Fractal Softworks Forum

Please login or register.

Login with username, password and session length
Advanced search  

News:

Starsector 0.9.1a is out! (05/10/19); Blog post: Personal Contacts (08/13/20)

Pages: 1 2 [3] 4 5 ... 9

Author Topic: 0.9.1a Balance Testing Case Study: Condor  (Read 4666 times)

SCC

  • Admiral
  • *****
  • Posts: 1983
    • View Profile
Re: 0.9.1a Balance Testing Case Study: Condor
« Reply #30 on: July 11, 2020, 08:44:14 AM »

I admit the ships used by the enemy might have been unfairly skewed against Condors. I didn't think about it at the time, since most of the fights were resolved in a binary fashion: either Onslaught died and so did the enemy fleet, or it didn't; but accidental flankers might wreck Condors and give the enemy enough power to get past the player-fleet-wrecking edge.
I will probably remake and split the enemy fleet into high-tech-ish and low-tech-ish fleets... Later. Enough testing for me now.

Comparing this to a real campaign situation: you would realistically have a lower number of Condor (those left from early game), and more non-carrier ships, and probably dedicated escorts. This what I have in my current campaign: 2 Condors, 3 Herons, 8 frigates, 2 destroyers, 3 cruisers, 1 Legion recently added. Condors are not being targeted that much, and have a permanent 1 Shepherd escort. I have left them unmanaged since early game with 100% survival rate.
There's a 1:2 ratio of carriers to warships, which I don't consider outlandish. I'd say that with fewer carriers, though, Condors are even worse off, since fighters are more likely to be lost and drain their replacement rate, whereas Drovers can leverage their ship system to prevent and mitigate losses.

Thanks very much for the information.  I don't suppose you know if the Admiral AI you are using the same one used in the AI battles mods?  I used it assuming it'd been setup with at least some form of fleet AI for both sides.  Alternatively, I'd be interesting to know which mod to grab that has said Admiral AI.
I remember downloading it from Tartiflette on Discord, since Fleet Tester doesn't have any admiral AI. If it has one now, it has been added in the meantime. Whenever I paid attention to the battle going on, it didn't seem to do much; assign escorts to cruisers at the very start of the game, then discard escort orders and order assault at the objective in the centre of the battlefield, then discard that order and focus only on retreating individual ships until the end.

Were all individual ships set to steady AI?
Yes.
Although I'm wondering how much its affected by fighter selection versus base ship.  I'd like to try this fleet setup, but with all 4 carrier types using the same ratios and types of fighters, to help eliminate other variations beyond the base ship.
But why? Why wouldn't better carriers leverage the capability to use better fighters than others? We might have different priorities and I can change variants to what you consider better, but I won't run suboptimal variants just to make all carriers use the same fighters. Though Condors with reapers are probably a better choice against the fleet that I tested against (which I'll fix later).

FooF

  • Admiral
  • *****
  • Posts: 593
    • View Profile
    • Email
Re: 0.9.1a Balance Testing Case Study: Condor
« Reply #31 on: July 11, 2020, 09:02:39 AM »

So what I've learned from all of this is that fighter superiority is even more important that I thought originally. I knew that it's good to have a carrier or two in your fleet just to help with those pesky frigates or just pure distraction, but this, this is just wrong. I wonder how a fleet full of Converted Hangars would perform vs just the ''usual builds'' (I know it's not the thread for this, I'm just thinking out loud). So yeah long story short, fighters are too strong and AI is too dumb vs fighters.

Anecdotal evidence: I am currently running a fleet of an Odyssey, Eagle (XIV), Heron, 2 Drovers and 12 Hammerheads with CH with Sparks, Broadswords, and Thunders. I am demolishing endgame bounties. The only ship without fighters is the Eagle. The AI is really skiddish around swarms of fighters so most ships start backing off even when the enemy fleet ought to control the battlespace. I notice fights really get spread out and that's where the range of all the fighters become a force multiplier where traditional ships wouldn't be able to do anything about it.

