WW2 BAD-T III - BDAc AI Dogfight Tournament

[Van Disaster]

1 hour ago, Alioth81 said:

Yes i will try to do that - I have to get to know the forum a bit better.

That was a complaint about the forum, not you - I tried to edit my previous post to quote yours, and couldn't find a way. Edit: well I certainly found a way to break quoting, at least.

First-pass kills are something I'm very unhappy about ( despite being to my benefit in that last battle ) - head-on jousting was something that happened in WW1, but for anything past 1917 it's just ridiculous. Unfortunately if the AI's target is in a line with it's velocity vector it doesn't know it needs to increase angular deflection... this is something that can be fixed.

 

1 hour ago, Alioth81 said:

I am sure that is true it was more a subjective feeling that explosive ammo does not always "work" whereas the AP ammo types seem a bit more reliable (this could be realistic).

For sure I put 30mm on my plane to have a good chance to destroy whatever is hit :-)
What would you think about some restriction about how many large caliber cannons you can mount/clip into the engine (which was for sure a real restriction)? A limit of 30 mm per engine in total would be a limitation for cannons but you still can mount 2x 12mm or 4x7mm? At first I had my 30mm shooting through the propeller but it just did not feel right.
This way if you want multiple large guns you would be limited to wing positions outside the propeller diameter where they are a bit less effective. (or have an inverted propeller design where you need an appropriate nose cone.

You'd only ever have one gun in the engine IRL and I'm not sure you could do it in a radial at all - for a watercooled V engine you can stick it in the vee, the propellor is geared so the crankshaft is below the gun. For a radial you'd have to have it firing down the middle of the crankshaft, and I don't think hollow cranks are a good idea for 2400hp aero engines... well I guess you could have it firing through a pair of cyls in a single row engine, but no *sizeable* radial has a single row. In-game it doesn't really matter if they're in the engine or on it, so not a problem as far as I'm concerned, benefit of the mass offset would be tiny.

However *above* the engine, or in the wing roots is just fine - you'd have to have synchronisation gear, but that was solved in WW1 and for slow firing cannon it's not a real impediment. You get problems with blinding the pilot with muzzle flash/filling the cockpit/engine air intakes with smoke/cartridge cases going everywhere they're not meant to/hot engine causing gun jams/other issues.

Ammo's effectiveness on hit depends on the temperature tolerance of the part it hits, amongst other things. My craft has survived a couple of 30mm hits, but I'd not rely on that. A burst of 20mm will kill anything, armoured or not, just like it ought to ( unless it's a Sturmovik, and good luck dogfighting in that ). The BDA issue is it's far too easy to hit anything - but if you make it harder to hit you'll have fights ending with no ammo, or lasting an hour. Well there is also that we carry a lot more ammo than real planes ( @tetryds, did you ever consider either increasing ammo mass directly or decreasing the rounds in a case? I agree the points system does work really well though ).

Edited November 10, 2017 by Van Disaster

[tetryds]

1 hour ago, Van Disaster said:

@tetryds, did you ever consider either increasing ammo mass directly or decreasing the rounds in a case? I agree the points system does work really well though ).

You can carry as much ammo as you wish, if you bring in more than needed you pay with the extra weight. And besides, the AI is not ready to deal with ammo management.

The ammo box capacity is basically a relationship between the mass and how many bullets such crate would be able to physically carry, with adjustments, especially for very large calibers.

I guess the fact that BAD-T is not meant to be realistic was lost again somewhere along the thread.

Edited November 10, 2017 by tetryds

[Van Disaster]

11 minutes ago, tetryds said:

You can carry as much ammo as you wish, if you bring in more than needed you pay with the extra weight. And besides, the AI is not ready to deal with ammo management.

The ammo box capacity is basically a relationship between the mass and how many bullets such crate would be able to physically carry, with adjustments, especially for very large calibers.

I guess the fact that BAD-T is not meant to be realistic was lost again somewhere along the thread.

