27 replies to this topic

[feeWAIVER]

In a nutshell,

[Widowmaker1981]

I see your "people cant aim" and raise you "WTF is this then?"

(Assuming its bad hitreg on the Shadow Cat, since i have no idea what else it could be)

[CFC Conky]

Server latency?

[Tarl Cabot]

feeWAIVER, on 28 September 2023 - 04:35 PM, said:

In a nutshell,

Exactly what is the issue? No arm lock, so one laser in the torso/head hit at the multi-crosshairs) vs 2 in the arms following the small circular crosshair). And is that blob a PPC being fired from the right torso? So it is heading for the multi-crosshair like the head laser. Perfect timing if that is a PPC.

Will say that I wished the separated crosshair graphics were reversed.

Edited by Tarl Cabot, 29 September 2023 - 01:29 PM.

[Meep Meep]

CFC Conky, on 29 September 2023 - 08:14 AM, said:

Server latency?

And coding because not too many months back after a patch I noticed both the physics glitches and hit reg became worse. Getting stuck on walls and objects far more often and seeing entire alphas swallowed whole without so much as a color change on the paperdoll even if the reticle turns red and the paperdoll flashes.

[Ihlrath]

Meep Meep, on 29 September 2023 - 01:43 PM, said:

And coding because not too many months back after a patch I noticed both the physics glitches and hit reg became worse. Getting stuck on walls and objects far more often and seeing entire alphas swallowed whole without so much as a color change on the paperdoll even if the reticle turns red and the paperdoll flashes.

Man getting stuck on walls or random rocks has gotten terrible. I also can't count how many times I've put my nose into an assault's back, alpha'd and watching nothing hit somehow....

[LordNothing]

most hitreg complaints or bad aim. seeing a mech walk into a weapon trail and assuming that counts as a hit. the projectile was fired with too much lead, it passed the target, and the target's momentum walked it into the trail in the next frame. while it appears in that frame that the projectile passed through the target, one frame does not tell the whole story. physics is a four dimensional problem. this is why ppcs are reported phasing through more than things like acs and gauss, because those weapons dont have long, highly visible trails. its all the same code.

the worst offender ive seen is an awesome which i shot repeatedly though the ct with an erppc. it was a non moving target and i was hill humping. i figure if the weapon velocity is high enough the awesome's short axis can fit between the hitboxes of a projectile in to successive frames at 30 ticks/s. this may also be why lights seem to be able to shed damage like the scat in the above video. you could up the server tick, but i dont think events like this happen frequently enough to warrant a drastic change like that (not even sure their servers can handle it). easier fixes would be to limit velocity quirks, increase the size of projectile hitboxes (especially on the long axis), or make some of the smaller mechs larger (see most recent rescale thread).

Edited by LordNothing, 29 September 2023 - 06:50 PM.

[Quicksilver Aberration]

LordNothing, on 29 September 2023 - 06:47 PM, said:

the worst offender ive seen is an awesome which i shot repeatedly though the ct with an erppc. it was a non moving target and i was hill humping. i figure if the weapon velocity is high enough the awesome's short axis can fit between the hitboxes of a projectile in to successive frames at 30 ticks/s. this may also be why lights seem to be able to shed damage like the scat in the above video. you could up the server tick, but i dont think events like this happen frequently enough to warrant a drastic change like that (not even sure their servers can handle it). easier fixes would be to limit velocity quirks, increase the size of projectile hitboxes (especially on the long axis), or make some of the smaller mechs larger (see most recent rescale thread).

I'd take all of the above minus limit velocity quirks, but also 32 tick rate is super low for shooters even with HSR (which removes some of the need for faster tick rates).