As for the rest of the thread, I don't have much to offer but I am surprised the Condors are doing so well vs. the Drovers.
Logged

Hiruma Kai

  • Commander
  • ***
  • Posts: 177
    • View Profile
Re: 0.9.1a Balance Testing Case Study: Condor
« Reply #32 on: July 11, 2020, 09:16:23 AM »

But why? Why wouldn't better carriers leverage the capability to use better fighters than others? We might have different priorities and I can change variants to what you consider better, but I won't run suboptimal variants just to make all carriers use the same fighters. Though Condors with reapers are probably a better choice against the fleet that I tested against (which I'll fix later).

I'd argue Condors can run Broadsword/Daggers, just the same as drovers, and still have anti-fighter guns for when the lone drover sends its fighter back in response.  Generally, my thinking is if a carrier isn't intended to be near the front line, anything not spent on fighters or anti-fighter defense (the only thing that can really reach you), is not an effective use of OP.  If for most of the fight the Condor isn't in range to use the Harpoons, they're dead weight.  I personally think the extra 3 OP spent on Daggers is worth it, especially with fast frigates and destroyers in the mix.  Certainly the mirror matches seem to indicate Condors didn't suffer by using Broadsword/Daggers compared to Condors.

In the mean time, I definitely think there is some kind of AI difference going on, that or allowing retreats is pulling fighters off the board faster than no retreats.  Testing so far with AI Battles version 6.4 gives the following:

Condors with Broadswords/Drovers: 3 for 3 victories.
Condors with Broadsword/Peridition (3) and Thunder x2 (3): 2 for 3 victories (although these are literally going to 0% Cr for everyone but the capitals, and the one loss literally only had the Onslaught on the field at the end).


I'll take a look on Discord to see if I can find Fleet tester with admiral AI.  Alternatively, I'll download the latest fleet tester and see how its results compare to yours.  I definitely want to understand why my results are hinting at a different outcome.

Edit: Just double checking my configuration, I realized I had left the wolf out of the enemy fleet, which means I'm not comparing apples to apples.  I'll need to rerun the tests with the appropriate enemy force.
« Last Edit: July 11, 2020, 09:39:35 AM by Hiruma Kai »
Logged

Thaago

  • Global Moderator
  • Admiral
  • *****
  • Posts: 4712
  • Quantum Mechanic
    • View Profile
    • Email
Re: 0.9.1a Balance Testing Case Study: Condor
« Reply #33 on: July 11, 2020, 09:51:55 AM »

: Takes copious notes :

I suspect that exact loadouts are going to make a big difference - in particular, with so many frigates any bomber other than a Dagger is dead weight.

In other news: in campaign last night the only medium missile I had around for the Condor was the medium Reaper. The AI will, on occasion, use its system to double tap enemy ships with 2 consecutive reapers. In this case 2x Thunder Condor, so the enemy was flamed out. I don't think this is an optimal build or anything, as the Reaper is almost always wasted, but its very gratifying when its not.
Logged

Hiruma Kai

  • Commander
  • ***
  • Posts: 177
    • View Profile
Re: 0.9.1a Balance Testing Case Study: Condor
« Reply #34 on: July 11, 2020, 10:58:54 AM »

So, using  the Battle AI mod, and the same fleet loadout as SCC, except using 6x Condor with broadsword/dagger/2x Vulcan/1x Reaper/Expanded Deck Crew/1 capacitor

I have the following statistic: 3 wins out of 5 matches for the Condors

However, SCC and Tartiflette have kindly shared the admiral AI .jar for the fleet tester mod on discord, so I've downloaded that and will do the same test with that mod, and see what things look like.

Edit: I only watched 2 of the battles, but I think the daggers are much better at securing small kills, compared to the periditoins. Especially in the first exchange.  The last fight I watched the medusa came in fast, got fluxed up, then 9 dagger atropos took it out after it had burned its teleport charges.  This happened in like the first 15-20 seconds of contact.  And as I discovered with my incorrect fleet testing, even a 1 frigate difference at the beginning can snowball significantly.

