28 replies to this topic

[MyriadDigits]

I've noticed its especially bad with my panther, which currently has 16 Rear CT armor so it can survive an LRM volley to the face. However I've had my heavier mechs die to rear damage facing the LRMs that fell me.

[Xetelian]

Saw a twitch clip of proton using SRMs and hit the guy solid in the front and it did back armor damage. This is like a well known issue that will never be addressed.

[Gen Lee]

That's not the only issue. Here lately, there have been several times that I took a gauss slug to the chest, while having plenty of armor, and had my internals damaged without the armor being destroyed. There's been at least three times where I had full armor on a heavy, took a single gauss round to my CT, and it turned both my armor and internals yellow. Whatever is going on, it doesn't just affect armor.

[Maddermax]

Not sure if it's related, but the other day I stopped to kill a disconnected mech. Shot him straight in the back CT, but the damage kept showing up on the front torso. Had to finally shift fire to a different part of his back to start getting back shot so I didn't have to chew through his front armor. That was with lasers though, not missiles.

[Quxudica]

Xetelian, on 01 August 2018 - 05:06 PM, said:

Saw a twitch clip of proton using SRMs and hit the guy solid in the front and it did back armor damage. This is like a well known issue that will never be addressed.

Yep. Been around since Beta, probably earlier. Not sure if it's an unfixable issue with the engine or if PGI just never bothered trying but either way it's unlikely to ever change.

[LordNothing]

i just added hit detection to my game engine and i know why. a shell going 2000m/s is going to move 33 meters between frames (assuming 60 physics frames a second, and that may not be the case, if it was like 30/sec then 67m). unless you have a continuous collision detection system, that is a system that can detect collisions that occur between frames. that 33 meters is more than enough to completely pass through some mechs. even srms will move 6 meters between frames.

[Daggett]

Often such things also happen by overtwisting or by getting shot in the butt. I did that to a lot of people who then complained about hit-reg and cheating when in reality they just twisted a tad bit too much or did not realize that they can't torso-twist their butt out of harm...

Maddermax, on 01 August 2018 - 08:18 PM, said:

Not sure if it's related, but the other day I stopped to kill a disconnected mech. Shot him straight in the back CT, but the damage kept showing up on the front torso. Had to finally shift fire to a different part of his back to start getting back shot so I didn't have to chew through his front armor. That was with lasers though, not missiles.

Disconnected mechs often have their hitboxes shifted, they are not really comparable to how 'live' mechs behave.

Edited by Daggett, 02 August 2018 - 06:38 AM.

[oneproduct]

It's a pretty common problem with collision detection known at tunneling. Seen it a few times myself.

[Eisenhorne]

I lost an assassin to this last night... back was cored out, turned to face a streak huntsman, somehow his SSRM's hit me in the back and killed me when I had nearly full front armor.

Edited by Eisenhorne, 02 August 2018 - 07:06 AM.

[Ensaine]

This has been an issue since Beta.... we tell them, they know, they do nothing about it... wanna buy a mech pack?

[Dr Cara Carcass]

LordNothing, on 01 August 2018 - 08:45 PM, said:

i just added hit detection to my game engine and i know why. a shell going 2000m/s is going to move 33 meters between frames (assuming 60 physics frames a second, and that may not be the case, if it was like 30/sec then 67m). unless you have a continuous collision detection system, that is a system that can detect collisions that occur between frames. that 33 meters is more than enough to completely pass through some mechs. even srms will move 6 meters between frames.

I have been saying this for over a year now. But for pgi i think the server has 30 ticks per second. Would be nice if someone oculd confirm that teh gauss is traveling 130m between two ticks.

[jtyotJOTJIPAEFVJ]

LordNothing, on 01 August 2018 - 08:45 PM, said:

i just added hit detection to my game engine and i know why. a shell going 2000m/s is going to move 33 meters between frames (assuming 60 physics frames a second, and that may not be the case, if it was like 30/sec then 67m). unless you have a continuous collision detection system, that is a system that can detect collisions that occur between frames. that 33 meters is more than enough to completely pass through some mechs. even srms will move 6 meters between frames.

Look up ray tracing.

I cannot believe even PGI would be incompetent enough to use an engine that can't do continuous collision detection for projectiles. It's literally easier to do this properly than use any discrete solution.

More likely this is due to the hit box having gaps, allowing projectiles through the front armor and hitting the rear collider.