Yes, that's why I asked if you'd considered changing the mass or the capacity with the same mass, did I say anything about restricting the number of ammo boxes?. The effect would either be less pray & spray as you reduce firing distances, or you can just load up on ammo & pay the mass cost if you want to shoot at 2km ( currently the mass cost is not enough to make a noticeable difference, so it's not actually a choice at all ).

If I was talking realism I'd be looking for ways to make the pilot less accurate, don't remember mentioning that in that post.

Edited November 10, 2017 by Van Disaster

[tetryds]

The crate mass is always the same, calculated for a perfect box of steel at the aproximate ingame size, then adjusted, it weights about 8kg.

Did you mean increasing the box weight/making it smaller so that you use less?

Their size was adjusted so that a box contains around 350 rounds of 20mm bullets, about what a real mid/late ww2 fighter would have. Then the ammo count for the other bullets was calculated as described on the previous post.

You can change the amount of ammo the box contains on the editor after placing it. I fail to visualize the effect you described coming from adjusting the boxes sizes.

As a rule of thumb I would suggest one box for manual piloting and two for AI, regardless of the weapond type you use given the other BAD-T constraints. That way you will never run out of ammo.

Edited November 10, 2017 by tetryds

[SuicidalInsanity]

20 hours ago, Van Disaster said:

That was a little more decisive than I expected I'll release the Deimos later when it's not a spoiler. BTW, can you guys give some music credits?

Those on the other hand went about as I expected. I think the old pair could probably do with some AI tweaks given that's changed too ( SkyRebal definitely needs min alt set! ) otherwise yeah, results as I thought.

All of my vids have music credited in their descriptions. As for AI, I dunno. I agree about the SkyRebel - it aggressively lithobraked once or twice back in the BAD-T II finals as well, but what would the Tytonid be without its proclivity for ramming?

...

With the final battles posted, BAD-T III draws to a close. There may be one or two more bonus videos - possible battle royale for third place/what sort of craft is needed to shoot down the Wasp/etc - but if not, it's been a fun ride, and hopefully I've done tetryds' tournament justice.

Congratulations, @Eidahlil for making an absolute monster of an aircraft.

[Van Disaster]

20 hours ago, tetryds said:

The crate mass is always the same, calculated for a perfect box of steel at the aproximate ingame size, then adjusted, it weights about 8kg.

Did you mean increasing the box weight/making it smaller so that you use less?

Their size was adjusted so that a box contains around 350 rounds of 20mm bullets, about what a real mid/late ww2 fighter would have. Then the ammo count for the other bullets was calculated as described on the previous post.

You can change the amount of ammo the box contains on the editor after placing it. I fail to visualize the effect you described coming from adjusting the boxes sizes.

As a rule of thumb I would suggest one box for manual piloting and two for AI, regardless of the weapond type you use given the other BAD-T constraints. That way you will never run out of ammo.

I was asking if you'd considered increasing the mass of a round of ammo ( or the physica box itself ) such that carrying more has a noticeable effect and there was a choice to be made between being able to spray shells around from long ranges, or conserving ammo until you have a more guaranteed hit by reducing the distance guns fire at. I was asking that because you're the one who's run numbers on that sort of thing. I could have put a single ammo box on my craft & been fine, I doubt it'd have even used that, but adding another made absolutely no difference to it's performance.

@SuicidalInsanityI think you've run a fine event!

Oh, and this one is for @MightyDarkStar - I was testing a new recording environment & thought I might as well post it, it involves a lightweight Kestrel. There's a 1440p option if anyone can use it.

Spoiler

 

 

Edited November 11, 2017 by Van Disaster

[tetryds]

@Van Disaster oh, I understand now.

Well, why would you want it to make more difference than it already does? Just like with fuel, the point is not to force you to make a hard decision on it, but to present sensible choices.

If you throw a single box of ammo and 5 Berezins you are going to run out of ammo pretty fast, and for a light fighter that has an empty mass of 2.5 tons the extra few hundred kg of ammo will matter a lot.

The ammo weight was taken directly from real life parameters and that worked pretty well.

