ETPro HitBoxes

[S3ti]

Ok got it fixed seems something with the etpro package I got via wget was wrong, copied over from my local installation and now it works. Again I didn't fight someone, just connected a second client and with +forward; +moveleft...

 

Here some results, overall pretty accurate:

 

 

 

 

 

 

 

 

And now fasten your seatbelts

Crouching and moving/strafing over bumpy terrain:

 

 

 

 

 

 

 

 

 

 

The flying headbox seems to be reproducable by crouching over little edges where a short stand up animation is played, others which aren't that extreme more random. I didn't play around with net* or any other settings but I guess that proves that hitboxes in certain situation are...f***ed up and you could shoot anywhere on the model and hit nothing. And this is no real opponent with fast strafing,dodging or jumping, pretty sure there will happen more errors then.

Edited June 6, 2013 by S3ti

[Death_Reincarnated]

Yep, this is what happens lol.

 

The one thing I was always interested in is how your aim is affected when you are moving over rough terrain - i.e. while strafing you move over small rocks, bumps etc and when using crouching.

 

The reason is that I find that you should be aiming in odd places to get those headshots and overall body hits.

 

You should have added a picture with a guy proning - I bet his head was up his a..... lol

[SunLight]

Bullet origin error experiment with that tool (I will attach the exe file here when I finish it):

View Axis (0,-1, 0) (Facing negative Y, running towards positive X axis + sprint)

headbox size is 12

 

200fps: AVERAGE bull. offs. (49 samples) (2.278 0.964 -0.126)

166fps: AVERAGE bull. offs. (57 samples) (2.600 0.500 -0.259)

125fps: AVERAGE bull. offs. (66 samples) (3.338 0.497 -0.125)

76 fps: AVERAGE bull. offs. (61 samples) (5.232 0.534 -0.125)

43 fps: AVERAGE bull. offs. (59 samples) (8.560 0.504 -0.125)

 

The higher the fps, the closer your bullet is to the viewpoint you had when you shot. With 43 fps you would entirely miss the head by aiming at it (half headbox is 6)

 

X offset is always positive, it means that the bullet origin is always 'ahead', i.e. in the direction you are strafing (I was strafing towards higher X coordinates)

But there is an offset even if you don't move, it seems to be related to your position.

 

Y offset isn't much important here, since Y is the axis I was looging at (in front of me)

 

---

 

125fps, running towards *negative* X this time, as you can see the real bullet is always 'ahead'

AVERAGE bull. offs. (63 samples) (-3.058 0.395 -0.125)

 

125fps, crouching and moving towards positive X:

AVERAGE bull. offs. (63 samples) (0.135 0.575 -0.125)

Crouching is slow. The faster the player moves, the more the bullet origin is wrong. If jaymod is the same, that would explain the mega unhittability in servers with higher movement speed...

 

And now fasten your seatbelts

lol dat headbox...

[SunLight]

Which the sequence of events that the costumer comes first?

Hitsounds and bullet tracers come from the server *after* you shoot (If your ping is high, even more)

 

So you shouldn't listen to hitsounds or watch tracers, you should only watch your crosshair, but keep into account my previous post on bullet origin offsets...

 

when you run left your real bullet starts more to the left, so if you aim at the enemy head, it would hit to its left, the opposite if you move right

[PiNoY]

Yes, this is also true in knife. when u released the knife while you are about to or during strafing/running left, knife tends to move leftwards or to the left.

 

Quite nice to know the above.

[SunLight]

What happens if someone connects and he doesn't have that modified cgame? Did you put it on a pk3? How did you call the mod? etpro?

 

I could join the channel #1on1.et if you want to test with more players... maybe you could ask on Crossfire as well

 

Pity that those frozen players are visible only to those clients who create them. So screenshots are the only way to go.

 

p.s.

those blue lines are drawn even with knife shots (ofc you won't see boxes then), if you wanna try some sort of 'clientside hit detection' you can take knife and see if the blue line hits the frozen player

[S3ti]

Bullet origin error experiment with that tool (I will attach the exe file here when I finish it):

View Axis (0,-1, 0) (Facing negative Y, running towards positive X axis + sprint)

headbox size is 12

 

200fps: AVERAGE bull. offs. (49 samples) (2.278 0.964 -0.126)

166fps: AVERAGE bull. offs. (57 samples) (2.600 0.500 -0.259)

125fps: AVERAGE bull. offs. (66 samples) (3.338 0.497 -0.125)

76 fps: AVERAGE bull. offs. (61 samples) (5.232 0.534 -0.125)

43 fps: AVERAGE bull. offs. (59 samples) (8.560 0.504 -0.125)

 

The higher the fps, the closer your bullet is to the viewpoint you had when you shot. With 43 fps you would entirely miss the head by aiming at it (half headbox is 6)

 

X offset is always positive, it means that the bullet origin is always 'ahead', i.e. in the direction you are strafing (I was strafing towards higher X coordinates)

But there is an offset even if you don't move, it seems to be related to your position.

