Fractal Softworks Forum

Please login or register.

Login with username, password and session length
Advanced search  

News:

Starsector 0.97a is out! (02/02/24); New blog post: New music for Galatia Academy (06/12/24)

Pages: 1 [2]

Author Topic: AI Shielding too good?  (Read 7813 times)

Jazwana

  • Commander
  • ***
  • Posts: 108
    • View Profile
Re: AI Shielding too good?
« Reply #15 on: June 23, 2013, 08:18:58 PM »

back to the thread topic, I would agree with nanostrike that the AI's shield usage is what I'd call poor macro compensated by really good micro.  It makes sub-optimal (bad) strategic decisions about when to vent, i.e. it doesn't when a small window opens up where a skilled player will trade off a bit of armor or hull damage to empty 25%-50% of the flux cap.  Or, it waits in a long window to make absolutely sure no damage will hit it, when a skilled player would accept a bit of damage (if the AI suicided into them) in order to get the max speed boost, opening up tactical options for the player.    It makes sub-optimal decisions on when to raise shields, such as right after! it bumps into the asteroid, imparting velocity that would prevent a second hit but not before taking the armor damage.

To be fair, it is likely really really hard to program an AI to make optimal decisions.  I certainly can't think of a rule-set that would be required.  So, in part to compensate for poor flux reduction choices there is spot-on flux cap management.  It's a trade-off that can feel cheap sometimes, but what would the alternative be? 

edit:  and I guess to answer that question, I kind of like nanostrike's idea of crew levels imparting some differentiation within the AI's "micro" to use the above analogy, while the captain's traits would give differentiation to the "macro" AI decisions.
« Last Edit: June 23, 2013, 08:21:21 PM by Jazwana »
Logged

Nanostrike

  • Captain
  • ****
  • Posts: 357
    • View Profile
Re: AI Shielding too good?
« Reply #16 on: June 24, 2013, 03:14:18 PM »

back to the thread topic, I would agree with nanostrike that the AI's shield usage is what I'd call poor macro compensated by really good micro.  It makes sub-optimal (bad) strategic decisions about when to vent, i.e. it doesn't when a small window opens up where a skilled player will trade off a bit of armor or hull damage to empty 25%-50% of the flux cap.  Or, it waits in a long window to make absolutely sure no damage will hit it, when a skilled player would accept a bit of damage (if the AI suicided into them) in order to get the max speed boost, opening up tactical options for the player.    It makes sub-optimal decisions on when to raise shields, such as right after! it bumps into the asteroid, imparting velocity that would prevent a second hit but not before taking the armor damage.

To be fair, it is likely really really hard to program an AI to make optimal decisions.  I certainly can't think of a rule-set that would be required.  So, in part to compensate for poor flux reduction choices there is spot-on flux cap management.  It's a trade-off that can feel cheap sometimes, but what would the alternative be? 

edit:  and I guess to answer that question, I kind of like nanostrike's idea of crew levels imparting some differentiation within the AI's "micro" to use the above analogy, while the captain's traits would give differentiation to the "macro" AI decisions.

This is pretty much what I was saying.  I can beat the AI pretty easily.  It's just VERY annoying how easily they can stutter their shields on/off and perfectly time dropping them to avoid Sabots and such with a Green Crew while I watch my crew of Elites on an AI-Piloted ship do all sorts of stupid shennanigans with their shields, like being overloaded by Broadsword wings because they won't lower their shields and take the tiny bit of damage from those Swarm Missiles.

It's kinda at the point where you can't overload the AI's shields without either a Strike Weapon (Antimatter Blaster, ect) or some sort of missiles unless it makes a really boneheaded decision.
Logged

Megas

  • Admiral
  • *****
  • Posts: 12221
    • View Profile
Re: AI Shielding too good?
« Reply #17 on: June 24, 2013, 05:12:16 PM »

When I try to kill the AI with a beam only ship, say starting Wolf vs. enemy Lasher, the AI "stutters" its shield usage perfectly such that it takes the longest time possible to kill it.
Logged

Nanostrike

  • Captain
  • ****
  • Posts: 357
    • View Profile
Re: AI Shielding too good?
« Reply #18 on: June 25, 2013, 01:08:00 AM »

When I try to kill the AI with a beam only ship, say starting Wolf vs. enemy Lasher, the AI "stutters" its shield usage perfectly such that it takes the longest time possible to kill it.

And it'll do this at nearly full flux.  A human trying to play it that close would be at a HUGE risk of overload, but the AI can manage to perfectly time it.  Stuff like this is what I'm talking about.
Logged

targetstar

  • Ensign
  • *
  • Posts: 27
    • View Profile
Re: AI Shielding too good?
« Reply #19 on: June 25, 2013, 03:09:37 PM »

Actully i take advantage of the AI shielding use.

I use Tempest... they are awsome. and there drones are helpful but not a nessecity to me.

In the early game i was a single tempest with either 2 Phase Beam or 1 Phase and 1 Gravity. (i aimed for the 2 phase). and I would easily kill the pirates even against larger ships. I play smart.

On a 1v1, (so me vs a Lasher or a larger ship) i use the phase beam to hit him, and his shields stay up. I use the long range of the beams to get out of his assults gun range and use my shields to avoid being hit by his crippling misslies. When ever my flux get to high, i actully just vent flux but keep within range of him. The ai will some how think that i am still going to shot him despit me venting. He keeps his shields up and I get to do this all over again. And when his flux gets to high, and he has to drop it. I get to tear him a new one.

Useing this i like to have a squadren of tempest. useing Hit and Run tactics. By seperating them out and using a bunch of tempest lasers at once. I can easily overload even the larger ships like the Enforcers. I was actully able to take out the Pirate Armada Fleet with just 5 Tempest. (though more than i liked escaped and i just had to fight them a 2nd time to die)
Logged

Nooblies

  • Lieutenant
  • **
  • Posts: 72
    • View Profile
Re: AI Shielding too good?
« Reply #20 on: June 30, 2013, 10:19:54 AM »

The AI shield usage is quite good but has some noticeable flaws. On low tech ships, such as the dominator, it is far too cautious and keeps shields raised if you are so much as looking in its general direction. This prevents it from dissipating flux efficiently (due to the flux upkeep of the shield) for no real gain, as changes to the speed at which frontal shields raise has made it much more difficult to "snipe" or slowly pick away at the edges of the ship before shields can be raised to block the shots (which is most likely the reason for its overly cautious behaviour).

It is also predictable in always trying to block explosive missiles with the shield, to such an extent that raising the enemy ship to high flux and firing a harpoon will almost always cause it to raise shields and you can easily overload it by firing at it as the harpoon is on approach. I find doing so to be much more effective at overloading ships than using sabots funnily enough.
Logged

DatonKallandor

  • Admiral
  • *****
  • Posts: 718
    • View Profile
Re: AI Shielding too good?
« Reply #21 on: June 30, 2013, 11:57:59 AM »

Making shields have to drop at the same speed they deploy (or slower) would make dedicated anti-shield weapons with a travel time like Sabots much more useful. It'd also open up the possibility of a shield skill (and augment) that's more interesting than %upkeep reduction and %flux per damage. Perhaps even partly crew veterancy dependent.
Logged
Pages: 1 [2]