WW2 BAD-T IV BDAc AI Dogfight Tournament

[Pds314]

Another example of a ram-to-kill with only modest damage. Note the bomb and mount just sitting there waiting for the killed plane to maneuver away from it.

[Pds314]

3000 rounds from repeated .50 cal ground attack runs. You can see several non-essential spoilers and fuselage or wing pieces are destroyed, but I then went over with the damaged plane and shot the ground attacker to pieces. More concentrated fire may well render it inoperable or at least without power but poorly-aimed fire drained the ammo store before the ground attacker could do fatal damage.

Old BAD-t: "You got hit by a single 20 mm, plane burnt down."
New BAD-t: "You hit the target with 732 .50 cal rounds, critical hit to central gear leg."

Edited October 6, 2018 by Pds314

[Guest]

The AI likes ramming now. TWICE in a row my planes collided in the merge.

[SuicidalInsanity]

Pds314: That many bullet holes in the bottom pic tells me something isn't configured right on your end, the only AA weapon that should leave that many holes is the ShKAS. AA weapon damage should look like this: 

I can think of three possibilities - One, your BDA settings damage mult isn't set to 750, but you've said it is, two, you're using an old version of the AA patch, but that would likely lead to noticeable NRE spam, or three, you've run into the BDA HP inflation bug where parts will gain massive increases to their hitpoints, where what once had 500 HP now has thousands, but all I can say about that is I've sometimes had it trigger after extended sessions of Vessel mover spawning and reverting to launch, but restarting KSP seems to fix it.
So. Two questions: 1) What are the hitpoint totals for the parts on your plane? 500-1500 range, or much higher?, and 2), if you turn on debug labels, open up the Alt-F12 console, and set up a vertical wing panel or something to shoot at from ~100m or so, what is the damage per shot the test weapon is doing?

Worst case scenario, the issue isn't resolved and you can use that added resilience to better track weapon accuracy, etc. against an opponent; even if damage remains broken on your end, your craft's armament should behave properly during the tournament matches.

dundun92: The AI occasionally likes to do that, yes. That's an issue better addressed in the BDA thread; my best guess based on my own experience is either the AI's made multiple passes and this time the RNG just got unlucky, or weapon ranges are low enough either plane doesn't have enough time to get a few shots off to get the other to dodge and break off soon enough

Edited October 7, 2018 by SuicidalInsanity

[Pds314]

8 hours ago, SuicidalInsanity said:

Pds314: That many bullet holes in the bottom pic tells me something isn't configured right on your end, the only AA weapon that should leave that many holes is the ShKAS. AA weapon damage should look like this: 

I can think of three possibilities - One, your BDA settings damage mult isn't set to 750, but you've said it is, two, you're using an old version of the AA patch, but that would likely lead to noticeable NRE spam, or three, you've run into the BDA HP inflation bug where parts will gain massive increases to their hitpoints, where what once had 500 HP now has thousands, but all I can say about that is I've sometimes had it trigger after extended sessions of Vessel mover spawning and reverting to launch, but restarting KSP seems to fix it.
So. Two questions: 1) What are the hitpoint totals for the parts on your plane? 500-1500 range, or much higher?, and 2), if you turn on debug labels, open up the Alt-F12 console, and set up a vertical wing panel or something to shoot at from ~100m or so, what is the damage per shot the test weapon is doing?

Worst case scenario, the issue isn't resolved and you can use that added resilience to better track weapon accuracy, etc. against an opponent; even if damage remains broken on your end, your craft's armament should behave properly during the tournament matches.

dundun92: The AI occasionally likes to do that, yes. That's an issue better addressed in the BDA thread; my best guess based on my own experience is either the AI's made multiple passes and this time the RNG just got unlucky, or weapon ranges are low enough either plane doesn't have enough time to get a few shots off to get the other to dodge and break off soon enough

My plane is tougher than most (I increased the wing mass mass and number of connections on many structurally-important parts to deliberately boost their health, some to several thousand like the root part and the wing spars, the plane is quite heavy for its size, but most parts that aren't structurally important have 500-1500 HP).