[LordNothing]

Quicksilver Kalasa, on 29 September 2023 - 08:04 PM, said:

I'd take all of the above minus limit velocity quirks, but also 32 tick rate is super low for shooters even with HSR (which removes some of the need for faster tick rates).

distance traveled is velocity*time. and the time is 1/tickRate. for a weapon with a velocity of 2000ms that comes out to 62.5 meters between ticks. this distance goes up with velocity and down with tick rate. plenty of space in which to phase through a mech. longer projectile hitboxes are the cheap fix, upping tick rate is the expensive fix.

there is also the smart fix of inter-frame hit detection. take the hitboxes from two adjacent frames, if they are convex hulls or cubes, you can merge the geometry into new convex hull to represent the space those objects will move through between the frames. you can do the same with the projectiles, which might just be hit spheres to reduce the complexity (a capsule collider works here) then if these two larger objects intersect, you have a collision between frames, a test that might fail on either of the bracketing frames, and if you need more data you can lerp back to the contact point and run hit detection on those objects again. hit detection is hard, frankly im surprised that it works at all.

[MechMaster059]

LordNothing, on 30 September 2023 - 06:00 AM, said:

... for a weapon with a velocity of 2000ms ...

Weapon velocities don't have any unit of measurement attached to them. They're abstract numbers. I found out by testing with a stopwatch that the actual meters / second of a weapon is the weapon velocity x 80%.

A light PPC has a velocity of 1400. If you stand at 1,120 meters from a target and time how long it takes a light PPC burst to hit the target using a stopwatch it comes out to exactly 1 second.

Not trying to be pedantic here, just trying to be helpful and raise awareness about odd little discrepancies with how the game works. (I have no idea why the weapon velocities aren't straight up m/s values...)

[Quicksilver Aberration]

LordNothing, on 30 September 2023 - 06:00 AM, said:

distance traveled is velocity*time. and the time is 1/tickRate. for a weapon with a velocity of 2000ms that comes out to 62.5 meters between ticks. this distance goes up with velocity and down with tick rate. plenty of space in which to phase through a mech. longer projectile hitboxes are the cheap fix, upping tick rate is the expensive fix.

there is also the smart fix of inter-frame hit detection. take the hitboxes from two adjacent frames, if they are convex hulls or cubes, you can merge the geometry into new convex hull to represent the space those objects will move through between the frames. you can do the same with the projectiles, which might just be hit spheres to reduce the complexity (a capsule collider works here) then if these two larger objects intersect, you have a collision between frames, a test that might fail on either of the bracketing frames, and if you need more data you can lerp back to the contact point and run hit detection on those objects again. hit detection is hard, frankly im surprised that it works at all.

Which is why most FPS "cheat" and make things hitscan.

Edited by Quicksilver Kalasa, 30 September 2023 - 06:34 PM.

[Bulletsponge0]

hit reg has been .... not the greatest .... in this game for more than just a short amount of time. but PGI has never (at least not in the last 8 years) given any indication they they were trying, or even interested, in looking into improving it

[Doppelsoldner King Crab]

living in the boondocks 300-400ms ping since founders

hit reg is the reason why most of my mechs are autocannon or ppc based.
lasers are useable but you could definitely feel hitreg being wonky sometimes.

but on OP's topic. this is probably server lag. there was a recent match where half my team were shooting at a hunchback that was teleporting, it was 10-4 so we were winning.
the teleport was weird, it was like a blinking stalker/dark templar. surrounding the mech, 6 of us could not kill it in 30 seconds. he was not only teleporting horizontally, he was doing it vertically as well!

i thought nothing of it until the end screen showed then my teammates started talking about it in voicechat. we all saw the hunchie teleport so it wasn't only me seeing that

Edited by PrismArc, 30 September 2023 - 07:39 PM.

[LordNothing]

Quicksilver Kalasa, on 30 September 2023 - 06:34 PM, said:

Which is why most FPS "cheat" and make things hitscan.

most games are human scale and you can do things like that. speeds and object scales are significantly more modest and a bullet may as well get to the target instantly. approximations all day every day. or as i call it the boring way to do things.

Edited by LordNothing, 30 September 2023 - 09:11 PM.

