 3D complicates things. The trajectory is either parallel to the plane of the camera, or the bullet is moving relatively quicker at some other trajectory. Variable
    
  • Not knowing what fired the projectile complicates things. You should find out what fired to know if it's a threat, and how far the projectile can travel, and what layers it can fire at.
  • Figuring out the unit that fired may be tricky
    • You don't know how long the bullet has lived (you have to 'rewind' a variable amount)
      
    • The unit firing may have been moving (need to keep track of units too)
      
    • There may be many potential units that could have fired the projectile (a group of units in close proximity complicates things)
  • Figuring out what was fired at is similarly tricky, bonus points for target leading
• Any projectile meaningfully affected by physics complicates things
• Full knowledge of all projectiles and units needs to be accounted for, splash damage etc
  
• The camera complicates things. Zoom orient pan
• 3D terrain and obstacles complicates things, particularly buildings
  
• More hurdles will pop up if thought about hard enough

In simple cases you may be able to make something that works; for example restricting it to one friendly target, one unit type, no camera movement, swapping bullet textures for easier screen reading, etc. Which I'd still be impressed with and hope to see someone try from a technical pov.

No one would be stupid enough do to it with image recognition. That's at least what I hope. There is no need to intercept DX draw calls either, like the SC2 bots are doing it. PA is much more verbose about projectiles and units, if you know where to look. Once you have the projetile in world coordinates, you also know precisely where it is going and when it is going to arrive (for all non-tracking projectiles, of course it won't work for missiles.)

If you are modding on the server, maybe you can find live data on projectiles and in world coordinates. From the client, you cannot do this, short of hacking away at the executable (you don't even have live data for where units are in the standard mod environment, all I know of are one-time things like alerts there). If you've done something I'm unaware of and can provide more information, please post it.

exterminans: "This thing easy"

experienced players and modders "This thing hard"

...

Should probably stop replying trialq

You're just working based on different assumptions. With the current API it is pretty hard, but with the API that we wish for it would be easy.

And the netcode isn't so complex either when you know how to use a disassembler. Especially with the Linux build.

I think this kind of micro should stay in the game.

It's a neat little trick that helps you to outplay your opponents in the very early game if he doesn't know about it. Late game it isn't possible anymore and as long as it stays that way im ok with it.

Has anybody tried if you can get the same effect by simply rallying the dox on two points next to each other?
And if you have two dox getting microed that way are they just batteling to infinity?

And about the discussion with the mods and unfair adventages:

The only ones who should be able to judge if a mod is allowed or not are the devs. If a mod gives players unfair advantages over others, I dont think it should be in the game. But if a mod, like the hotkey mods which are currently out there is just helping a player or provides features which are currently missing in the game thats ok.

I don't mind it. Doing the micro takes time from your macro stuff. So if you micro a lot, someone else might be already building halleys to come and greet you