 

Y offset isn't much important here, since Y is the axis I was looging at (in front of me)

 

Could be it's trying to be aligned to the grid somehow, or just another fps dependant flaw? Maybe the server is trying it with hitboxes too and there are some calculation errors which cause that stuff.

 

Hitsounds and bullet tracers come from the server *after* you shoot (If your ping is high, even more)

 

On Jaymod hitsounds are clientside, that's why it may "feel" less laggy even on higher pings because you hear them instantly. It also happens now and then that you hear 3-4 headshots, get killed and remaining enemy hp says still 100 (even without adre)

 

Yes, this is also true in knife. when u released the knife while you are about to or during strafing/running left, knife tends to move leftwards or to the left.

 

Quite nice to know the above.

 

That is the way knifethwrowing is implemented in Jaymod and is no bug!

 

What happens if someone connects and he doesn't have that modified cgame? Did you put it on a pk3? How did you call the mod? etpro?

 

I could join the channel #1on1.et if you want to test with more players... maybe you could ask on Crossfire as well

 

Pity that those frozen players are visible only to those clients who create them. So screenshots are the only way to go.

 

You can connect with a normal client, didn't notice any strangeness (though you won't have that copied models then). It's not in a pk3 just threw it in the etpro folder, it's all a standard installation except that dll. So people willing to test it should download your modified cgame_mp_x86.dll beforehand, to be safe to a separate ET installation with ETPro installed.

 

b_anticheat 0

sv_punkbuster 0

sv_pure 0

sv_allowDownload 0

 

so there should be no "accidents".

 

I wouldn't expect much from crossfire today except "2005 called" or "nice trojan"

 

Server is up again and I'll leave it running, if you want to see some things yourself add me in xfire and we can test some stuff there.

I'm curious how it will work with fast movements or higher ping connections.

 

 

Bullet origin error experiment with that tool (I will attach the exe file here when I finish it):

View Axis (0,-1, 0) (Facing negative Y, running towards positive X axis + sprint)

headbox size is 12

 

200fps: AVERAGE bull. offs. (49 samples) (2.278 0.964 -0.126)

166fps: AVERAGE bull. offs. (57 samples) (2.600 0.500 -0.259)

125fps: AVERAGE bull. offs. (66 samples) (3.338 0.497 -0.125)

76 fps: AVERAGE bull. offs. (61 samples) (5.232 0.534 -0.125)

43 fps: AVERAGE bull. offs. (59 samples) (8.560 0.504 -0.125)

 

The higher the fps, the closer your bullet is to the viewpoint you had when you shot. With 43 fps you would entirely miss the head by aiming at it (half headbox is 6)

 

Another thought, could that be simply added turnspread? Because the closer your fps gets to your mousesampling rate the lower your turnspread gets. If you have double the fps than your mouserate you get no turnspread at all, ET just doesn't recognize a mousemovement every second frame.

Edited June 6, 2013 by S3ti

[PiNoY]

when u say turn spread and bullet spread, are you saying they are different entity ? or i understand it wrong?

[S3ti]

when u say turn spread and bullet spread, are you saying they are different entity ? or i understand it wrong?

 

Yes, a little bullet spread is always there (bullet endpoint is random and also the origin as SunLight showed)!

 

But you get added spread by moving your mouse, you can make it visible with cg_drawSpreadscale 2. It will show you a growing bar on the left side accordingly. It's good to be aware of that and control your movement -> rather strafeaim with your keyboard when it's possible you will notice that moving your mouse at certain speeds for some seconds you get quite huge spread. That's the reason why 180° scripts give you an unfair advantage and were forbidden in leagues because ET recognizes no mouse movement and you get 0 turnspread with those scripts.

[SunLight]

Could be it's trying to be aligned to the grid somehow, or just another fps dependant flaw? Maybe the server is trying it with hitboxes too and there are some calculation errors which cause that stuff.