Edited by jtyotJOTJIPAEFVJ, 02 August 2018 - 05:37 PM.

[LordNothing]

jtyotJOTJIPAEFVJ, on 02 August 2018 - 05:34 PM, said:

Look up ray tracing.

I cannot believe even PGI would be incompetent enough to use an engine that can't do continuous collision detection for projectiles. It's literally easier to do this properly than use any discrete solution.

More likely this is due to the hit box having gaps, allowing projectiles through the front armor and hitting the rear collider.

ray collisions are the most basic collision detection mode. its pretty foolproof and gives you high accuracy contact points. its why all the hit scan weapons are so damn reliable. ive got orineted bounding boxes and convex hulls so far. i also have rays, spheres and aabbs defined but not really fleshed out yet. the latter 2 are needed for the broad phase to speed things up, but im not at the point yet where i got so many objects that i need to do a broad phase pass. i think im going to go with capsule colliders for determining interframe collisions but i need to do some reserch in that area. if you get a hit in the cylinder part but not the sphere ends then you know you need to rewind time to that point and retest those objects. only reason i havent done rays yet is they collide differently with different types of objects and so require algorithms specific to each.

when a game dev really nails collision detection without including a 3rd party physics engine they really earn my respect. unfortunately developers who can do that are few and far between. i will be lucky if i can just get something that is remotely usable.

Edited by LordNothing, 02 August 2018 - 06:14 PM.

[Akillius]

Ensaine, on 02 August 2018 - 08:44 AM, said:

This has been an issue since Beta.... we tell them, they know, they do nothing about it... wanna buy a mech pack?

Yup this.
Seen it before with pretty much every weapon and when it happens to me I generally don't play for a week.
Otherwise on daily.

[KoalaBrownie]

Dunno about missile damage but on one occasion I thought I was hitting a guy in the rear torso with an AC and it was the front for some reason. I thought he turned around after I hit him as well Didn't get it on video though so can't be 100% sure

[El Bandito]

Solaris full of such shenanigans. Sometimes shots hit the back, other times they do not register.

[jtyotJOTJIPAEFVJ]

LordNothing, on 02 August 2018 - 06:10 PM, said:

i think im going to go with capsule colliders for determining interframe collisions but i need to do some reserch in that area. if you get a hit in the cylinder part but not the sphere ends then you know you need to rewind time to that point and retest those objects.

Just test the collisions in the target's reference frame. That way you can use the tightest possible bounding boxes and there's no need to faff about with rewinding time.

LordNothing, on 02 August 2018 - 06:10 PM, said:

only reason i havent done rays yet is they collide differently with different types of objects and so require algorithms specific to each.

Congrats, you only need to do the easiest part then!

LordNothing, on 02 August 2018 - 06:10 PM, said:

when a game dev really nails collision detection without including a 3rd party physics engine they really earn my respect. unfortunately developers who can do that are few and far between. i will be lucky if i can just get something that is remotely usable.

Shouldn't that be the bare minimum standard of quality rather than something impressive? It's really not as hard as you make it out to be. I had working collision detection in my 0-budget game engine and I had no clue what I was doing.

[Tetatae Squawkins]

I just had a match where I teleported about 300-500 meters right into the middle of the enemy blob. That was new.

Edited by Tetatae Squawkins, 03 August 2018 - 02:31 AM.

[LordNothing]

jtyotJOTJIPAEFVJ, on 03 August 2018 - 02:17 AM, said:

Just test the collisions in the target's reference frame. That way you can use the tightest possible bounding boxes and there's no need to faff about with rewinding time.



Congrats, you only need to do the easiest part then!


Shouldn't that be the bare minimum standard of quality rather than something impressive? It's really not as hard as you make it out to be. I had working collision detection in my 0-budget game engine and I had no clue what I was doing.

i kind of need to rewind time because im doing an engine for a newtonian space sim. i need a lot of space. torchships and all, high velocities. so long as it kind of works il be happy. of course i dont have to worry about animations or anything like that. it also doesnt help that the whole thing is written in standalone lua. its also just a hobby i break out once in a blue moon.

Edited by LordNothing, 03 August 2018 - 04:21 AM.

[B0oN]

oh hey, lookie … one of CE´s oldest bugs again
I dunno how, but MWLL Dev´s fixed that issue with frontal shots hitting rear and vice versa .
So yeah, OP, that problem is actually quite known, but I guess PGI is too busy working away on their new game