WW2 BAD-T 2 - BDA AI Dogfight Tournament [SEMIFINALS]

[tetryds]

Won't spoil, but final exams... I'm so sorry D:

[gridghost]

NP, finals are a tinsy bit more important

[Pds314]

Interesting point here. I tried throwing together a compliant plane with the latest version of everything, cut the power on the Brutus to 50% because I wanted something like a Spitfire Mk.XIV, and BAD-T2 engines are extremely OP, as even the most powerful World War 2 fighter engines with 2000+ HP could output MAYBE 15 kN of thrust at peak thrust, and likewise, even the huge Double Wasp on a Corsair weighs a bit more than a tonne, not 1.4 like our biggest engines. Made a resized and re-spec'd Juno, and had it fight an Me-262.

The results were.. Interesting. The AI DOES NOT understand how to properly fly WWII-era jets. I could pretty well boom and zoom in the Schwalbe enough to kill the Spitfire (or several of them) without risking much, but the AI seemed to have a tendency to run away, then make a sharp turn and put the Griffon Spitfire back on its tail repeatedly. Changing the settings didn't help because at high settings, the AI would stay just outside 700 meter gun range, while at low settings, the AI couldn't track the Spitfire once it zoomed in for the "kill." It's not that the AI can't aim, it's actually a while lot better at it than I am, and if given the opportunity, it'll get a kill from a 1 or 2 second burst, whereas early in I was firing up to 200 shots of mk108 30mm per kill. Not good, since I only gave it 320 rounds of ammo. Again, cannot stress that with the 108, however much you feel like leading, lead a lot more. I sometimes feel like I've missed and then the Spitfire goes boom. It doesn't help that the 262 burns ammo like an A-10. 2400 30mm shells a minute... With 8 seconds worth of ammo... (8+2 of half-output IRL, but I didn't want to add a third ammo box.)

 

I wonder if Dynamic deflection could be used to "teach" the AI to keep above ~350-400 mph when flying a jet and only turn sharp when it has pretty good speed advantage. I know it can help with 4.5/5th gen fighters that have crazy amounts of pitch authority. That's not exactly the issue here, but I feel it might help. The only two AI victories I saw against the Spit was first when the Spit.. Went.. For a head-on-pass with the 262... Yeah. You're not coming out of that... And second when the Spit tried to follow the 262 into a vertical dive at top speed despite having about half the wing strength. Again, good friggin' luck.

 

On a related note, maybe I should make some Яussian Bias planes, IDK, a Yak-15P or maybe a MiG-15bis. MiG-15bis vs. Spitfire Mk24 or Me-262 is fair, right?

 

Wait.. Holy heck! This is KSP. I know what plane to make! An Me-163!

Edited June 29, 2016 by Pds314

[Wjolcz]

16 hours ago, Pds314 said:

Interesting point here. I tried throwing together a compliant plane with the latest version of everything, cut the power on the Brutus to 50% because I wanted something like a Spitfire Mk.XIV, and BAD-T2 engines are extremely OP, as even the most powerful World War 2 fighter engines with 2000+ HP could output MAYBE 15 kN of thrust at peak thrust, and likewise, even the huge Double Wasp on a Corsair weighs a bit more than a tonne, not 1.4 like our biggest engines. Made a resized and re-spec'd Juno, and had it fight an Me-262.

The results were.. Interesting. The AI DOES NOT understand how to properly fly WWII-era jets. I could pretty well boom and zoom in the Schwalbe enough to kill the Spitfire (or several of them) without risking much, but the AI seemed to have a tendency to run away, then make a sharp turn and put the Griffon Spitfire back on its tail repeatedly. Changing the settings didn't help because at high settings, the AI would stay just outside 700 meter gun range, while at low settings, the AI couldn't track the Spitfire once it zoomed in for the "kill." It's not that the AI can't aim, it's actually a while lot better at it than I am, and if given the opportunity, it'll get a kill from a 1 or 2 second burst, whereas early in I was firing up to 200 shots of mk108 30mm per kill. Not good, since I only gave it 320 rounds of ammo. Again, cannot stress that with the 108, however much you feel like leading, lead a lot more. I sometimes feel like I've missed and then the Spitfire goes boom. It doesn't help that the 262 burns ammo like an A-10. 2400 30mm shells a minute... With 8 seconds worth of ammo... (8+2 of half-output IRL, but I didn't want to add a third ammo box.)

