Rammed someone and he survived

[=X51=VC_]

So I saw a Pe-2 and dived on him. He was smoking and not firing back so I pressed the attack, pulled up at the last second and flew over him (I thought). Next thing I know my screen is red, prop bent, wings missing, then dead and kicked to external view. I can see him flying away just fine... eventually some of the guys I was with on TS took him down but it took a while. He was rubber-banding quite badly when I dived on him so I guess... lag? But it still doesn't make sense if a collision was recorded on my end that it didn't register for him.

[OrLoK]

TBH I wouldn't fret about this, especially if there was lagging involved.

 

Although one would expect that the collision would be registered on the other plane, we have to remember that the idea of reality,mass etc is not "real" in the virtual world and that odd results can happen when things are not running as they should.

 

Now, if this becomes a common occurrence, then we need to take notice

[BubiHUN]

Lagging, and netcode can cause this. Just try to do that in quick missions, and you will see how the devs wanted it to happen. I mean, everything. 

[=EXPEND=Tripwire]

It does make sense and happens occasionally due to lag.

 

Which option in this situation would you prefer.

 

Your flying in combat. You near another plane but avoid a collision. The other pilots game registers a collision.

 

A) Your plane explodes for what you assume is no reason as you clearly avoided a collision.

B) The other guy gets to ghost through your plane seeing internal structure etc.

C) Only he explodes as per what he saw onscreen. You don't as you saw no collision.

 

It doesn't occur every time. I'd prefer C, exactly as it is in-game now. It's nothing to do with how strong the other plane was, or how bad it's damage model is as per other threads on this same topic.

 

Yes it's a bitter pill to swallow seeing the other plane fly away but that's online life.

[216th_Jordan]

This doesn't have to mean a lot. I exploded so often when colliding with other planes while they just flew away, likely because my ping was lower? (It was mostly with people who had latency issues)

[=X51=VC_]

Yeah, I'm not complaining as much as sharing an amusing anecdote, I did pass him close enough to give the pilot a haircut with my prop. But it does make sense now with the lag and the way it's implemented. Although, the situation is reversed to how you described. Visually, I passed him, but my plane was the one that registered a collision I didn't see. By your logic, if he was lagging he should have gone boom and I would be fine

[Ehret]

Lag is not constant and hit boxes aren't (I think?) fully precise. Then we have our local graphic settings, for example number of pre-rendered frames which affects perceived latency even further.

[A_radek]

I have a theory and someone that knows otherwise or more please correct me.

Aircraft in Box are divided in to sections, LeftCenterWingPiece, LeftMiddleWingPiece, LeftWingTip and so on. Each and every of those 'parts' has a set amount of healthpoints. If your engine 'part' of that many HP collides with a pe-2 left side tail section that has a bigger amount of them, pe-2 tail will get damaged but not necessarily destroyed.

[Finkeren]

I had this happen twice online last week and someone asked, if I was a cheater (not accused, asked, and politely so)

[=X51=VC_]

I had this happen twice online last week and someone asked, if I was a cheater (not accused, asked, and politely so)

Were you flying a Pe-2 last night?

[SJ_Butcher]

simple, stalinium is much more stronger than any german steel

Edited January 17, 2018 by SJ_Butcher

[SAS_Storebror]

A few days ago online on our FAC test server, Skylla, Blueberry, Conky and I flew a couple of sorties.

On my 1st sortie (with quite some success), on landing I overlooked the NDB placed at one end of the runway and ran right into it on approach (all images clickable for full size):

 

The resulting damage was interesting.

The prop got bent but the wing went through the NDB (or vice versa) with no damage:

 

The poor Focke got kicked to the side, and even though the tail didn't touch the NDB at all, the right horizontal stab was ripped off:

 

The aftermath fell within expectations first.

Ditched the right wing, structure showed a crack, then toppled over to the left, ditched that wing too, struture cracked again:

 

Finally I got rid of my tail wheel and came to a halt, twisted around by 180°:

 

The plane was standing like this for about a minute, when all of a sudden the right main gear strut collapsed and I fell down on my right flaps:

 

Conclusion: Funny stunt, albeit not all that realistic 

 

Cheers!

Mike

[Max_Damage]

The game calculates things frame by frame. if you are so fast that you fly past an objects between the frames then you dont take damage.

[SAS_Storebror]

While the explanation sounds reasonable at first glance, on second glance we see:

I was on my final approach, just about to flare for touchdown. Speed was 160-170 km/h.

170km/h equals 47.2 m/s.

With 50 SPS (Simulations per second) on the server, the plane moves less than 1m per simulation.

The average wing chord of a 190 (correct me if I'm wrong, no figures at hands at the moment) is about 1.7m.

Add to that the pole strut diameter of the NDB (about 30cm? maybe).

 

This makes it impossible to pass the NDB between two "frames" (simulations on server side).

In order to pass it between three simulations, first of all the server would have to drop one. This would have been visible on the server window, our server is far from it's resource limits and simply doesn't drop simulations so far.

