The early game

[TechSY730]

So that is what focus fire does. It never was very clear by the tooltip.

Anyways, I guess that the preferred behavior for auto-targeting is already in place, its just that auto-targetting is so paced (that is, limited by the amount of time spent on it each frame), the non auto-premptability of auto-chosen targets, and ships can be so short lived in battle, that we rarely get to see great benefits of this behavior.

Its one of those things that the CPU is so stressed out, it can't fully carry out the auto-micromanagement that it could do given enough time. In these cases, you really have no choice but to step in and micro some yourself. Either the CPU has to step it up or you do, and the CPU is already near its limits.

[keith.lamothe]

Well, really it should be doing fine as-is.  If fewer than 50 eligible targets are in range it considers all eligible targets on the planet so it won't generally get "stuck" on a suboptimal-that-came-in-range-first very long.  Also, with aggregate targeting it often doesn't actually have to do very many full scan-and-sort operations; if all your fighter mkIs are relatively close together they just all use the same base list (without focus fire they spread out a bit, but it's the same set of targets).

If someone can come up with a reproducible case where the autotargeting is being highly sub-optimal, I'd like to see the save and steps to reproduce.

[BobTheJanitor]

Not really, focus-fire actually computes "how many times would I have to shoot this for it to die" for each eligible target, and sorts them in ascending order of that (this counts the partial-shots too, though those are not realistically different than a whole shot).  There are other considerations that can override, but that's probably the operative one in most cases.

So then what is non-focus-fire doing? Does it consider all the friendly ships in an area and all the enemy ships in the area and decide to assign them each different targets based on what would be the quickest method to make everything in firing range dead? One seems to be described as 'faster' while the other is 'more efficient' and trying to figure out how those terms differ in this context is making my head hurt. 

[TechSY730]

So then what is non-focus-fire doing? Does it consider all the friendly ships in an area and all the enemy ships in the area and decide to assign them each different targets based on what would be the quickest method to make everything in firing range dead? One seems to be described as 'faster' while the other is 'more efficient' and trying to figure out how those terms differ in this context is making my head hurt. 

Well, maybe faster and more efficient are bad terms. Maybe a better description would be prioritizing maximization of kills per second vs prioritizing maximization of damage per second would be a better way to think about it. Paradoxically, in many cases those two goals do not lead to the same result.

[Draco18s]

If someone can come up with a reproducible case where the autotargeting is being highly sub-optimal, I'd like to see the save and steps to reproduce.

I don't think I have a save (would have to dig through), but I did end up group-moving a fleet blob around a system and the frigates decided that some target under a shield sphere was a priority kill and WOULD NOT MOVE OFF until those targets were dead (despite having never received an explicit attack order).  Every other ship kept flying around the system killing with impunity.

Does that count?

[TechSY730]

If someone can come up with a reproducible case where the autotargeting is being highly sub-optimal, I'd like to see the save and steps to reproduce.

I don't think I have a save (would have to dig through), but I did end up group-moving a fleet blob around a system and the frigates decided that some target under a shield sphere was a priority kill and WOULD NOT MOVE OFF until those targets were dead (despite having never received an explicit attack order).  Every other ship kept flying around the system killing with impunity.

Does that count?

That is a limitation of the fact the auto-given targets are not premptable by the auto-target selection. Don't bother asking for it, I already did, twice. Apparently the auto-targeting logic is so tight on CPU that even including in the loop ships currently having auto-given attack targets would hurt performance too much.

[Zeba]

Doesn't anyone ever unlock up to mark III riot starships in the beggining? I find they make taking the first 5~7 worlds remarkably easy to do when kitted out with full shields and a balanced mix of guns. Add in a dozen merc bombers and fighters and your early game should be easy peasy.

edit; playing on difficulty 8 no less.

[Lancefighter]

Riot starships are too expensive early on...

[Zeba]

Riot starships are too expensive early on...

Waiting an extra few minutes to churn out a full cap of mark III riots to roll the initial ai worlds seems a good tradeoff of resource cost vs attention span to me. But then again I have successfully mined in eve so maybe I have an extra measure of patience than most.   

[Lancefighter]

mmm so true.

I can only ever mine on eve if I'm suitably distracted..

[Ozymandiaz]

I mined in EVE once, then I never did it again

However I do blow miners up from time to time...