As I said, it took only one pass with the damaged plane to kill the ground attacker, even though it had some parts with 7000 hp. 20 mm does reasonably-significant damage with a few good hits to the same part.

That being said, it does seem like my dummies take quite a few hits to kill.

Not sure if I have an outdated AA patch.

Edited October 7, 2018 by Pds314

[Pds314]

Advice on how to get the AI to aim well without sacrificing too much maneuverability, energy retention, or stability? I am seeing it fire 2 or 3 seconds continuously 20+ meters off target. Not a spray of bullets with a CEP of 20 meters but a death ray* a meter wide focused 20 meters from where it needs to be. The only way I can see to fix it is to reduce damping and ki and increase steering, but this is quite risky as many planes will stall, flail, or sideslip with such an aggressive pilot tuning.

 

*it SHOULD be a death ray, most of my designs use 15-23mm so far and have all nose-mounted guns. In war thunder it would kill in the scenarios I'm seeing where it manages to hit. It's not in my KSP because something is wrong.

Edited October 8, 2018 by Pds314

[Guest]

3 hours ago, Pds314 said:

Advice on how to get the AI to aim well without sacrificing too much maneuverability, energy retention, or stability? I am seeing it fire 2 or 3 seconds continuously 20+ meters off target. Not a spray of bullets with a CEP of 20 meters but a death ray* a meter wide focused 20 meters from where it needs to be. The only way I can see to fix it is to reduce damping and ki and increase steering, but this is quite risky as many planes will stall, flail, or sideslip with such an aggressive pilot tuning.

 

*it SHOULD be a death ray, most of my designs use 15-23mm so far and have all nose-mounted guns. In war thunder it would kill in the scenarios I'm seeing where it manages to hit. It's not in my KSP because something is wrong.

The aiming is a BDA issue. As far as tweaking the AI, look at the link in my signature about "building the most competitive BDArmory Fighter aircraft". Although the guide is designed for modern drones, the AI tweaking tips can be applied to WW2 aircraft. 

[Pds314]

Question as to victory conditions:

 

If my plane runs out of gas, while the other is flying with fuel but not ammo, and mine lands itself intact, is that a draw or do I lose?

19 minutes ago, dundun92 said:

The aiming is a BDA issue. As far as tweaking the AI, look at the link in my signature about "building the most competitive BDArmory Fighter aircraft". Although the guide is designed for modern drones, the AI tweaking tips can be applied to WW2 aircraft. 

I have read your tutorial maybe 3 or 4 times. It's interesting. In fact, it's what made me decide to reinforce the heck out of all wing spars.

Edited October 9, 2018 by Pds314

[Pds314]

Okay apparently my planes CAN tear the dummy to bits, they just need to concentrate fire on it, which is not likely to happen in a head-on pass.

Shadowbird Mk II vs dummy:




Shadowbird Mk IV vs dummy:

 

I wonder if a flying wing would be competitive?

On an aside, it has a single-piece mostly non-protective canopy that can open and close (although it slides sideways).

Edited October 9, 2018 by Pds314

[SuicidalInsanity]

Up to 7000 HP? No wonder stuff took forever to die in your tests. Just keep in mind the Eidahill Clause.

Re: Victory conditions. I don't think it's come up before, so I'm unsure as to what precedent is in this case. While gut feeling is last plane in the air is the victor, that makes for a boring fight if it comes down to watching a plane run out of fuel, and choosing how much ammo to bring is an important design consideration. As such, I'm going to rule that if one side's planes run out of ammo, they immediately forfeit the match, as they no longer have any means to shoot down their opponents, short of RNG kamikazi.

[Pds314]

On another note. Someone earlier was saying how I wasn't gonna build a craft that could potentially knock out a top level pilot.

Maybe not, but anyone else would DEFINITELY have lost consciousness in this turn.

