Matrix Games Forums

Forums  Register  Login  Photo Gallery  Member List  Search  Calendars  FAQ 

My Profile  Inbox  Address Book  My Subscription  My Forums  Log Out

Observations on micromanagement vs ai

 
View related threads: (in this forum | in all forums)

Logged in as: Guest
Users viewing this topic: none
  Printable Version
All Forums >> [New Releases from Matrix Games] >> Command: Modern Air / Naval Operations >> Observations on micromanagement vs ai Page: [1]
Login
Message << Older Topic   Newer Topic >>
Observations on micromanagement vs ai - 10/9/2013 12:57:36 PM   
bvoid

 

Posts: 32
Joined: 5/16/2013
Status: offline
I have some observations from my most recent game, I played the scenario with Indian CVBG vs Chinese CVBG: Tiger and Dragon 2019. Had a lot of fun, but I feel my tactics relied more on the ai being bad.

As India I was able to wipe out the entire Chinese fleet. I did take some losses (5xmigs 1xfrig 2xdd) but that was more because of me being a noob at the start. Am about to try playing as China but am confident using the same tactics I can win, the 2 sides seem balanced.

There are lots of "tricks" you can do to exploit the AI:

1. Dance a fighter around at edge of SAM range.
When you see incoming missiles turn around and afterburn out of there. Can waste 2 vampires per fighter easily doing this. I was even able to get close enough to launch an AA-11 at the carriers ASW/AEW helis, and then escape the vampires. In real life I imagine the sam operator would think "I see what you did there..." and adjust their dlz after the first failure, but yes I understand programming this is non-trivial.

Possible bug: "Pessimistic Dynamic Launch Zone" is not pessimistic enough. Too easy to escape by just afterburning away. I do understand this is difficult to program, since the aggressor may not know it's target is an extremely fast Mig29K.

Related possible bug: AEW should at least make some kind of effort to gtfo when the hostile fastmover is incoming.

2. Exploit "bad readings" to launch SS missiles early.
Maybe this isn't a bug or exploit, but just because of optimistic vs pessimistic DLZ. I usually use "hold fire", and the game won't let me manually launch ssm at long range. However when an "inaccurate reading" appears closer I can then launch. Then I can manually retarget the missiles to a proper trajectory.

Suggested fix: Maybe this isn't a bug at all, but just because I am using pessimistic dlz. I would suggest a small UI change, make manual fire use the optimistic DLZ? If I am firing manually then please let me worry about the DLZ myself - just make sure the target is inside the "absolute max".

Related bug: This affects the AI too, they will see a "false reading" close by and launch massive amounts of missiles at it. This often results in the ai having depleted magazines by the time the real target gets in range.

Suggested fix: AI should be more pessimistic about ambiguous contacts.

3. Precise targeting of sea skimmers:
If you manually retarget your skimmers so they only go active at the very last moment you can almost completely bypass enemy defences. This is a bit weird, the Chinese CVBG was not running it's radars etc - but at the last moment when they detected my vampires they would all light up. But too late! Using this precise targeting I was able to send in 2xharpoonski kh-35 at a time at each ship. Not a saturation bombardment at all, the cvbg could easily shoot these down if they detected them earlier.

Suggested fix: The enemy fleet should have had it's sensors on in this situation. I had shot down their AEW helo about an hour beforehand, and shot up the other escort groups. It's madness to keep radars turned off when stuff is blowing up around you!

edited for clarity etc


< Message edited by bvoid -- 10/9/2013 1:29:51 PM >
Post #: 1
RE: Observations on micromanagement vs ai - 10/9/2013 1:53:26 PM   
ExMachina


Posts: 456
Joined: 9/25/2013
Status: offline
quote:

I would suggest a small UI change, make manual fire use the optimistic DLZ


Agreed! I'm coming to the general conclusion that Command needs to allow the player more control over the AI's sometimes wasteful and naive weapons use. A corollary to that is what you say: the player should be allowed a wider range of discretion when choosing weapons' allocation.

(in reply to bvoid)
Post #: 2
RE: Observations on micromanagement vs ai - 10/9/2013 3:22:22 PM   
$trummer

 

Posts: 224
Joined: 7/27/2013
Status: offline
A version of this exists in sub combat; the AI will not take high-aspect torpedo shots but will maneuver to get into the target's baffles. This is often good doctrine but not always.

(in reply to ExMachina)
Post #: 3
RE: Observations on micromanagement vs ai - 10/16/2013 3:42:35 PM   
Dimitris


Posts: 9824
Joined: 7/31/2005
Status: offline
quote:

ORIGINAL: bvoid
1. Dance a fighter around at edge of SAM range.
When you see incoming missiles turn around and afterburn out of there. Can waste 2 vampires per fighter easily doing this. I was even able to get close enough to launch an AA-11 at the carriers ASW/AEW helis, and then escape the vampires. In real life I imagine the sam operator would think "I see what you did there..." and adjust their dlz after the first failure, but yes I understand programming this is non-trivial.