Edit 2: Another thought.  When I'm playing in the campaign, my default 2 fighter slot carrier are using either broadsword/daggers or longbow/daggers (or x2 Longbows on the Odyssey).  My 3 slot carriers will have broadsword/longbow/dagger.  I might switch up for a station attack, but those are my go to general setups when I don't know what I might be engaging.  Do thunders really bring that much to a fight compared to bombers (especially daggers).
« Last Edit: July 11, 2020, 11:30:20 AM by Hiruma Kai »
Logged

Aereto

  • Captain
  • ****
  • Posts: 275
    • View Profile
Re: 0.9.1a Balance Testing Case Study: Condor
« Reply #35 on: July 11, 2020, 12:27:06 PM »

: Takes copious notes :

I suspect that exact loadouts are going to make a big difference - in particular, with so many frigates any bomber other than a Dagger is dead weight.

In other news: in campaign last night the only medium missile I had around for the Condor was the medium Reaper. The AI will, on occasion, use its system to double tap enemy ships with 2 consecutive reapers. In this case 2x Thunder Condor, so the enemy was flamed out. I don't think this is an optimal build or anything, as the Reaper is almost always wasted, but its very gratifying when its not.

Reapers to force them to raise their shields, then either Longbow's sabots or Dagger's torpedoes to force an overload. That's a Condor's Gambit.

Of course the reapers are best when used on overloaded ships.
« Last Edit: July 11, 2020, 01:09:07 PM by Aereto »
Logged

Morrokain

  • Admiral
  • *****
  • Posts: 1445
  • Megalith Dreadnought - Archean Order
    • View Profile
Re: 0.9.1a Balance Testing Case Study: Condor
« Reply #36 on: July 11, 2020, 12:48:41 PM »

So what I've learned from all of this is that fighter superiority is even more important that I thought originally. I knew that it's good to have a carrier or two in your fleet just to help with those pesky frigates or just pure distraction, but this, this is just wrong. I wonder how a fleet full of Converted Hangars would perform vs just the ''usual builds'' (I know it's not the thread for this, I'm just thinking out loud). So yeah long story short, fighters are too strong and AI is too dumb vs fighters.

Anecdotal evidence: I am currently running a fleet of an Odyssey, Eagle (XIV), Heron, 2 Drovers and 12 Hammerheads with CH with Sparks, Broadswords, and Thunders. I am demolishing endgame bounties. The only ship without fighters is the Eagle. The AI is really skiddish around swarms of fighters so most ships start backing off even when the enemy fleet ought to control the battlespace. I notice fights really get spread out and that's where the range of all the fighters become a force multiplier where traditional ships wouldn't be able to do anything about it.

As for the rest of the thread, I don't have much to offer but I am surprised the Condors are doing so well vs. the Drovers.

It's really the AI that is the biggest problem imo. Fighters in large numbers aren't necessarily stronger than combat ships - especially since the more stacked they are the more actual connections per shot of larger weapons are possible - creating higher fleetwide flux efficiency as numbers increase..

The thing is, the presence of fighters completely cripples the AI's willingness to leverage fleet firepower or close with the enemy and allows them to be picked off piecemeal. It is far too conservative to handle large fighter numbers because there will always be some fighters on the field - causing the enemy to retreat instead of close the distance and pressure the carrier.

That is even before kiting comes into play.

Combat ships have to ignore fighters for positional purposes or this will always be the case. Even if fighters are weakened to the point of uselessness the AI can still be abused by this behavior so the player can pick their fights a lot easier than with pure combat ships on the field.
Logged

Hiruma Kai

  • Commander
  • ***
  • Posts: 177
    • View Profile
Re: 0.9.1a Balance Testing Case Study: Condor
« Reply #37 on: July 11, 2020, 02:33:19 PM »

So I've got the fleet tester mod with admiral AI setup, and running and its basically producing the same results as the AI Battles mod (which was a tournament mod).

This is again using SCC's suggested line up.

Anyways, so before with AI Battles mod: 3 wins out of 5 matches for Condors
With Fleet Tester + Admiral AI .jar: 3 wins out of 5 matches for Condors

This again is with 60 DP of "Broadsword/Dagger/Reaper/2x Vulcan/Expanded Deck Crew/rest into caps" Condor, to match up against the Drover setup.

