WW2 BAD-T III - BDAc AI Dogfight Tournament

[Matuchkin]

On 09/12/2017 at 9:16 PM, SuicidalInsanity said:

The Wasp is quite the difficult opponent to counter, I eventually had to resort this:

 

Not a fair fight by any standards. You pretty much decimated that propeller fighter with your Korean-War era jet.

[SuicidalInsanity]

Looks like it's time for the Failure Reel.

 

[Matuchkin]

@SuicidalInsanity well look, I understand that the Wasp is very powerful and difficult to beat, but at least this challenge will keep its integrity (and challenging quality) if no one attempts to throw the equivalent of a cold-war MiG at it.

[tetryds]

More updates about BAD-T WW1 rules:

The minimum wing strength will probably be lowered, since most of airplanes will have much more strict weight balancing (in other news, water is wet).

7mm weapons must be able to deal damage even when there are only two of them, I intend to balance it in such a way that the weapon count does not go past three weapons for fighters.

Since now part strength seems to not be tied to the mass anymore (or something like that), it will have to be reviewed, which is kind of a pain but fine (@SuicidalInsanity help plox )

Ammo won't have to be limited, but beware that weight is now precious, and low caliber weapons have a high fire rate.

It will not require size limitations.

No armor, even if we get heavy fighters, BAD-T III showed us that the armor system is not balanced well enough to be used reliably at this moment.

The point system will have to be overhauled, I am currently studying alternatives, and will try to keep it as close to the current implementation as possible.

 

Edit: It may seem like things are getting simpler, less engines and weapons, fewer weapon and engine choices, but that is exactly the point, we are going to be so busy with weight and performance that we cannot spend too much time picking a good setup and iterating between them, we will be iterating just as much, to make it fly graciously.

Edit2: Also, triplanes look awesome, why not?

Edited December 22, 2017 by tetryds

[Van Disaster]

1 hour ago, tetryds said:

Edit2: Also, triplanes look awesome, why not?

Unsure about FAR wing interaction with those ( biplanes work if you can keep wings far enough apart at least ) - although I never did get a clear picture of FAR & close coupled wing interaction other than some loss of lift. In theory, absolutely for triplanes!

[SuicidalInsanity]

2 hours ago, tetryds said:

7mm weapons must be able to deal damage even when there are only two of them, I intend to balance it in such a way that the weapon count does not go past three weapons for fighters.

Since now part strength seems to not be tied to the mass anymore (or something like that), it will have to be reviewed, which is kind of a pain but fine (@SuicidalInsanity help plox )

I can help with that. I'm guessing the intent is to use the BDA beta? The change to a HP system in the beta means ensuring 7mm weapons do damage is easier, since now even chip damage is persistent and accounted for, rather than potentially radiating away under the current heat system, though some fiddling with part resilience will be needed either way.

[DoctorDavinci]

23 minutes ago, SuicidalInsanity said:

I can help with that. I'm guessing the intent is to use the BDA beta? The change to a HP system in the beta means ensuring 7mm weapons do damage is easier, since now even chip damage is persistent and accounted for, rather than potentially radiating away under the current heat system, though some fiddling with part resilience will be needed either way.

With the beta you can now change the armor thickness and hitpoints of each part through a config file which should help with making everything balanced

[SuicidalInsanity]

2 hours ago, DoctorDavinci said:

With the beta you can now change the armor thickness and hitpoints of each part through a config file which should help with making everything balanced

Can you elaborate? I've tried reducing the default armor value to something below 10 via MM and custom configs, but neither method works. Asking because trying to do some basic benchmark damage tests for 7mm weapons results in all of them doing ~ 0.01 damage per hit, compared to ~ 60 for .50 cals, and I suspect that even at a measly 10, the  intrinsic part armor is responsible.

[DoctorDavinci]

10 hours ago, SuicidalInsanity said:

Can you elaborate? I've tried reducing the default armor value to something below 10 via MM and custom configs, but neither method works. Asking because trying to do some basic benchmark damage tests for 7mm weapons results in all of them doing ~ 0.01 damage per hit, compared to ~ 60 for .50 cals, and I suspect that even at a measly 10, the  intrinsic part armor is responsible.