Possible bug: "Pessimistic Dynamic Launch Zone" is not pessimistic enough. Too easy to escape by just afterburning away. I do understand this is difficult to program, since the aggressor may not know it's target is an extremely fast Mig29K.


As you say, this is quite hard to get right. The ultimate expression of firing conservatism would be to rely exclusively on NEZ instead of DLZ. But that gets you into wierd H2-era situations like an F-14 and a MiG-25 launching against each other at the same time. (And it doesn't solve the problem entirely either; the other guy may waltz-in slow deliberately in order to "fool" your calcs, and turn around and run with AB once you start firing.)

In RL, what are the factors that prevent pilots from over-using this tactic to win/survive missile engagements?

< Message edited by Sunburn -- 10/16/2013 3:45:59 PM >


_____________________________


(in reply to bvoid)
Post #: 4
RE: Observations on micromanagement vs ai - 10/16/2013 4:06:35 PM   
Tomcat84

 

Posts: 1942
Joined: 7/10/2013
Status: offline

quote:

ORIGINAL: Sunburn
In RL, what are the factors that prevent pilots from over-using this tactic to win/survive missile engagements?


Timing, if theres a set TOT window that you must adhere to (or the rest of the COMAO coming up behind you) then at some point (depending on risk level) you'll have to press forward to make it.

Also: probably the enemy will switch tactics and hold their fire for longer if he observes this.

But it is certainly a valid tactic, especially against strategic SAMs whose location is known.

For mobile SAMs one thing I am experimenting with right now for a WIP scenario is using the event editor to teleport mobile SAMs in when aircraft are at a closer range, simulating unlocated SAMs being quiet until aircraft are close and then going active and launching.

In future hopefully this can be simplified by having event editor actions to toggle EMCON and toggle hold fire, that way taking out the teleporting part (and giving you the option to detect it sooner if you have capable assets (maybe JSTARS for example).

Another thing I could think of just off the top of my head is maybe having a unit/side settable (and event editor changeable!) doctrine slider that you could set from 100% max range to any lower value you want if you want your guys to hold shots. Would not have to be % but could also be like, Max, normal, RTR or something.

(in reply to Dimitris)
Post #: 5
RE: Observations on micromanagement vs ai - 10/16/2013 4:13:50 PM   
ExMachina


Posts: 456
Joined: 9/25/2013
Status: offline
quote:

In RL, what are the factors that prevent pilots from over-using this tactic to win/survive missile engagements?


I'd guess not being as fully aware of the battle space (like the player is in CMANO). In CMANO every "pilot" knows what's happening everywhere; he has really good estimates of specific enemy AAW capabilities, he can usually count all enemy shots, and he can cross reference all that information with the enemy unit data base.

Generally, I think of CMANO as an idealized (or "textbook") imagining of modern engagements, one that allows the player to explore the "what ifs" while relying on the AI to help keep him on track..



< Message edited by ExMachina -- 10/16/2013 4:24:43 PM >

(in reply to Dimitris)
Post #: 6
RE: Observations on micromanagement vs ai - 10/16/2013 4:34:32 PM   
madflava13


Posts: 1511
Joined: 2/7/2001
From: Alexandria, VA
Status: online

quote:

ORIGINAL: bvoid


1. Dance a fighter around at edge of SAM range.
When you see incoming missiles turn around and afterburn out of there. Can waste 2 vampires per fighter easily doing this. I was even able to get close enough to launch an AA-11 at the carriers ASW/AEW helis, and then escape the vampires. In real life I imagine the sam operator would think "I see what you did there..." and adjust their dlz after the first failure, but yes I understand programming this is non-trivial.

Possible bug: "Pessimistic Dynamic Launch Zone" is not pessimistic enough. Too easy to escape by just afterburning away. I do understand this is difficult to program, since the aggressor may not know it's target is an extremely fast Mig29K.

Related possible bug: AEW should at least make some kind of effort to gtfo when the hostile fastmover is incoming.



How is this a bug? In real life, SEAD missions frequently "bait" SAM units to launch, then evade using tactics similar to this. We generally know the range and speed of enemy SAM units, thanks to the CIA/DIA/whomever, so this isn't an unrealistic tactic in my mind. I do not have the game (yet), so I may be missing the point here - if so I apologize.

_____________________________

"The Paraguayan Air Force's request for spraying subsidies was not as Paraguayan as it were..."