Forcing people to carry more ammo may make some battles end by lack of ammo, it has happened on BAD-T 2. Currently all you have to do is "carry enough" but not "too much", its a soft balance restraint.

When balancing, there are roughly three types of balancing subjects: things that matter a lot, things that matter a bit and things that do not matter. The balance restraints can be set up for each of them in three ways: hard, soft and free. Things may start to make sense now, because it should sound obvious that things that matter a lot should have hard restraints, etc.

Well, it is not always like that, and the game itself already applies a great range of restraints. Things like the aero model already have a lot of balancing subjects sorted out. Now if you break the balance between the restraints you apply you affect the experience of the player. Imagine if you had to go to bed and sleep every day for 8 hours a day on GTA for instance.

On the case of ammo, it is something that matters a bit, afterall, you need ammo to battle, and the proportional weight increase already applies a soft restraint to it (remember it also affects momentums and the COM/COL relationship). In the end, no artificial restraint or adjustment is required on this case, especially since having more will not give you any sort of advantage.

 I may draw a diagram later to better explain how these balancing factors affect each other.

Edited November 11, 2017 by tetryds

[Van Disaster]

I'm aware of most of the reasons ( I'm not you, so obviously I don't know all your thought process ) - I brought it up because I don't think it makes enough difference that there's any point choosing to take less ammo and I wondered if you'd run numbers on different masses. TBH without gross changes in mass - and changes in other areas of balance like wings - I don't think ammo box counts will ever be more than fine tuning unless guns had to have their ammo in close proximity, and as the game doesn't enforce that that's a non-starter anyway - but maybe your data says otherwise.

I don't want to force people to take more ammo; the thought I had was to give people the option to take less ammo and gain a noticeable performance advantage that they could exploit to get closer where they've set their guns up so as to not waste ammo spraying it around at 2km. Right now if I pull half the ammo out of my entry it doesn't make a measurable difference to anything ( perhaps a fractional increase in climb rate ). Right now we're all ( from the videos, literally all ) showering anything that moves at almost as far as the guns can shoot because you don't gain anything by not doing that. But, as I said, on reflection I don't think fiddling with ammo box mass is really going to do that without changes elsewhere too.

Edited November 11, 2017 by Van Disaster

[tetryds]

Before we go further into balancing details, it would be great if all BAD-T III participants could answer this poll:

http://www.strawpoll.me/14376072

It will help more than you can imagine.

If you have any further suggestions just post them here as usual.

[MightyDarkStar]

4 hours ago, Van Disaster said:

Oh, and this one is for @MightyDarkStar - I was testing a new recording environment & thought I might as well post it, it involves a lightweight Kestrel. There's a 1440p option if anyone can use it.

  Hide contents

 

 

Oh, now this was cool. You've completely transformed the Kestrel, not to mention how good a fight that was. The Huginn is quite a pretty thing, I adore your building style. 

Did you tweak any of the Ai settings on the Kestrel M? I don't know what half the stuff means lmfao.

[Van Disaster]

2 hours ago, MightyDarkStar said:

Oh, now this was cool. You've completely transformed the Kestrel, not to mention how good a fight that was. The Huginn is quite a pretty thing, I adore your building style. 

Did you tweak any of the Ai settings on the Kestrel M? I don't know what half the stuff means lmfao.

The important parts are just fancy labels for the gains on a PID controller: Steer Factor is the proportional gain, Steer Damping is the differential gain, and Pitch Ki is the gain for an integral controller ( I haven't checked if it really only just works in pitch ). I set steer factor to 5.5, damping to 1.5 & pitch Ki to 2.5, but those wouldn't be final numbers by any means - it just means it maneuvers a bit more smoothly - which isn't always a good thing because it's like having slow reactions, but it does mean less energy burned off usually. I think the Rolf-75 suffered a little from over-enthusiastic control movement.  Most of the difference in the Kestrel performance is simply because it shed a tonne & a half or so.

Quite impressed that it took three Mk108 rounds without losing anything important!

I'll see if I can resurrect any of my civvie planes from 1.1 contests, I like those too.