I'll do up a base config for you that will plaster all parts with custom armor thickness and hitpoints 

One thing to note is that the lower caliber weapons won't do very much if there is armor ... Ill set the thickness to 1 and set a low hitpoint value, hopefully this will help mitigate the issue

Also it is still a beta and we are still getting new builds every day or so from the guru's 

EDIT - Here's the config ... should work as is to change the hitpoint and armor values of all parts (HP=100 and Armor=1)

Quote

@PART[*]:NEEDS[BDArmory]
{
    MODULE
    {
        name = HitpointTracker
        HitPoints = 100
        maxHitPoints = 100
        Armor = 1
        ArmorThickness = 1
        ExplodeMode = Never
    }
}

 

EDIT 2 - I also queried the lads and this might help as well:

Quote

probably need to adjust apBulletMod = 0 upwards, BUT 7mm did/do no reall damage unless you are real close, it's passable on bare unarmored parts, but on anything else, ie normal aero parts you'll be fining all day and only kill by attrition

So dropping the armor to 0 sounds like what would be needed here as well as some tweaking of the bullet modifier in the gun part config

EDIT 3 - So apBulletMod has not been implemented yet, however the bulletdamage mult should be able to do what you need by increasing the damage of the round itself ... add this to the weapon cfg and adjust to suit, bulletDmgMult = 1.3 (SMI .50's use this modifier)

Send me a PM and let me know how I can help if you need further assistance

Edited December 22, 2017 by DoctorDavinci

[DoctorDavinci]

I forgot to add a caveat to the above ... there are some stock aero parts that needed a custom HP patch, the patch can be found in the MMpatches directory in BDArmory

For ease I am posting in the spoiler below ..

Spoiler

///////////////////////////////////////////
//Fixes for Mass/Density anomoalies in KSP
///////////////////////////////////////////

@PART[turboJet]
{
    MODULE
    {
        name = HitpointTracker
        ArmorThickness = 10
        maxHitPoints = 2000
        ExplodeMode = Never
    }    

}

@PART[elevon*]
{
    MODULE
    {
        name = HitpointTracker
        ArmorThickness = 10
        maxHitPoints = 500
        ExplodeMode = Never
    }    

}

@PART[B9.Aero.Wing.Procedural.TypeB]
{
    MODULE
    {
        name = HitpointTracker
        ArmorThickness = 10
        maxHitPoints = 500
        ExplodeMode = Never
    }    

}

@PART[winglet3]
{
    MODULE
    {
        name = HitpointTracker
        ArmorThickness = 10
        maxHitPoints = 650
        ExplodeMode = Never
    }    

}

@PART[tailfin]
{
    MODULE
    {
        name = HitpointTracker
        ArmorThickness = 10
        maxHitPoints = 750
        ExplodeMode = Never
    }    

}

@PART[AdvancedCanard]
{
    MODULE
    {
        name = HitpointTracker
        ArmorThickness = 10
        maxHitPoints = 750
        ExplodeMode = Never
    }    

}

@PART[CanardController]
{
    MODULE
    {
        name = HitpointTracker
        ArmorThickness = 10
        maxHitPoints = 750
        ExplodeMode = Never
    }    

}

@PART[nacelleBody]
{
    MODULE
    {
        name = HitpointTracker
        ArmorThickness = 10
        maxHitPoints = 750
        ExplodeMode = Never
    }    

}

These parts will either need to be directly edited in the above config (000000_HitpointModule_PartFixes) or have an editing MM config that is set to FINAL

Edited December 22, 2017 by DoctorDavinci

[tetryds]

5 hours ago, DoctorDavinci said:

probably need to adjust apBulletMod = 0 upwards, BUT 7mm did/do no reall damage unless you are real close, it's passable on bare unarmored parts, but on anything else, ie normal aero parts you'll be fining all day and only kill by attrition

The bullets have to be adjusted in a way that a pair of 7mm weapons can take down an aircraft through air superiority.

I don't give a damn if that is realistic or not, that is how it has to be, and is kinda how it actually was, but since AI is not smart enough to shoot at the important spots, it has to inflict damage anywhere it hits.

[Van Disaster]

7 hours ago, tetryds said:

The bullets have to be adjusted in a way that a pair of 7mm weapons can take down an aircraft through air superiority.

I don't give a damn if that is realistic or not, that is how it has to be, and is kinda how it actually was, but since AI is not smart enough to shoot at the important spots, it has to inflict damage anywhere it hits.

That would need it to know what an important part actually is, which goes a bit beyond just the autopilot. Could make it shoot at engines easily enough ( I will have a look at that, might simply be a case of taking an engine transform instead of CoM ) but anything else you'd have to mark up in a part config. Not sure how you'd mark up a wing root, shape recognition is a little way down the road I think...

Anyone actually looked at how well WW1 aircraft guns were mounted? I have a suspicion they vibrated a fair bit, I bet they weren't terribly accurate.

Edited December 23, 2017 by Van Disaster

[SuicidalInsanity]

The armor reduction doesn't help, nor did the Bulletdmgmod, 7mm weapons still do 0.01 dam/hit. There's going to need to be some serious HP reduction for a WWI/interbellum tourney, even equipping test craft with dual Berezin UBZs it still took dozens of hits to destroy a part. Though if nothing else, I at least have a damage benchmark to work with now.

Speaking of AI smarts, the other main issue I noticed during tests was the AI's newfound suicidal insanity. Have there been major changes to the AI?