16.5 G turn at 227 m/s on the initial merge. 360 degree turn in 10 seconds, less 76 m/s. Granted, the turn is not something that could be sustained indefinitely, although it did gain some altitude overall.

56 minutes ago, SuicidalInsanity said:

Up to 7000 HP? No wonder stuff took forever to die in your tests. Just keep in mind the Eidahill Clause.

Re: Victory conditions. I don't think it's come up before, so I'm unsure as to what precedent is in this case. While gut feeling is last plane in the air is the victor, that makes for a boring fight if it comes down to watching a plane run out of fuel, and choosing how much ammo to bring is an important design consideration. As such, I'm going to rule that if one side's planes run out of ammo, they immediately forfeit the match, as they no longer have any means to shoot down their opponents, short of RNG kamikazi.

This happens VERY frequently with the dummy. It really can't do meaningful damage to anything past 400 meters or so but it will fire away what should be a rather massive amount of ammunition (900 rounds per gun) in about 8 minutes of flying. WRT 7000 HP, that was an accident lol. The ground attack plane just happened to have tons of stuff bolted on to an ordinary Mk1 cockpit. Even the dummy though has some parts with pretty crazy levels of HP, including the cockpit.

The planes I've built that are actually competition-worthy have most parts at HP 500-1500, but some wing spars and base plates for the cockpit and wings and such exceed this. The spars on my Shadowbird Mk IV have about 2000-3000 HP each but are rather small targets in addition to being tough. My other plane had spars and its spine up to 4500 HP. That being said, these parts are heavy. You could not just go around building an entire plane out of super narrow mass/strength 2.5 wings 5 meters long with more wings attached and still expect it to fly well, and even then, someone will eventually engine-snipe you I should think.

If you want an example of super crazy parts, some of the plates on my 27-tonne flying tank plane spawn with up to 13000 HP. Considering it is multiple plates thick in some areas, it actually takes like 5-8 75mm main gun rounds to destroy a section of the frontal armor. I.E. it could resist a Sherman tank shell.

One case I can think of where running out of bullets doesn't cause the plane to be incapable of killing that's practically unique to the dummy is if one side runs out of fuel and the other runs out of ammo, the AI will attempt to BOMB anything that lands on the ground but is still considered a valid target. This actually happened in my earlier test and it successfully killed Val after she did a crash-landing (ok no she ran straight into the ground at 450 mph and managed to survive, it wasn't really a landing) due to diving too fast.

Edited October 9, 2018 by Pds314

[Pds314]

Jeb wins the terrible lottery of being AI module sniped, decides 400 mph dive straight into the ground is the best solution.

(Of course, being Jeb, he survived the impact, even if most of his plane did not).

[Pds314]

Val has taught us all an important lesson: if you land on water you will sink and be consumed by the Deep Sea Kraken.



However, Val managed to escape her doomed aircraft as it was being consumed and began moving toward the surface.



Val is moving quickly toward the surface. Perhaps TOO quickly. Is 34.4 m/s a good speed to ascend at while diving that deep?



Val is launched from the sea and pulls her chute to stop her from falling back into it too hard.

Edited October 9, 2018 by Pds314

[Pds314]

Map of the fight: all four Kerbals survived (although two had significant injuries due to bullet wounds).

Top: Val and an Ammo box that survived the sea Kraken.
Bottom: Jeb's crash sight due to AI module sniped.
Right: Injured Kerbal ran out of fuel over the water but my other plane floats nose up, barely.
Left: Damaged plane with slightly injured Kerbal did a perfect landing without fuel under AI control.

Edited October 9, 2018 by Pds314

[Pds314]

That might be a kill.



OKAY THAT'S DEFINITELY A KILL!

 

Notice the lack of a Kerbal or any fuel in the remaining debris.

Edited October 9, 2018 by Pds314

[Pds314]

>No elevator.
>Missing fuselage sections
>Folding in half even under 1 G.
>Missing Aileron
>Missing most of the wing area.
>Missing part of horizontal stabilizer.
>Cockpit is crushed with pilot leaning out the window.
>Somehow not dead.