 

I wonder if Dynamic deflection could be used to "teach" the AI to keep above ~350-400 mph when flying a jet and only turn sharp when it has pretty good speed advantage. I know it can help with 4.5/5th gen fighters that have crazy amounts of pitch authority. That's not exactly the issue here, but I feel it might help. The only two AI victories I saw against the Spit was first when the Spit.. Went.. For a head-on-pass with the 262... Yeah. You're not coming out of that... And second when the Spit tried to follow the 262 into a vertical dive at top speed despite having about half the wing strength. Again, good friggin' luck.

 

On a related note, maybe I should make some Яussian Bias planes, IDK, a Yak-15P or maybe a MiG-15bis. MiG-15bis vs. Spitfire Mk24 or Me-262 is fair, right?

 

Wait.. Holy heck! This is KSP. I know what plane to make! An Me-163!

Well... The AI doesn't seem to know how to conserve energy (afaik). The engines were purposely made OP just so the AI doesn't struggle all the time. The guns are also not realistic, because if they were everyone would simply take the best gun. They were made the way they are to keep the balance of the tournament.

[tetryds]

3 minutes ago, Veeltch said:

The guns are also not realistic, because if they were everyone would simply take the best gun. They were made the way they are to keep the balance of the tournament.

How dare you. No, that is wrong, the guns are as realistic as they can possibly be, and are balanced out on specs that have nothing to do with bullet speed, muzzle velocity, fire rate and such.

Heck, even the tracer colors are accurate.

[Wjolcz]

Just now, tetryds said:

How dare you. No, that is wrong, the guns are as realistic as they can possibly be, and are balanced out on specs that have nothing to do with bullet speed, muzzle velocity, fire rate and such.

Oh dear lord! Please forgive me! I didn't know! ;_;

That's actually pretty cool to know though. I always thought they were made balanced just so one type of gun doesn't own everything else.

[The Optimist]

5 minutes ago, tetryds said:

How dare you. No, that is wrong, the guns are as realistic as they can possibly be, and are balanced out on specs that have nothing to do with bullet speed, muzzle velocity, fire rate and such.

Heck, even the tracer colors are accurate.

wow that's a savage roast

[Pds314]

13 minutes ago, Veeltch said:

Oh dear lord! Please forgive me! I didn't know! ;_;

That's actually pretty cool to know though. I always thought they were made balanced just so one type of gun doesn't own everything else.

If the guns were balanced on things like rate of fire, the SHkas wouldn't be far and away the fastest-firing 7.x mm cannon in existence and do practically the same damage as 

Edited June 29, 2016 by Pds314

[tetryds]

@Pds314 about dynamic deflection, it would be interesting to see how it does with the AI.

Not trying to hype anyone while still doing so, my last test is this Friday.

@Veeltch I understand, it kinda seems like so, it took me about a month or more to balance these weapons and that worked out better than I expected.

[Pds314]

4 hours ago, tetryds said:

@Pds314 about dynamic deflection, it would be interesting to see how it does with the AI.

Not trying to hype anyone while still doing so, my last test is this Friday.

@Veeltch I understand, it kinda seems like so, it took me about a month or more to balance these weapons and that worked out better than I expected.

Out of curiosity, what *are* they balanced on? Accuracy? Overheat? Aerodynamics? Impact damage? I know that, IRL, late war Spitfires had 7.7mm, Mustangs had 12.7mm, A whole variety of planes, including Spitfires, had 20mm, and many German planes had 30mm. A few planes here and there had bigger cannons, but what's the point? a few 30mm hits will kill a bomber if it hits anything critical. And a 30mm vs. a fighter is either going to spark if it hits something weak or blow the thing to kingdom come if it hits meat or metal of any description. Remember that a Mk108 round weighs about the same as an A-10 round (EDIT: no it doesn't, that's the Mk103 round, which I'm fairly sure isn't in the pack), albeit with barely over half the muzzle velocity. That should be more than enough to kill a World War 2 fighter if it hits anything substantial.

(Of course, for ground attack, the big cannons really shine. Like, considering many WWII tanks would be penetrated by a 40mm to anything but the frontal armor.., although even so, I'd still probably opt for the "spray and pray" approach available to something like a 262 armed with four 30mm vs. the sniping approach of the "Narwhal" 262 armed with the 50mm cannon)

Wow, I realize that the Me-262 is like an A-10 if the A-10 had poor accuracy and reliability and would turn into a fireball if hit with so much as a rifle round.. 

Edited June 30, 2016 by Pds314

[tetryds]

@Pds314 They are mainly balanced on overheat and cooldown time and bullet spread.

Since we don't have different types of ammo, BDArmory received updates to make them more balanced, using bullet energy instead of momentum, this also helped to soften a possible russian bias.

Now with those changes you get something more similar to what would happen in real life, while still having a balanced gameplay.

The final tweak though was the points system, which also worked surprisingly well.

[Pds314]

25 minutes ago, tetryds said:

@Pds314 They are mainly balanced on overheat and cooldown time and bullet spread.

Since we don't have different types of ammo, BDArmory received updates to make them more balanced, using bullet energy instead of momentum, this also helped to soften a possible russian bias.

Now with those changes you get something more similar to what would happen in real life, while still having a balanced gameplay.

The final tweak though was the points system, which also worked surprisingly well.

Яussian bias might also apply somewhat to the differences in altitude performance of the engines. Again, since the AI has trouble maintaining energy and all battles are at nearly point blank, It's unlikely that high-altitude planes will even reach their desired altitudes, instead, they're rapidly clubbed down to the deck.

I also wonder if a bit of forced dynamic deflection on the AOA of all surfaces could, somewhat, mimic how non-boosted controls get "heavy" at high speeds or during a dive. I guess the problem there is that IRL, they solved this problem with tabs, whereas making rules about tab shape and size and forced dynamic deflection to simulate control stiffness would be hideously complex, and ultimately futile without better AI to maintain enough energy to get the required dynamic pressure stiffen the controls.

On a related note, is there any mod with a decent set of WW2 fighter cockpits? The stock cockpits are lacking. The closest thing is the Mk1 inline, but it weighs an entire tonne, which, considering the weight of some WW2 fighters, is most-definitely prohibitively heavy. It's acceptable for something like a 262 where you can save weight elsewhere by using relatively light parts, but it ends up screwing with the CoM position and over-weighting smaller planes. Like I tried to make a Volksjager once and then realized it was nearly 3 tonnes empty vs. like 1.9 empty in reality.

Edited June 29, 2016 by Pds314

[SuicidalInsanity]

There's the Striker Aerospace mod, that has a bunch of WWII or WWII-ish give-or-take-10-years type cockpits. Not sure if it ever got updated to KSP 1.1.X or what the cockpits weigh, though.

Alternatively, since I have some free time coming up, I suppose I could throw together a BAD-T cockpits pack, shouldn't take too long to do.

[tetryds]

@blackheart612 mentioned he plans to make cockpits in the future for the mod AirplanePlus.

Since it's yet unknown when BDArmory will update you could give it a try and see what you can come up with, or we cand wait to see if blackheart will make them before it updates.

Also, given now I am working on my own game, BAD-T 3 will have a limited and not large amount of spots and because of that part of the airplanes will go straight to Tier 2 (if @DoctorDavinci decides to continue with them).

Depending on how many airplanes there are on Tier 2, losing on Tier 1 may disqualify you straight away, this way it's not a bad thing to go to Tier 2 as the chance of winning there will be higher.

It would be possible to have a Tier 1.5 too, with Tier 1 losers, that only depends on how many people are willing to help and how much time they have.

In resume, work hard on your airplanes for BAD-T 3 when it comes, it's going to be much less forgiving and much more fun.

[Pds314]

Grr... I'm trying to make my Ho-229 go faster without cheating. I _could_ sort-of-not-exactly-cheat by area-ruling it a little (Germany knew about that?? right?? right? They must've! German engineering and stuff??)... Basically, the drag coeffcient more than triples between Mach 0.5 and Mach 0.8.

Thrust is accurate, to the extent it can be on a prototype that's barely not a paper plane. Nominally 10 kN, which means 8.8 for normal flight.

Weight and fuel fraction are accurate, as well as fuel tank locations.

Turning circle is about 700 meters. No idea what the real thing could do in that regard.

Drag rudders might be a *tiny* bit bigger than IRL.

Edited June 30, 2016 by Pds314

[Pds314]

3 hours ago, tetryds said:

@blackheart612 mentioned he plans to make cockpits in the future for the mod AirplanePlus.

Since it's yet unknown when BDArmory will update you could give it a try and see what you can come up with, or we cand wait to see if blackheart will make them before it updates.

Also, given now I am working on my own game, BAD-T 3 will have a limited and not large amount of spots and because of that part of the airplanes will go straight to Tier 2 (if @DoctorDavinci decides to continue with them).

Depending on how many airplanes there are on Tier 2, losing on Tier 1 may disqualify you straight away, this way it's not a bad thing to go to Tier 2 as the chance of winning there will be higher.

It would be possible to have a Tier 1.5 too, with Tier 1 losers, that only depends on how many people are willing to help and how much time they have.

In resume, work hard on your airplanes for BAD-T 3 when it comes, it's going to be much less forgiving and much more fun.

Is BAD-T 3 going to be set in the same era, i.e. mid-late WW2 props?

[tetryds]

43 minutes ago, Pds314 said:

Is BAD-T 3 going to be set in the same era, i.e. mid-late WW2 props?

Yes.

[Van Disaster]

Dynamic deflection works quite well with the AI ( or at least worked well in older BDAs ), I'm surprised it *wasn't* in the list of mods for this tournament. The only problem from experience was if your craft gained too much altitude and you'd tuned it to prevent over-pulling and stalling, you'd have very little control deflection at any speed. Fixing that properly makes things complicated ( what exactly do you limit deflection with, and what UI do you use to set it up... ) - but the AI tends not to drift that far vertical anyway if it's actually trying to fight and not doing it's buggy reach-for-the-stars. I think I only have one video of a fight going over 5km despite that tournament having a floor of 1.4km, and that came back down to the deck eventually.

[blackheart612]

@tetryds I have a feeling that if a stable 1.1.x patch comes out soon, BDArmory will be updated at once along with some withheld mods.

[DoctorDavinci]

14 hours ago, tetryds said:

Also, given now I am working on my own game, BAD-T 3 will have a limited and not large amount of spots and because of that part of the airplanes will go straight to Tier 2 (if @DoctorDavinci decides to continue with them).

I will continue with hosting the BAD T3 tournament as long as we get working mods for it

I have tried a few in 1.1.3 and found vessel mover don't work right and there are various other glitches in other mods that need to be ironed out in order to make setting up the battles not a major hassle

[Bananders]

8 hours ago, DoctorDavinci said:

I have tried a few in 1.1.3 and found vessel mover don't work right

Try this fix by @SpannerMonkey(smce) 
-snip-

Edited July 1, 2016 by tetryds
No license.

[tetryds]

@Bananders Whoops, you need to include a license to redistribute that.

IIRC Vessel Mover is CC-BY-SA, so all you need to do is to add it to the download.

[Bananders]

 

9 hours ago, tetryds said:

Whoops, you need to include a license to redistribute that.

I'm just gonna link the original post by SpannerMonkey(smce)
Link

 

Edited July 1, 2016 by Bananders

[SpannerMonkey(smce)]

@tetryds Hi I've added the MIT license as was included with last release of VM (will i need to make the source available even though i made no code changes? all new to me so unsure )

[Adam Kerman]

Question: is the stock gull with prop for the engine, ok?