So overall, 6/10, which at this sample size is pretty indistinguishable from the Drover victory odds using Broadswords and Daggers as well.

So I submit fighter selection has a far greater effect than, say filling the medium missile slot on a Condor.  Also, PPT actually plays a huge part in these fights.  Generally the destroyers are at 0% CR by the end, and any cruisers are in critical malfunction territory at the end.

Although, there was one perfect win for the Condors, no losses and only 3 retreats.  Such is the randomness of AI.

So next up is the Thunder x2/Broadsword + Peridition Drover tests to confirm its more an issue of fighter selection rather than base ship.
« Last Edit: July 11, 2020, 02:34:53 PM by Hiruma Kai »
Logged

Thaago

  • Global Moderator
  • Admiral
  • *****
  • Posts: 4712
  • Quantum Mechanic
    • View Profile
    • Email
Re: 0.9.1a Balance Testing Case Study: Condor
« Reply #38 on: July 11, 2020, 02:37:18 PM »

I'm not surprised to learn that CR/passive play is a big problem. Between fighters flying around and no admiral to give eliminate/fighter strike/engage commands, and no player to push things, its a perfect storm to make for slow play!
Logged

Morrokain

  • Admiral
  • *****
  • Posts: 1445
  • Megalith Dreadnought - Archean Order
    • View Profile
Re: 0.9.1a Balance Testing Case Study: Condor
« Reply #39 on: July 11, 2020, 02:41:09 PM »

I'm not surprised to learn that CR/passive play is a big problem. Between fighters flying around and no admiral to give eliminate/fighter strike/engage commands, and no player to push things, its a perfect storm to make for slow play!

I think the Admiral AI does give commands, or am I wrong there?

So the only thing actually missing is the player piloting something if that is the case.
Logged

Thaago

  • Global Moderator
  • Admiral
  • *****
  • Posts: 4712
  • Quantum Mechanic
    • View Profile
    • Email
Re: 0.9.1a Balance Testing Case Study: Condor
« Reply #40 on: July 11, 2020, 07:56:39 PM »

Oh my bad - I thought they were using the AI version with all that turned off. Depending on the version the tournaments very much mess with AI in order to makes things even and less swingy.
Logged

Hiruma Kai

  • Commander
  • ***
  • Posts: 177
    • View Profile
Re: 0.9.1a Balance Testing Case Study: Condor
« Reply #41 on: July 12, 2020, 12:45:43 PM »

I'm not sure how to tell what the admiral AI is doing.  At the beginning of the AI Battles mod or Fleet tester with the updated jar, there are some initial orders to claim nav/sensor bouys.  That shows up on the map layer.  Then the only other orders I see on the map layer are retreat orders.  The local ship AI is clearly assigning itself to escort though (I've seen drovers follow other ships that are retreating to the edge of the map, then once they retreat, head back to the fight).  And generally in the initial clash, most of the bombers go after the same ship, but then again, that might just have been the first ship seen, or they all had the same selection criteria at the beginning of the fight, so it tends to make them select a single target.

And all I can say is random AI is random.  Well, and steady AI is really, really good at staying alive, with the side effect of not going for the kill in a lot of situations.  Which to be fair, if you're actually playing the game, generally what you want out of your standard AI.  The fact that these fights are going to CR, frigates and destroyers are hitting 0% with several different composition means things can't be too far out of balance.

Although it does makes me wonder if we're testing things the right way.  As it is right now, fleet composition in terms PPT is arguably more important than base ship effectiveness.  A few more cruisers and a few less destroyers and you'll have a significant advantage at the end.  The other thing is, the game isn't intended to be balanced AI vs AI.  Its intended to be player vs AI, and provide a fun and interesting challenge.  A player thrown into the mix will make fights go much shorter, at which point PPT doesn't matter as much.

It also begs the question, of which player's balance?  Someone just piloting their ship and ignoring the AI is going to have different performance from one that switches to the map layer and issues appropriate orders mid-battle.  Keeping your carriers together and having them send a bomber wave at a target you know is going to be high flux in 10-20 seconds, is going to see different effectiveness.  We also know at some point there's a transition where you have so many fighters on your side, the AI can't handle it optimally.