But of course, then we have things like this... clip off a wing? Not a problem.



TFW you're missing a wing and can still dodge bullets so well your opponents both run out of ammo and you have >700 rounds left but probably can't hit anything with them.

Edited October 9, 2018 by Pds314

[53miner53]

Working on my entry, "Delta". 

https://drive.google.com/open?id=1Z6BopQUl-U1IyQCYBsla4FujL0vnSmPg

Somehow, Kerbals will survive a crash in this plane at 50+m/s head on. Or when landing...

Also, If you shoot off the tail from the Dummy craft, it can still fly. 

[sturmhauke]

I know what a wing spar is in real life, but what does it mean in this context? Is that like using a narrow Pwing part in the center, and attaching leading and trailing Pwing panels to it?

[SuicidalInsanity]

Pretty much.

A unnecessarily over-engineered example taking the concept up to eleven, but the base concept remains the same. Reinforced structural element supporting wing elements, and lighter weight wing sections added to it. The idea is even if a wing section takes enough damage to be destroyed, the wing as a whole remains more-or-less functional vs. having half or more of a wing suddenly detach from the aircraft. It will help against smaller munitions, against 20+mm HE cannons, perhaps not so much.
If you choose to go this route, just keep it under 60 parts please.
 

[sturmhauke]

I think I'm ready for the competition now.

Ok but seriously though. I'm a little surprised it actually got off the ground, but now it's flexing all over the place. Less so if I use autostruts, but it still flaps like a weird skeletal bird.

[Joseph Kerman]

On 10/9/2018 at 8:39 AM, Pds314 said:

Question as to victory conditions:

 

If my plane runs out of gas, while the other is flying with fuel but not ammo, and mine lands itself intact, is that a draw or do I lose?

2

It's the opponent's victory; I had made a jet in Season 3 that runs out of fuel before my opponent. He won only because of that, since my guns are better.

[SuicidalInsanity]

Sturmhauke: Remember that twin engine craft need to meet a minimum weight of 5000kg, and Kerbals in lawnchairs need to be more-or-less inside a constructed cockpit. Since that's a more subjective judgement I really should post some examples of the things I'm looking for.

 

[sturmhauke]

21 minutes ago, SuicidalInsanity said:

Sturmhauke: Remember that twin engine craft need to meet a minimum weight of 5000kg, and Kerbals in lawnchairs need to be more-or-less inside a constructed cockpit. Since that's a more subjective judgement I really should post some examples of the things I'm looking for.

 

Yeah, I know. That was a joke, not a finished airframe. The idea was to test the basic geometry and then put more stuff on top. What is with the flexing though? It's not visible in that picture, but at the moment I do any maneuvers the wings and structural parts bend way more than stock parts would, even with autostruts on.

[SuicidalInsanity]

Fair enough. Flexing generally occurs when something heavy is attached to something light, and in the pic it looks like you have lightweight wing spars connected to lightweight fuselage elements, with decently large wings attached to the spars. Any sort of maneuver is going to put the kiloNewtons of force generated by lift and drag of those wings on that relatively lightweight joint and cause it to flex. Best answer I can give from here is bulk up the strength of major joints (bigger mass of structural elements/struts), or reengineer where those joints are (moving the fueltanks inwards to the center and using their relatively greater mass to root the wing spars, for instance) to reduce joint flexibility.

[SuicidalInsanity]

A note on Constructed Cockpits, and what I'm looking for:

1: No. No windscreen, no protection, kerbal cannot, by any stretch of the imagination, be considered "in a cockpit".
2: Modest improvement, but no. Kerbal is still mostly fully exposed.
3: Edge case, but acceptable. Kerbal somewhat exposed, but cockpit structure is recognizable.
4: Ideal constructed cockpit. Kerbal is recognizably in a cockpit, and has good protection from both airflow and incoming fire. (Yes, kerbals can be killed, and it will result in the plane crashing).
5: Kerbal is fully enclosed within a constructed canopy. Also an ideal cockpit.