[Draco18s]

Doesn't anyone ever unlock up to mark III riot starships in the beggining?

Does anyone unlock Riot IIIs at all?
Riot IIs are so much more awesome.
(Does anyone build Riot Is first?)

[Sunshine!]

Some triangle rebalancing:

    * The rationale here is that bombers have been having their way with forcefields a bit too much, and having fighters be so much more "general-dps" than the other two has made them much less a natural predator of the Bomber. Also, the Missile Frigate is still being reported as the least desirable by a significant margin.
    * Fighters (including the tachyon and bulletproof variants) :
          o Bonus vs Polycrystal from 2.4 => 5.
    * Bombers:
          o Bonus vs UltraHeavy from 10 => 6.
          o Bonus vs Structural from 10 => 6.
          o Bonus vs Heavy from 10 => 6.
          o Bonus vs Artillery from 10 => 6.
          o Base Attack Power from 1900*mk => 2400*mk.
    * Missile Frigates:
          o Bonus vs Light from 10 => 6.
          o Bonus vs UltraLight from 10 => 6.
          o Bonus vs Swarmer from 10 => 6.
          o Bonus vs Neutron from 10 => 6.
          o Bonus vs Composite from 10 => 6.
          o Bonus vs Refractive from 10 => 6.
          o Base Attack Power from 1600*mk => 2400*mk.
          o Base Crystal Cost from 700 => 500.

Honestly, I don't think this actually fixes the problem.  10 mk1 bombers still does 1.4 million damage to a forcefield - if the AI gets a group of 60 of them on-planet, a player can still expect to be in a world of pain.

Part of the problem is that the AI doesn't need to deal with production costs in most cases (exogalactic strikes are an exception, I believe), and only worries about what kind of cap a ship has.  So bombers, which are four times as expensive as fighters for human players, show up with the same prevalence in AI forces.  A wave comes in, and it can be 100 fighters or 100 bombers (if I'm understanding the mechanic correctly), even though the 100 bombers are exponentially more dangerous to the player as a wave because they'll punch through player defenses a billion times faster than fighters.  Some kind of modifier because of cost (and not just cap) would go a long way towards reducing player gripes over bombers without requiring a general nerf of the bombers, especially when we still have 350 million HP forcefields we need to deal with.

This also applies to things like Cutlasses (which are even more dangerous than bombers in a lot of cases since they just LOVE eating power plants) - Cutlasses spawn in huge swarms because their ship cap is twice that of fighters, and yet they cost more to produce than fighters.

[Draco18s]

Honestly, I don't think this actually fixes the problem.  10 mk1 bombers still does 1.4 million damage to a forcefield - if the AI gets a group of 60 of them on-planet, a player can still expect to be in a world of pain.

Check your math.  1 volley from 10 bombers is 144,000 damage.

Part of the problem is that the AI doesn't need to deal with production costs in most cases (exogalactic strikes are an exception, I believe), and only worries about what kind of cap a ship has.  So bombers, which are four times as expensive as fighters for human players, show up with the same prevalence in AI forces.  A wave comes in, and it can be 100 fighters or 100 bombers (if I'm understanding the mechanic correctly), even though the 100 bombers are exponentially more dangerous to the player as a wave because they'll punch through player defenses a billion times faster than fighters.  Some kind of modifier because of cost (and not just cap) would go a long way towards reducing player gripes over bombers without requiring a general nerf of the bombers, especially when we still have 350 million HP forcefields we need to deal with.

This also applies to things like Cutlasses (which are even more dangerous than bombers in a lot of cases since they just LOVE eating power plants) - Cutlasses spawn in huge swarms because their ship cap is twice that of fighters, and yet they cost more to produce than fighters.

A very good point, but I don't think Keith/X will go in the suggested direction.

[keith.lamothe]

A very good point, but I don't think Keith/X will go in the suggested direction.

He's good

The main thing there is that the AI really isn't supposed to be treated like it's acting under the same sorts of resource constraints as the humans.  It's got the better part of 2 galaxies (at least) under its control.  If you had even 30 planets with good econ tech, would you have much problem churning out bombers at the same speed as fighters?  Just a matter of placing enough Space Docks and engineers, the actual resource cost is not much of a concern there.