And even then, at the end of the previous simulation I would have had to be exactly a millimeter away from the NDB pole, and after the next simulation but one, just a millimeter behind. Unlikely.

 

Add to that, all of this doesn't explain what ripped off the stab.

But at least it's funny 

 

Cheers!

Mike

[=X51=VC_]

He he, I like quirky incidents like that, thanks for sharing!

[A_radek]

Storebror.

 

Max_damage is on the right track. Hit detection is demanding as the frequency of checks has to be so high. Especially for fast traveling things like bullets. It's not linked to framerate as you concluded. If I were to choose I'd spend more resources on bullet hit detection than aircraft collisions.

 

In an online environment it gets more complicated but my knowledge ends there.

 

If you watch someone online make a landing at a high descent angle you will see them dipping through the ground before wheels are above ground again. A simillar situation, and the negative side of optimizations. Interpolation that smooths things out at an unfortunate moment.

Edited January 17, 2018 by a_radek

[Pupo]

...

 

Mike

 

 

You found a very likely candidate method to reproduce a somewhat serious bug in the game. Someone should try and repeat your scenario a couple of times, and see if the results are reproducible. If there is indeed a bug, this is the best chance we have to have it fixed. 

[SAS_Storebror]

It's not linked to framerate as you concluded.

 

I'm not talking about framerate. Max_Damage called it "frames" - what I'm talking about is the "Simulations per Second", that's what the dserver.exe shows as "SPS" value.

On a well working server, this value is 50 +/- 0.1 SPS, with no variation, and that's exaclty what we have on our FAC server.

 

 

If you watch someone online make a landing at a high descent angle you will see them dipping through the ground before wheels are above ground again. A simillar situation, and the negative side of optimizations. Interpolation that smooths things out at an unfortunate moment.

 

Correct. On client side flight paths are being interpolated until the next location update comes in.

Ping times from both clients involved plus the SPS tick delay sum up to a value where you can observe things like you've described.

That however has nothing to do with what we observed here.

I'm the only client involved.

The NDB is a server thing.

My ping to the server is 20ms (even when we add that to the distance travelled, we still don't get the required two meters), but that doesn't matter because each of my client's location updates has that same delay, so effectively on server side they are received and calculated in the same 50 SPS rate again.

 

 

You found a very likely candidate method to reproduce a somewhat serious bug in the game. Someone should try and repeat your scenario a couple of times, and see if the results are reproducible. If there is indeed a bug, this is the best chance we have to have it fixed. 

 

I do have the recorded track, so if the Devs are interested, just drop me a PM.

 

Cheers!

Mike

Edited January 17, 2018 by SAS_Storebror

[AndyJWest]

Storebror's screenshots clearly show that the top of the NDB broke off in the collision. The break isn't exactly in the right place, but expecting the damage model for a beacon to depict exactly every collision is unrealistic. Allowing for that, I don't see much to complain about really. The beacon broke. The plane broke. If you want a simulation that models the complete physics in real time for every possible collision between in-game objects, I suggest you start saving for a Cray to run it on.

 

As for aircraft collisions online, people have been complaining about such issues for many years. Fundamentally, it is a consequence of where and when collisions are modelled. To get totally consistent results you'd have to model everything on the server, and put up with the inherent lag involved. Which would make flying rather uncomfortable.

[A_radek]

I'm not talking about framerate. Max_Damage called it "frames" - what I'm talking about is the "Simulations per Second", that's what the dserver.exe shows as "SPS" value.

On a well working server, this value is 50 +/- 0.1 SPS, with no variation, and that's exaclty what we have on our FAC server.

 

 

 

Correct. On client side flight paths are being interpolated until the next location update comes in.

Ping times from both clients involved plus the SPS tick delay sum up to a value where you can observe things like you've described.

That however has nothing to do with what we observed here.

I'm the only client involved.

The NDB is a server thing.

My ping to the server is 20ms (even when we add that to the distance travelled, we still don't get the required two meters), but that doesn't matter because each of my client's location updates has that same delay, so effectively on server side they are received and calculated in the same 50 SPS rate again.

 

 

 

I do have the recorded track, so if the Devs are interested, just drop me a PM.

 

Cheers!

Mike

Storebror my online example was picked to give me an easier time explaining interpolation. As I read you know exactly how that works, and unless I'm misunderstanding you fundamentally: don't you think same type of optimization goes here? There are many objects in a Box map to check collision against So the resolution of your flight path/collisions will be limited even to the degree of "chunks", as in your wing passing through and your stabilizer also receiving damage. While for how long your wing was able to travel on the other side of the nbd without a reaction would equal the exact time it took for the collision to be detected.

 

 

I really thought max was talking about frame rate. Sorry max!

Edited January 17, 2018 by a_radek

[Max_Damage]

I dont think that there is any similar problem with bullets. I ve not once seen hit detection not work for me. As for collisions im fine with it. its a computer program after all what did you want. It cant divide by zero either.

Edited January 18, 2018 by Max_Damage