And as promised here's the Deimos from this contest: https://kerbalx.com/VanDisaster/DAC-Deimos-Mk-X

Tetrydis, most of my frustration with building for this is just the same as other contests, the AI limitations.

Edited November 11, 2017 by Van Disaster

[SuicidalInsanity]

6 hours ago, Van Disaster said:

the thought I had was to give people the option to take less ammo and gain a noticeable performance advantage that they could exploit to get closer where they've set their guns up so as to not waste ammo spraying it around at 2km.

Was kind of the thought behind the Stribog's armament choice - I wanted it as light as it could be made, and 30mm shells were heavy, hence the less than full box.
On the other hand, setting engagement ranges to something more realistic like ~800-1200m with the current AI will only result in guaranteed kills during the first pass, as both craft have time to line up shots and begin firing from a range that's too close to dodge the incoming fire.
 

5 hours ago, tetryds said:

Before we go further into balancing details, it would be great if all BAD-T III participants could answer this poll:

I suspect that some part of this may be influenced by each contestant's relative (un)familiarity with FAR, given its recent return after a long hiatus; There were craft that looked as if they were a first foray into FAR, some that looked like the work of FAR veterans, and some in the middle  Yes, I know, decent tutorials on FAR are somewhat lacking.

If nothing else, perhaps the inclusion of a basic Tips 'n Tricks section on things like FAR Wing Strength/Mass, basic intro definition of terms like turnfighting, BnZ, and energy, etc is something that the host of the next BAD-T could include in the OP.

[Van Disaster]

4 minutes ago, SuicidalInsanity said:

On the other hand, setting engagement ranges to something more realistic like ~800-1200m with the current AI will only result in guaranteed kills during the first pass, as both craft have time to line up shots and begin firing from a range that's too close to dodge the incoming fire.

"Realistic" would be 100-300m and to be fair I think most of the kills in the tournament ( definitely not all mind ) are probably below 400m. The closer craft get to each other the more slight errors in heading vs target start causing bigger control movements & the craft can diverge from head-on, that needs some testing - it'll quite happily go head-on through a bullet hail for hundreds of metres anyway, so there's no difference whether it's firing or not. Ideally the AP would recognise head-on behaviour & break off, but that's another matter ( far too many collisions still anyway ).

[tetryds]

3 hours ago, SuicidalInsanity said:

If nothing else, perhaps the inclusion of a basic Tips 'n Tricks section on things like FAR Wing Strength/Mass, basic intro definition of terms like turnfighting, BnZ, and energy, etc is something that the host of the next BAD-T could include in the OP.

That, or I can get to make those tutorials that I promised some 3 years ago.

[OmegaForce]

I haven’t posted in a long time but have been fallowing this thread from the start and was glad to see the conclusion of BDA II. I wondered how the sky rebel would do against the Tytonid if we made it to the finals. Not surprised at the outcome as it was Ferram's own creation but getting to second place was a win in itself.

 

For this competition I didn’t have enough time to tweak and test my design as I would have liked nor was I up to date on how the AI has changed in BDA. However even with further improvements I don’t think Tornbird would have made it to the finals.

If a future tournament is held around a WWI theme I have a few suggestions.

Planes from that time period were very light so a maximum weight limit might be added to keep designs from using overly strong wings then guns could be limited to lower calibers and still be effective. Armor could be left out completely or greatly reduce the amount allowable to keep the vulnerability factor high. Also full moving control surfaces from B9 procedural wings I feel shouldn’t be allowed along with leading edge control surfaces.

Building cockpits out of B9 Pwings works but for those wanting a simpler solution the open cockpit mod would be ideal.

https://spacedock.info/mod/1126/MK-1 Stockalike open cockpit ( inline external seat )

Engines could come from the early set in the Airplane Plus mod. They might need some tweaking but visually they fit the era. The mod as well has a bi plane style fixed landing gear we could put to use.

 

[Van Disaster]

3 hours ago, OmegaForce said:

Engines could come from the early set in the Airplane Plus mod. They might need some tweaking but visually they fit the era. The mod as well has a bi plane style fixed landing gear we could put to use.