So anyways, I did a bunch of broadsword/peridition +2x thunder setups.  Both Condor and Drover.

I'm including in the spoiler tags the setup files I'm using for the condor fights, just in case someone can spot a mistake.  The Venture_Pulse is just a Venture_Balanced with a Pulse Laser instead of mining blaster.   I did make a mistake on the first 2 Drover fights, and gave them 3x broadsword/peridition and 3x Thunders.  Anyways, I include this because I've got fairly different results from SCC.

Condor fights:
8 matches, 4 wins for the Condors, 3 losses, 1 tie (literally every ship was at 0% CR and retreating off the field - although Conquest did retreat while Onslaught had been destroyed...)

Drover fights
6 Drovers (3 Broad/Perdition, 3 Thunder)
1 win, 1 loss
5 Drovers (3 Broad/Perdition, 2 Thunder)
5 wins, 3 loss. 
Note: The 5 wins were all in a row, and the 3 losses were all in a row at the end - so it was looking like 5/5 at the beginning of testing and I was very confused about why drovers with thunders were so much better than broadsword/dagger).

I have been running some of these matches 2 at a time (two instances of starsector running) to speed up testing, but I don't think that should matter?

Anyways, thunders are better than I had traditionally given them credit for, at least in a AI vs AI setup.  I'm not seeing statistically significant differences between 5 Drover fleets and 6 Condor fleets, at least with this limited testing.  Again, it really is vagary of the AI.   I mean, I've seen the Onslaught destroy the Omen during the first few moments of enemy contact, when they wanted to go in opposite directions and the omen crashed into the Onslaught with its shields down.  Sometimes a destroyer will zoom ahead, flux up, and then the bombers come in.  Some times, the bombers slit their attention on two different ships at the beginning of the fight, running up flux, but not securing a kill.

I'm tempted to switch over to a new scenerio, with superior enemy forces, and seeing how long the carrier fleet lasts.  Probably something like a reckless SO Luddic fleet 180 DP versus 120 DP.  And then see how long the fleet lasts and how much they kill.  That I'm willing to bet is going to show case the Drover's advantages.

player0_fleet.csv
Spoiler
#personality: The captain's personality. Can be "timid", "cautious", "steady", "aggressive", "reckless".
#flagship: boolean to set the flagship of the fleet
#DO NOT RENAME THIS FILE, DO NOT CHANGE THE TOP LINE.
1,conquest_Elite,steady,false
2,eagle_Assault,steady,false
3,venture_Pulse,steady,false
4,enforcer_Elite,steady,false
5,enforcer_Elite,steady,false
6,hammerhead_Balanced,steady,false
7,centurion_Assault,steady,false
8,centurion_Assault,steady,false
9,wolf_Assault,steady,false
10,wolf_Assault,steady,false
#11,drover_Broad_Perd_Harpoon,steady,false
#12,drover_Broad_Perd_Harpoon,steady,false
#13,drover_Broad_Perd_Harpoon,steady,false
#14,drover_Thunder_Harpoon,steady,false
#15,drover_Thunder_Harpoon,steady,false
11,condor_Broad_Perd_Harpoon,steady,false
12,condor_Broad_Perd_Harpoon,steady,false
13,condor_Broad_Perd_Harpoon,steady,false
14,condor_Thunder_Harpoon,steady,false
15,condor_Thunder_Harpoon,steady,false
16,condor_Thunder_Harpoon,steady,false
[close]

player1_fleet.csv
Spoiler
#personality: The captain's personality. Can be "timid", "cautious", "steady", "aggressive", "reckless".
#flagship: boolean to set the flagship of the fleet
#DO NOT RENAME THIS FILE, DO NOT CHANGE THE TOP LINE.
1,onslaught_Standard,steady,false
2,eagle_Balanced,steady,false
3,falcon_Attack,steady,false
4,heron_Strike,steady,false
5,mora_Strike,steady,false
6,condor_Support,steady,false
7,drover_Starting,steady,false
8,medusa_Attack,steady,false
9,sunder_CS,steady,false
10,shrike_p_Attack,steady,false
11,shrike_p_Attack,steady,false
12,tempest_Attack,steady,false
13,omen_PD,steady,false
14,lasher_Assault,steady,false
15,wolf_Strike,steady,false
[close]

drover_Thunder_Harpoon.variant
Spoiler
{
    "displayName": "Strike",
    "fluxCapacitors": 3,
    "fluxVents": 0,
    "goalVariant": true,
    "hullId": "condor",
    "hullMods": ["expanded_deck_crew"],
    "permaMods": [],
    "variantId": "condor_Thunder_Harpoon",
    "weaponGroups": [
        {
            "autofire": true,
            "mode": "LINKED",
            "weapons": {
                "WS 001": "vulcan",
                "WS 002": "vulcan"
            }
        },
        {
            "autofire": false,
            "mode": "LINKED",
            "weapons": {"WS 003": "harpoonpod"}
        }
    ],
    "wings": [
        "thunder_wing",
        "thunder_wing"
    ]
}
[close]

condor_Broad_Perd_Harpoon.variant
Spoiler
{
    "displayName": "Strike",
    "fluxCapacitors": 0,
    "fluxVents": 0,
    "goalVariant": true,
    "hullId": "condor",
    "hullMods": ["expanded_deck_crew"],
    "permaMods": [],
    "variantId": "condor_Broad_Perd_Harpoon",
    "weaponGroups": [
        {
            "autofire": true,
            "mode": "LINKED",
            "weapons": {"WS 001": "vulcan"}
        },
        {
            "autofire": false,
            "mode": "LINKED",
            "weapons": {"WS 003": "harpoonpod"}
        }
    ],
    "wings": [
        "broadsword_wing",
        "perdition_wing"
    ]
}
[close]
Logged

SCC

  • Admiral
  • *****
  • Posts: 1983
    • View Profile
Re: 0.9.1a Balance Testing Case Study: Condor
« Reply #42 on: July 12, 2020, 01:35:52 PM »

If you were using half bomber, half thunder Condors and had way more wins than I did, then I either unknowingly had other changes in my files I don't remember, or my Condors had one hell of a bad luck. I'll run some battles tomorrow and if they stay abysmal like that, then I'll have to reinstall my Starsector and check if that changes anything.
I had Combat Chatter, Console Commands, Fleet Tester, Flux Reticle, Lazylib, Magiclib, Practice Targets and Graphicslib on when doing the tests, but none of those should change the results.
I'm not sure how to tell what the admiral AI is doing.  At the beginning of the AI Battles mod or Fleet tester with the updated jar, there are some initial orders to claim nav/sensor bouys.  That shows up on the map layer.  Then the only other orders I see on the map layer are retreat orders.  The local ship AI is clearly assigning itself to escort though (I've seen drovers follow other ships that are retreating to the edge of the map, then once they retreat, head back to the fight).  And generally in the initial clash, most of the bombers go after the same ship, but then again, that might just have been the first ship seen, or they all had the same selection criteria at the beginning of the fight, so it tends to make them select a single target.
If you keep the map open and the game unpaused, you can see what admiral AI is doing. As I mentioned earlier, first it gives escort orders on cruisers, then dismisses those orders and orders assault on the objective, then dismisses that and only retreats individual ships. I don't think it's possible to give invisible orders, but I'm not so sure of it now.
The biggest benefit for me was that it acted sort of like the player, so I did not have to make any input myself.

Morrokain

  • Admiral
  • *****
  • Posts: 1445
  • Megalith Dreadnought - Archean Order
    • View Profile
Re: 0.9.1a Balance Testing Case Study: Condor
« Reply #43 on: July 12, 2020, 01:59:40 PM »

^ In regards to player piloting balance vs player admiral balance: (Bit of a tiny derail but I think it's relevant to the balance discussion)

The player piloting is much, much more effective as a whole. The player can make nuanced decisions about when to strike, what to tank on shields, armor, etc - and generally has a much better understanding of the tactical positioning of the enemy units and what can realistically punish the player's decision to go for the kill to a pressured or overloaded enemy ship.