I think this, but I'm still not sure: maybe your bullet is actually shot in your next (client) frame, so of course you will have moved.

 

That explains the fact that the more fps you have, the less the error is, and the fact that if you crouch or move slower there is less error.

 

The error when you don't move could be some rounding or calculation thing.

 

On Jaymod hitsounds are clientside, that's why it may "feel" less laggy even on higher pings because you hear them instantly. It also happens now and then that you hear 3-4 headshots, get killed and remaining enemy hp says still 100 (even without adre)

I know that the remaining hp thing feels bugged in jaymod, but the hitsounds, I'm not sure.

imho they are sent from the server with the playerstate in the snapshot, at least in my demostats tool I take jaymod hs information from there...

 

maybe your ping is never high enough to notice

[SunLight]

when u say turn spread and bullet spread, are you saying they are different entity ? or i understand it wrong?

there is only one spread, but it depends both on bullets and mouse moves

 

if I remember correctly the formula is something like this:

 

1. every time you shoot a bullet, it goes up a certain amount

2. every frame it goes up, based on your mouse move, but there is a maximum threshold (see below)

3. every frame it goes down a fixed amount (higher for crouch, prone)

 

The interesting thing about #2 is, there is a maximum amount it can go up each frame. So if you cover a big angle quickly in few frames with highsens or turn scripts, you will hit this threshold and you will get few spread

 

If you move slowly over several frames, your turnspread will never be cut and you get the whole amount. In the end, because of this, medium speed is what gives you more turnspread.

 

---

 

If you shoot when you crouch (not in jay3, since bullets are 2 fast) the 'down' component is enough to keep the spread at the minimum. If you stand, it goes up faster than it goes down so it will increase.

 

for some weapons (engi rifle) spread is always max, and if you jump (unless you crouchjump) it's high as well

[S3ti]

I know that the remaining hp thing feels bugged in jaymod, but the hitsounds, I'm not sure.

imho they are sent from the server with the playerstate in the snapshot, at least in my demostats tool I take jaymod hs information from there...

 

maybe your ping is never high enough to notice

 

Jaybird wrote that he moved the hitsounds completely to clientside, but dunno if it makes that independant from the real hit detection on the server...can't read that code If so, it may underlie client prediction errors too?! Anyway, if I play Jaymod on US servers with 150 ping (in scoreboard) it feels way less laggy than etpro or silent with same ping and I hear the hitsounds instantly. But Jaymod has another antilag system too...

 

http://forums.warchest.com/showthread.php/11574-problem-making-hitsounds

[SunLight]

Jaybird says (btw, it took me some work to find out back then watching inside jaymod, lol if I only knew he wrote it clearly somewhere )

 

Well, yeah, serverside detection. Basically I used Quake 3's version of hitsound: ps->persistant[PERS_HITS]

 

So it's still serverside.

ps is the playerstate sent from the server, not the predicted playerstate I think. imho he meant there are no more 'events', when he said clientside, so there must still be a delay.

 

If it was really clientside, you could hear a hs hitsound and maybe you didn't headshot at all... (which sometimes on jaymod I had the impression of)

 

edit:

p.s. is your xfire like your nick? I don't wanna add someone else

[S3ti]

Jaybird says (btw, it took me some work to find out back then watching inside jaymod, lol if I only knew he wrote it clearly somewhere )

 

 

So it's still serverside.

ps is the playerstate sent from the server, not the predicted playerstate I think. imho he meant there are no more 'events', when he said clientside, so there must still be a delay.

 

If it was really clientside, you could hear a hs hitsound and maybe you didn't headshot at all... (which sometimes on jaymod I had the impression of)

 

edit:

p.s. is your xfire like your nick? I don't wanna add someone else

 

Exactly, or the opposite happens too, a quite unbelievable amount of hs and the guy still standing (maybe because of adre but still strange sometimes) Or you see hitpuffs on the model and no hitsound is played.

 

Yep, same nick.

Edited June 6, 2013 by S3ti

[SunLight]

most of the boxes we tested were correct, we even tested 15maxpackets and other warp settings

 

boxes are generally ok, but I could post some lol ones:

 

here it looks like he is carrying a headbox maybe bumpy terrain again...

 

when changing strafe direction, some bugs can happen as well...

 

crouching bug:

 

but normally they were quite good: