WW2 BAD-T - World War 2 BDArmory AI Dogfight Tournament [FINISHED]

[TUKE]

The new BDA AI is actually quite good and avoids head-on attacks.

[Phearlock]

I thought your plane was really cool Azimech, but yeah it didn't seem like it was designed to force the AI to fight the way you wanted with it. If I were making an aircraft for a human I'd probably not end up designing the KF-Spike 5 the way I did. Tetryds managed to set his aircraft up for B&Z'ing tactics which was cool. I decided to exploit the v8 BDAI's tendency to be terrible at going defensive by simply giving the Spike a bunch of instantaneous rate. With a dash of >1TWR when slow and decent low speed maneuverability to make up for the fact that I basically intend for the AI to bleed all its energy on purpose to get good rate and small radius.

I'm still speculating on what sort of design will do best with the new AI. I won't have to design a craft designed to survive head-on passes as much at least, so that's nice.

IL-2 was fun, though never played it too much. 500+ hours in both BMS and DCS though.

Edited September 14, 2015 by Phearlock

[tetryds]

Yeah, the new AI seems to be much better, I did not get to test it yet but I have only heard good stuff about it.

We will probably have to relearn how to use it, and I hope that it allows us to modify as much as possible on the behavior (did you know there is even a setting for how far the airplane will go before turning back for attacking on BDA 0.8.x? But it's global)

You may have wondered why didn't I post more battles already.

Well, keep wondering.

Btw, Azimech's plane goes straight to the HOF, I have never seen a plane sustain that much damage and still fly.

It may not be as resistant as the VZ-38 but it can be teared to pieces and go like like nothing happened.

Also, the winner's craft will always be public.

This way I expect everybody to work on their crafts to be at least as good as the winner of the previous tournament, the quality of our planes is going to grow exponentially.

But I don't want this to be an elite challenge, if it becomes too hard to compete I will have to create a "training" version, for people to learn, wouldn't be as serious as this one, mostly public crafts where people run battles by themselves to improve their planes.

Anyway, I don't think that will be needed, because...

Edited September 14, 2015 by tetryds

[TUKE]

Well the L-19 was designed to counter what was becoming the standard design of two aerosports, so it was supposed to climb like crazy then BnZ with it's fairly high damage but high RoF guns.

[Azimech]

I had no idea you could tweak the AI to do things like BnZ. I wish to learn about it though, but even better would be one generic setting. Put Erich Hartmann in an LAGG-3 and he still would've scored kills, and I want to compare airplanes, not pilots. The airplane should influence what kind of tactic the AI chooses.

And a future version of my plane will sure be quicker to scramble.

- - - Updated - - -

I have IL-2 Sturmovik Ultimate Edition as well as 1946 (But the disk got scratched), Forgotten Battles (Lost a disc) and the original.

Any ideas how to get it run with a 1920x1080 screen?

Install the latest patch and you're good :-)

[TUKE]

Install the latest patch and you're good :-)

Where do i get the patch from?

[tetryds]

I had no idea you could tweak the AI to do things like BnZ. I wish to learn about it though, but even better would be one generic setting. Put Erich Hartmann in an LAGG-3 and he still would've scored kills, and I want to compare airplanes, not pilots. The airplane should influence what kind of tactic the AI chooses.

And a future version of my plane will sure be quicker to scramble.

Yep, exactly, use the same settings as mine on another plane and it's not going to work at all.

You tweak the AI to handle a certain design, not the other way around.

I still want to check out the new AI and suggest some stuff.

[TUKE]

Yep, exactly, use the same settings as mine on another plane and it's not going to work at all.

You tweak the AI to handle a certain design, not the other way around.

I still want to check out the new AI and suggest some stuff.

The new AI can aim a lot better and will try and evade when shot at, instead of just flying blindly. However it does now have a tendency to fly into he ground a lot more

[Littlerift]

The new AI can aim a lot better and will try and evade when shot at, instead of just flying blindly. However it does now have a tendency to fly into he ground a lot more

Better at evading?! My current aircraft is already almost immune to being shot in close range combat due to the way it tends to spin and turn. I need to try this new BdArmoury.

[ferram4]

In my experience the current BDArmory AI evasion is actually a death sentence for the evading fighter. Planes evade, but blindly and with no concern for the energy they're losing or where the opponent they're evading is going. I've seen planes with much more energy extending away from a slower plane suddenly go insane and bleed off all their speed because that slower plane got a single shot off and triggered evasion... and then the was-slower-but-not-anymore plane caught up and destroyed the target plane. I've seen other planes panic themselves into the ground after a near head-on at low altitude, and I've seen planes that should have simply continued the turn in the direction they were going "evade" back into their pursuers guns.

Whoever gets guns near enough to fire first nearly automatically wins the engagement, because the other one will be too busy panicking rather than actually trying to delay dying, let alone attempting to gain a reversal. The only time this doesn't happen is when the chasing fighter bleeds off so much energy that it falls below the minimum altitude set in the AI and it stops pressing the attack to climb out of that and the reprieve lets the other fighter switch things up.

Older BD AI with current better-than-real-life prop engines seems like it could be tweaked to do BnZ, but current AI with real-life prop engines are stuck turnfighting, lawnmowering and making things easier for their opponent.

Edited September 16, 2015 by ferram4
whoops, more clarity

[tetryds]

I have just released the weapons pack that will be used on the next version of this challenge!

Have fun playing with it, the points system to determine the amount of each weapon you can use is still to be set.

I heavily recommend testing each weapon to see which one fits your aircraft design style the best.

http://forum.kerbalspaceprogram.com/threads/134745

[Littlerift]

I have just released the weapons pack that will be used on the next version of this challenge!

Have fun playing with it, the points system to determine the amount of each weapon you can use is still to be set.

I heavily recommend testing each weapon to see which one fits your aircraft design style the best.

http://forum.kerbalspaceprogram.com/threads/134745

This is awesome - I need to lock myself away for a little bit and test.

[Darren9]

Great weapons pack. I wonder if anyone will use that huge canon, I'm thinking no 2 ShVAK with 2 crates of ammo seems quite effective.

[Littlerift]

Wow, the new BdAI constantly rams other aircraft, and seems to quite often just fly in straight lines in front enemy aircraft, and it tends to fly into the ground. This is upsetting, I actually prefer the old system.

Edited September 21, 2015 by Littlerift

[TUKE]

Wow, the new BdAI constantly rams other aircraft, and seems to quite often just fly in straight lines in front enemy aircraft, and it tends to fly into the ground. This is upsetting, I actually prefer the old system.

Really? I found it better, apart from the flying into the ground (But hey, i've got min alt set to, well, min).

[Wjolcz]

Really? I found it better, apart from the flying into the ground (But hey, i've got min alt set to, well, min).

Same. The new AI actually tries to use some sort of tactics by applying airbrakes and such.

Just finished my plane thanks to the released guns and I'm ready for when the tournament restarts.

Edited September 21, 2015 by Veeltch

[tetryds]

The first Semi-final!

Was a good match.

[Littlerift]

Really? I found it better, apart from the flying into the ground (But hey, i've got min alt set to, well, min).

It seems much better at deflection shots that the previous AI - my new design is forever taking the wing off of my old design in high-angle, close range attacks, but it seems very twitchy even with damping turned way up. And it tends to fly in straight lines to build up some separation, which allows the plane behind it to kill it with ease. My new design is superior in every way to my previous design, but it can only win if the AI doesn't do something ridiculously dumb.

[ferram4]

I think that the min alt setting should be required to be at least 600m rather than allowing the 150m minimum. With the 150m minimum, pretty much all planes need to be turnfighters so that they can avoid slamming into the ground during a dive or evasion since they're allowed to get so low. Problem is that if someone decides to set the minimum to 600m independently of everyone else to avoid that, everyone who does go for the 150m minimum can chew up the 600m climber's rear ends with no competition.

But yes, the AI is scarily good at deflection shots now.

[Littlerift]

I think that the min alt setting should be required to be at least 600m rather than allowing the 150m minimum. With the 150m minimum, pretty much all planes need to be turnfighters so that they can avoid slamming into the ground during a dive or evasion since they're allowed to get so low. Problem is that if someone decides to set the minimum to 600m independently of everyone else to avoid that, everyone who does go for the 150m minimum can chew up the 600m climber's rear ends with no competition.

But yes, the AI is scarily good at deflection shots now.

I like this idea. At the moment the challenge heavily favours aircraft that are built to turn fast and have the minimum altitude set to 150, because if you have a fighter that is designed to BnZ or fight at higher altitudes then you can be sure that your plane will be shot down while climbing by any well designed competitor following the former design philosophy. It would also mean that things like climb rate actually matter, whereas at the moment you can get away with things like high wing-loading because the trade-off of wing area for weight means you're penalised in an instant-action, ultra low altitude dogfight. It would also help the turboprop a little bit, as while my current design uses a turboprop some testing has shown that switching out for the radial engine, which is lighter, has faster acceleration and is more powerful, almost ensures a victory in under a minute against the standard model. Hell, my radial engine variant can engage and win against 3 of the turboprop variants simultaneously, simply due to the fact that it will be in the air and turned around before the other planes have even built up any speed or altitude.

On another note, however, I've loving the weapons pack - the models are fantastic: stock-alike but still with a very specific character and very high quality, and they mostly seem balanced against one another (although the cannons could perhaps use a weigh increase: one cannon weighs less than two mgs and ensures a much faster kill).

[tetryds]

While I agree that 600m min should be the limit low, an aircraft built to get to 150m faster will also reach 600m faster, even more if the other airplane does not accelerate as much while climbing.

What I am thinking about doing is spawning both very far away from each other, but I am still thinking of how would I do that.

By the way, the next semifinal match was just ran!

But first I need to see with erasmusguy if he would be fine with it being made public, it was not one of the most fun matches, you know.

But there we have our finalists!

Maelstorm vs KF-Spike 5

Both airplanes were designed to be light and powerful, with 20mm cannons and a lot of turn rate.

The battle will happen soon.

Who are you betting on?

Remembering that the winner's craft will be made public after the final.

About the losers branch, I am thinking about quick best of three 1x1 battles, no upload, just to show who wins.

[Phearlock]

I'm 95% certain the VZ-38 will take the losers bracket pretty easily, it's a really good plane and I'm sad it got knocked down in the first round.

And whoo, finals, I hope my little plane will manage. Will the focus on low speed rate, high TWR and high drag during turns pay off as well as I hoped when I made it?

[Wjolcz]

KF-Spike 5 is a real beast and I think it will win. Mainly because of it's all-moving elevators. Seems it gives a lot of advantage.

[Phearlock]

All moving surfaces usually start showing clear advantages over elevators around the transsonic region and beyond. At these speeds, you can get the same performance out of simply sticking an oversized elevator on the back (pretty easy with procedural wings). Though the all-moving stabilator was much easier to use. I believe they're set to a maximum deflection of about 13 degrees, as the plane is unstable enough that it doesn't really need as much control input for pitch as you'd expect. And I did ask if the two all-moving surfaces were ok before submitting the craft.

It's turning capability is mostly inspired by the 109, with the slats giving it a very tight initial turn at the cost of a lot of drag. Though unlike the 109 it has enough TWR and lift that the low speed sustained rate is actually very good, hence it doesn't suffer as much from shedding all its airspeed as a lot of other designs would.

[Wjolcz]

All moving surfaces usually start showing clear advantages over elevators around the transsonic region and beyond. At these speeds, you can get the same performance out of simply sticking an oversized elevator on the back (pretty easy with procedural wings). Though the all-moving stabilator was much easier to use. I believe they're set to a maximum deflection of about 13 degrees, as the plane is unstable enough that it doesn't really need as much control input for pitch as you'd expect. And I did ask if the two all-moving surfaces were ok before submitting the craft.

It's turning capability is mostly inspired by the 109, with the slats giving it a very tight initial turn at the cost of a lot of drag. Though unlike the 109 it has enough TWR and lift that the low speed sustained rate is actually very good, hence it doesn't suffer as much from shedding all its airspeed as a lot of other designs would.

Oooooh! So it's slats that makes it so manouverable? Neat. Had to actually rewatch the video to notice them.