How a typical match ended. Nice to see the AI recruits only the finest Ork fighter pilots. 
Planes ineffectually shoot at each other for ~10 minutes, before deciding to ram into each other, usually disabling one plane and knocking the pilot out of the other. If by some chance both craft survived without major damage, then they'd generally try it again a few minutes later.
 

[Guest]

2 hours ago, SuicidalInsanity said:

The armor reduction doesn't help, nor did the Bulletdmgmod, 7mm weapons still do 0.01 dam/hit. There's going to need to be some serious HP reduction for a WWI/interbellum tourney, even equipping test craft with dual Berezin UBZs it still took dozens of hits to destroy a part. Though if nothing else, I at least have a damage benchmark to work with now.

Speaking of AI smarts, the other main issue I noticed during tests was the AI's newfound suicidal insanity. Have there been major changes to the AI?

How a typical match ended. Nice to see the AI recruits only the finest Ork fighter pilots. 
Planes ineffectually shoot at each other for ~10 minutes, before deciding to ram into each other, usually disabling one plane and knocking the pilot out of the other. If by some chance both craft survived without major damage, then they'd generally try it again a few minutes later.
 

Funny you said that. TWICE in a row, my jets crashed into each other in the beta.

[Eidahlil]

6 hours ago, SuicidalInsanity said:

Speaking of AI smarts, the other main issue I noticed during tests was the AI's newfound suicidal insanity. Have there been major changes to the AI?

I think there haven't been any major changes (mandatory "but I am not a dev"), and the suicidal insanity is due to a bug in guard mode "firingAtMe" detection. Previously the planes would get out of the way early because they would be dodging each other's gunfire, now planes mostly don't dodge gunfire. The detection angles are too small because of a forgotten square root. Pretty sure it was already in the 1.0 release.

The collision detection, with a detection radius of 30m and a prediction interval of 0.5s firing for planes heading at each other at 100m/s each will mostly miss (70%), but it has been this way for a long while. It's fired every 20 updates, so it has multiple chances of detecting the other one, but in a protracted fight it's very likely to fail at some point.

[SuicidalInsanity]

Making progress zeroing in weapon damage, definitely getting a feel for how the new ballistics system works. tetryds, about how many rounds do want a typical kill to take?

Also, discovered a new way for test matches to end: Pilot KIA. Appropriate, given the target era...

 

[Van Disaster]

We may have to up the part limit a bit...

[tetryds]

10 hours ago, SuicidalInsanity said:

Making progress zeroing in weapon damage, definitely getting a feel for how the new ballistics system works. tetryds, about how many rounds do want a typical kill to take?

An average sized wing piece should take about 4 or 5 shots to be taken down. Let's use that as a baseline, then adjust from there, because it's not that simple, you know.

 

10 hours ago, SuicidalInsanity said:

Also, discovered a new way for test matches to end: Pilot KIA. Appropriate, given the target era...

That is super cool.

 

Edit:

About parts, we may have to have custom struts, the current ones weight 50kg each, which is a LOT.

And I don't think the part count has to be increased, most previous BAD-T tournaments had crafts well under the limit, and when two aircraft with part count close to the limit flew the framerate started to suffer a bit.

Edited December 24, 2017 by tetryds

[Van Disaster]

Strutting a triplane drives the part count through the roof, unfortunately ( as well as actually building triplane wings ). At least we only have two guns.

[Guest]

11 hours ago, SuicidalInsanity said:

Also, discovered a new way for test matches to end: Pilot KIA. Appropriate, given the target era...

"Pilot knocked out, hold J to revive"

Spoiler

War Thunder memes at their finest

 

[SuicidalInsanity]

Part count isn't too bad - the red triplane is 45 parts, but uses single segment wings; the kamikaze biplane was 80, but that was due to some crazy over-engineered wings, redoing the wings brought it down to 50. There's also some MM mass adjustment at work because, yes, 50kg per strut is ridiculous; the triplane has a dry mass of 655kg, the biplane 554kg,
Here's an alpha of what I've got so far of MM mass tweaks/weapon calibration/AirplanePlus engine configs that I've been using in my tests.

[Alioth81]

On 23.12.2017 at 2:04 PM, Eidahlil said:

planes mostly don't dodge gunfire. The detection angles are too small because of a forgotten square root. Pretty sure it was already in the 1.0 release.

I can more or less confirm this.
When I designed a (non BAD-T III compliant) plane that could shoot down the Wasp I used BDc 1.0 and the Red Arrows would reliably take at least one Wasp down first pass.
Once I downgraded to 0.3 the planes started showing constant "evasive" behaviour in the debug text and did barrell rolls until they were shot down instead of facing the danger :-)

 

 

[Joseph Kerman]

A cool example for WWI designs: the P-1

[SuicidalInsanity]

I think I've gotten part/engine/weapon configs just about finalized.

 

[Van Disaster]

Ver' nice. Is the audio actually synced with the fire rate though? guns fire rate is kinda high if it is...