Now, I'm not saying giving tactical orders are useless or terrible or anything. They *mostly* do their job and do it well. But there are two very notable exceptions to this:

1) Capturing strategic points.

2) Dealing with fighters.

This is because the AI always values staying alive over any player given orders at the moment. They retreat from capturing a position when enemies are in strong numbers, or move away from capturing a position to engage or chase enemies if they have the advantage. It is quite frustrating. Sometimes it goes well, don't get me wrong, but the times when they ignore my orders highly irritates me. I think the reason more players don't mind this behavior as much is because tactical points aren't very powerful right now. They do something, but unless you have a lot of them they are less noticeable in the overall scheme of things.

However, AI behavior when dealing with fighters, specifically, drives me absolutely crazy. It works in standard vanilla campaign balance but only there and it is the most inflexible part of the AI right now. It is balanced for a very, very specific use case: A small number of fighters mixed in with lots and lots of combat ships such that the combat ships can huddle together and reduce the fighter replacement rate or kite away and do the same. Large numbers of fighters break this concept, and they break it very hard. Also, it is difficult to conceptually create a faster reinforcing fighter without running into this issue. Even if the fighter is very fragile, as long as the carrier can send a couple or have one on the field at all times the AI will continue to backpeddle and never really make any ground against the carrier. It either leads to the warship dying or a stalemate situation that drastically slows the battle down until CR runs out. It quite literally takes active player intervention to speed the battle up and eliminate carriers.

Speaking of eliminate, this default behavior would actually be just fine if it weren't for the fact that an "Eliminate" command doesn't really do that much to help this. Standard AI should be conservative and focused on staying alive, but though the warship will be more aggressive while under an eliminate command, certainly, large numbers of fighters still lock it into a backpeddling position because the key assumption that the AI makes is that the carrier will eventually "run out of fighters to send" for some kind of window. When that assumption proves false the AI can't make the decision to just go for the kill and hope it makes it - which is what I think eliminate should be about. It needs to be a binary "this ship must die" order that assumes the player is ok with the ship in question from dying. When I give that order I am making a gamble to sacrifice ships to eliminate a high-threat target that I find to be more important to take out vs keeping those ships alive in the tactical sense of the battle. I don't really want them to be smart I want them to do their job and carry out my order because it's priority one in this circumstance.

If you keep the map open and the game unpaused, you can see what admiral AI is doing. As I mentioned earlier, first it gives escort orders on cruisers, then dismisses those orders and orders assault on the objective, then dismisses that and only retreats individual ships. I don't think it's possible to give invisible orders, but I'm not so sure of it now.
The biggest benefit for me was that it acted sort of like the player, so I did not have to make any input myself.

Is this changed from the vanilla admiral AI? If the Admiral AI (speaking purely on the enemy AI side here) isn't giving eliminate orders then I would argue that it is a limitation. Sometimes they are very necessary and this would provide nice ambush mechanics to challenge the player. The enemy can stand to be wasteful, unlike the player fleet (at least generally).
Logged

Hiruma Kai

  • Commander
  • ***
  • Posts: 177
    • View Profile
Re: 0.9.1a Balance Testing Case Study: Condor
« Reply #44 on: July 12, 2020, 02:08:51 PM »

If you keep the map open and the game unpaused, you can see what admiral AI is doing. As I mentioned earlier, first it gives escort orders on cruisers, then dismisses those orders and orders assault on the objective, then dismisses that and only retreats individual ships. I don't think it's possible to give invisible orders, but I'm not so sure of it now.
The biggest benefit for me was that it acted sort of like the player, so I did not have to make any input myself.

Oh you're right.  Looking mid-game doesn't help, but clearly right at the beginning there are some escort orders.

My mods list when doing this is: Fleet Tester 1.0
LazyLib 2.4
MagicLib 0.27
ZZ GraphicsLib 1.4.1

I'll do a reinstall as well and see if it my results change signficantly.  It'd be bad if I've got an unintentional edits somewhere that is affecting results.

Logged
Pages: 1 2 [3] 4 5 ... 9