Kinda looks the part, ( if you're building a SPAD ) - performance curve is kinda wierd though, gets up to 125mph fast and then just stops. Good find though, will go looking around for more.

The 1.25m rotary is kinda funny - that'd have the plane on it's back on the ground, I reckon

I've been looking around for something like a Bristol Mercury/Jupiter, or original P&W Wasp - the classic 20s-30s install had the distinctive annular cowl.

Ed: I like this one, very inter-war.

Spoiler

 

 

Edited November 13, 2017 by Van Disaster

[SuicidalInsanity]

Not sure how light one can go with stock parts - that's what, 400kg of struts alone on the above biplane?
It would be ideal if game-changing edits weren't required for everyone who wanted to compete, but if low mass is the name of the game, while it probably wouldn't be as simple as @PART[*] {@mass /=2}, it might be worth considering some sort of part mass reduction MM patch.

[tetryds]

Yes, a mass reduction MM patch sounds good.

Procedural parts are usually very light, and IIRC empty procedural tanks weight less than nosecones or fuselage.

Not that many parts require weight adjustments though, Aviator Arsenal parts remain untouched. However landing gears are insanely heavy.

We certainly need an engine patch too, just like with the current BAD-T version.

I like that open cockpit mod, we could use that and offer the choice to make a custom open cockpit as well.

Since airplanes are much lighter, low caliber weapons end up dealing much more damage to them. On my tests I found out the control surfaces are very fragile and usually what's taken down first.

When messing around with Airplane Plus I remember hating those engine settings, just like @Van Disaster said, everything gets up to speed instantly then hit a wall at 200 km/h, there is absolutely no way you can design a faster (or slower) airplane using them, which is super bad. I think I even talked about it on the thread, but for people using the mod casually it doesn't matter.

[MightyDarkStar]

How are those centrifugal jets coming along btw? @SuicidalInsanity

[SuicidalInsanity]

@tetryds - I/m not surprised you had the results got got with the AP engines. Looking at their engine curves, the optimum flight envelope is ~15m/s at something like 20,000m. That's... not where they should be.
Try these:

Spoiler

    velCurve
    {
key = 0 1
key = 0.09987181 1.02265
key = 0.219359 0.8320512
key = 0.2994871 0.498718
key = 0.6 0 0 0
    }
    atmCurve
    {
key = 0 0
key = 0.2155983 0.001505733 0 0
key = 0.3220931 0.1720689
key = 0.5087607 0.6007528
key = 0.7771367 0.9021378
key = 0.8989316 0.9685513
key = 1 1
    }

@Mightydarkstar: I finished them awhile ago:

The next BAD-T looks like it'll be the interwar years, so these won't be needed, but if you want to fool around with them, you can grab them from here.

[MightyDarkStar]

11 hours ago, SuicidalInsanity said:

@tetryds - I/m not surprised you had the results got got with the AP engines. Looking at their engine curves, the optimum flight envelope is ~15m/s at something like 20,000m. That's... not where they should be.
Try these:

  Hide contents

    velCurve
    {
key = 0 1
key = 0.09987181 1.02265
key = 0.219359 0.8320512
key = 0.2994871 0.498718
key = 0.6 0 0 0
    }
    atmCurve
    {
key = 0 0
key = 0.2155983 0.001505733 0 0
key = 0.3220931 0.1720689
key = 0.5087607 0.6007528
key = 0.7771367 0.9021378
key = 0.8989316 0.9685513
key = 1 1
    }

@Mightydarkstar: I finished them awhile ago:

The next BAD-T looks like it'll be the interwar years, so these won't be needed, but if you want to fool around with them, you can grab them from here.

Gorgeous!

[Guest]

14 hours ago, SuicidalInsanity said:

@tetryds - I/m not surprised you had the results got got with the AP engines. Looking at their engine curves, the optimum flight envelope is ~15m/s at something like 20,000m. That's... not where they should be.
Try these:

  Reveal hidden contents

    velCurve
    {
key = 0 1
key = 0.09987181 1.02265
key = 0.219359 0.8320512
key = 0.2994871 0.498718
key = 0.6 0 0 0
    }
    atmCurve
    {
key = 0 0
key = 0.2155983 0.001505733 0 0
key = 0.3220931 0.1720689
key = 0.5087607 0.6007528
key = 0.7771367 0.9021378
key = 0.8989316 0.9685513
key = 1 1
    }

@Mightydarkstar: I finished them awhile ago:

The next BAD-T looks like it'll be the interwar years, so these won't be needed, but if you want to fool around with them, you can grab them from here.

Amazing!!!

[Alioth81]

On 14.11.2017 at 5:58 AM, SuicidalInsanity said:

I finished them awhile ago:

Nice work

Can you quickly explain how they are balanced? Some seem to have the same weight as the Jaeger but seem much more powerful

It motivated me to build a jet fighter according to BAD-T III rules with the original "Jaeger"

https://kerbalx.com/Alioth81/AL-7B

It can barely take of the Oasis runway where there is no Hangar at the end. (maybe the take off speed could be a bit lowered).

However the AI is severly hindered by the regain energy function that lets the plane as soon as it is above min height... and then it has to climb up again. Also the take off angle is too steep for my taste.
I made seom very little changes to BD armory AI and they in my opinion greatly increased the performance of crafts with a higher combat speed setting and especially low power jets that are below their combat speed when the go above min height.

1. Reduced the climb angle from 0.3*surface speed every 100 m to 0.15* surface speed every 100 m. (e.g. at 100m/s the change is from around 17° to 8°) Helps to climb at a more efficient angle and therefore increases the speed (which help fighters with higher min combat speed).

2. Changed the regain energy function so that there is a buffer of 150m above min height where it actually does not dive (which was useless anyway) and above that the dive is a bit more gradual until it gets to the same level as it is in the current version. Also here fighters with high "min combat speed" do not make a useless dive when the go above min height (they might still do a hard turn however).

		void RegainEnergy(FlightCtrlState s, Vector3 direction)
		{
            debugString.Append($"Regaining energy");
            debugString.Append(Environment.NewLine);

            steerMode = SteerModes.Aiming;
			Vector3 planarDirection = Vector3.ProjectOnPlane(direction, upDirection);
			float angle = (Mathf.Clamp(MissileGuidance.GetRadarAltitude(vessel) - minAltitude-150, 0, 1500) / 1500) * 55;
			angle = Mathf.Clamp(angle, -3, 55) * Mathf.Deg2Rad;
			Vector3 targetDirection = Vector3.RotateTowards(planarDirection, -upDirection, angle, 0);
			targetDirection = Vector3.RotateTowards(vessel.Velocity(), targetDirection, 10f * Mathf.Deg2Rad, 0).normalized;
 
			AdjustThrottle(maxSpeed, false);
			FlyToPosition(s, vesselTransform.position + (targetDirection*100));
		}

As I am not a programmer I am sure there are better ways to do it

Edited November 16, 2017 by Alioth81

[Van Disaster]

The best way is to not have a seperate behaviour, it should be something that's done while doing everything else ( just like avoiding the ground should be ). Haven't found the magic just yet though.

[Alioth81]

6 hours ago, Van Disaster said:

The best way is to not have a seperate behaviour, it should be something that's done while doing everything else

Actually this "fix" (at least the one in the regain energy funtion) is not seperate it improves the fast low TWR plane behaviour without affecting slow high TWR planes.
The original implementation had all planes dive below minimum altitude if they did not reach minimum combat speed after the steep climb. As soon as they dive below minimum altitude they would go into a steep climb (so they actually do not regain energy at all).

This cannot be intended behaviour and must be a "bug"

I simply  decreased the dive angle close to min height (in fact it has a very slow rise close to minheight to partially compensate for height loss when banking hard close to min height).

High TWR planes with low combat speed are not affected at all because the rarely experience such a situation.

But you are right it is just a drop in the ocean and not a magical solution.
To whom should this be adressed? Should I put this in the BD armory thread? I guess with KSP stock engines such problems probably never occur.