(in reply to bvoid)
Post #: 7
RE: Observations on micromanagement vs ai - 10/16/2013 4:54:33 PM   
ExMachina


Posts: 456
Joined: 9/25/2013
Status: offline
quote:

How is this a bug? In real life, SEAD missions frequently "bait" SAM units to launch, then evade using tactics similar to this


I once depleted a Brit frigate of all her Sea Darts using this tactic with a single pair of fighters.

It's the whole idea of "fool me once shame on you, fool me twice shame on me"

(in reply to madflava13)
Post #: 8
RE: Observations on micromanagement vs ai - 10/16/2013 5:08:10 PM   
madflava13


Posts: 1511
Joined: 2/7/2001
From: Alexandria, VA
Status: online

quote:

ORIGINAL: ExMachina

quote:

How is this a bug? In real life, SEAD missions frequently "bait" SAM units to launch, then evade using tactics similar to this


I once depleted a Brit frigate of all her Sea Darts using this tactic with a single pair of fighters.

It's the whole idea of "fool me once shame on you, fool me twice shame on me"


Ok, that makes a little more sense. To be technical, I don't think it's a "bug," but more properly called a feature that isn't working as intended? I am not a programmer, so I don't know how tough it is to change this - but it seems like the obvious response is "don't do that to the AI if you think it's gamey."

_____________________________

"The Paraguayan Air Force's request for spraying subsidies was not as Paraguayan as it were..."

(in reply to ExMachina)
Post #: 9
RE: Observations on micromanagement vs ai - 10/16/2013 5:22:36 PM   
thewood1

 

Posts: 3981
Joined: 11/27/2005
Status: offline
Harpoon devs (various ones) worked on this issue for decades with almost no positive results. In a couple instances, they broke more than they fixed. Programming learning/memory into an AI is task that is usually beyond any but the best systems developers.

(in reply to madflava13)
Post #: 10
RE: Observations on micromanagement vs ai - 10/16/2013 6:08:38 PM   
CV32


Posts: 1041
Joined: 5/15/2006
From: The Rock, Canada
Status: offline
quote:

ORIGINAL: thewood1
Harpoon devs (various ones) worked on this issue for decades with almost no positive results. In a couple instances, they broke more than they fixed. Programming learning/memory into an AI is task that is usually beyond any but the best systems developers.


I wouldn't say we achieved "almost no" positive results but its certainly true that modeling an intuitive, flexible, responsive AI is a Herculean task. As for micro management, I find myself still doing it with CMANO despite the obviously much more complex models.

_____________________________

Brad Leyte
HC3 development group member for HCE
Author of HCDB official database for HCE
Harpgamer.com Co-Owner

(in reply to thewood1)
Post #: 11
RE: Observations on micromanagement vs ai - 10/16/2013 6:24:26 PM   
thewood1

 

Posts: 3981
Joined: 11/27/2005
Status: offline
I am going back to the 360 days...they screwed this up a few times.

(in reply to CV32)
Post #: 12
RE: Observations on micromanagement vs ai - 10/17/2013 6:01:37 AM   
MR_BURNS2


Posts: 974
Joined: 7/18/2013
From: Austria
Status: offline

quote:

ORIGINAL: Tomcat84

Another thing I could think of just off the top of my head is maybe having a unit/side settable (and event editor changeable!) doctrine slider that you could set from 100% max range to any lower value you want if you want your guys to hold shots. Would not have to be % but could also be like, Max, normal, RTR or something.


+1, i had similar thoughts, other thoughts:

reduce DLZ engagement range automatically when ammo reaches a certain limit.

for aircraft, program certain simple basic gameplans, like bracket, crank after firing missiles to reduce closure, grinder etc.

being able to set them up in a certain formation with a single click would also help, like spread or trail.
Doing it manually over the F4 menu is a bit labourious.

I don´t know how hard it would be to program things like that, but surely easier then create a smarter independent AI. It would give the scenario designer means to make the AI attacks harder


_____________________________

Windows 7 64; Intel(R) Core(TM) i7 CPU 920 @ 2.67GHz (8 CPUs), ~2.7GHz; 6144MB RAM; NVIDIA GeForce GTX 970;



(in reply to Tomcat84)
Post #: 13
Page:   [1]
All Forums >> [New Releases from Matrix Games] >> Command: Modern Air / Naval Operations >> Observations on micromanagement vs ai Page: [1]
Jump to:





New Messages No New Messages
Hot Topic w/ New Messages Hot Topic w/o New Messages
Locked w/ New Messages Locked w/o New Messages
 Post New Thread
 Reply to Message
 Post New Poll
 Submit Vote
 Delete My Own Post
 Delete My Own Thread
 Rate Posts


Forum Software © ASPPlayground.NET Advanced Edition 2.4.5 ANSI

0.127