[LordNothing]

MechMaster059, on 30 September 2023 - 02:47 PM, said:

Weapon velocities don't have any unit of measurement attached to them. They're abstract numbers. I found out by testing with a stopwatch that the actual meters / second of a weapon is the weapon velocity x 80%.

A light PPC has a velocity of 1400. If you stand at 1,120 meters from a target and time how long it takes a light PPC burst to hit the target using a stopwatch it comes out to exactly 1 second.

Not trying to be pedantic here, just trying to be helpful and raise awareness about odd little discrepancies with how the game works. (I have no idea why the weapon velocities aren't straight up m/s values...)

did you account for how much time it took to hit the button? the time it takes to observe the weapon hit, have your brain recognize what has happened, and press the button again to stop it? account for ping? the error bars of that test are so wide as to account for the difference.

besides all the numbers in xml are in meters/sec, and physics code is a lot easier to write when you use an existing units system. whether or not that scale is actually used internally is a different matter. but i wouldnt do that unless i have weird scale requirements (kerbal does this but it needs to fit an entire solar system). i avoid unit conversions whenever necessary and only modify the units for display if needed (say kph for mech speed readouts). that way i dont have to do a bunch of bug inducing conversions each time i need to run physics code. thats kind of a noob mistake i dont think the ce devs would do, sort of like using left and right hand transformations interchangably.

Edited by LordNothing, 30 September 2023 - 10:32 PM.

[MechMaster059]

LordNothing, on 30 September 2023 - 10:18 PM, said:

did you account for how much time it took to hit the button? the time it takes to observe the weapon hit, have your brain recognize what has happened, and press the button again to stop it? account for ping? the error bars of that test are so wide as to account for the difference.

...

Yup. Try it yourself. I'm using the stopwatch on my smartphone. I know when I'm going to press the mouse fire button and have my thumb hovering millimeters over the stopwatch "start" button. I press them together and watch the LPPC round until it bursts.

Do this 10x and by the 10th time you can get it accurate to within a tenth of a second. No joke.

You can try it with other weapons too. Try it with LRMs. They move slow so it's even easier to time it.

[LordNothing]

MechMaster059, on 01 October 2023 - 12:30 AM, said:

Yup. Try it yourself. I'm using the stopwatch on my smartphone. I know when I'm going to press the mouse fire button and have my thumb hovering millimeters over the stopwatch "start" button. I press them together and watch the LPPC round until it bursts.

Do this 10x and by the 10th time you can get it accurate to within a tenth of a second. No joke.

You can try it with other weapons too. Try it with LRMs. They move slow so it's even easier to time it.

a better measurement can be obtained by recording and comparing the timestamps.

[MechMaster059]

LordNothing, on 01 October 2023 - 06:46 AM, said:

a better measurement can be obtained by recording and comparing the timestamps.

Great, why don't you go ahead and do that and prove it to yourself rather than sitting here second-guessing me.

I know I'm right and if you're too lazy to go in game and measure it yourself it's no big deal.

[Widowmaker1981]

MechMaster059, on 01 October 2023 - 07:28 AM, said:

Great, why don't you go ahead and do that and prove it to yourself rather than sitting here second-guessing me.

I know I'm right and if you're too lazy to go in game and measure it yourself it's no big deal.

I just did the test using video recording and timestamps and.. you appear to be wrong.

Using AC5s with 1500 m/s, on an unskilled mech with no velocity quirks, target was 750m away precisely in testing grounds (used the Cataphract on Alpine). From moment of firing (judged by moment weapons went on cooldown) to moment of impact (judged by reticule flash) took 0.502 seconds. Thats bang on as it should be.

If you dont believe me, i can prove it with the video, just cant really be bothered to upload it to youtube.

Edit: The confusion might be caused by the fact the animation is slightly delayed, if i judge by the animation of the explosion instead of the reticule flash it adds an extra 0.1s or so.

Edited by Widowmaker1981, 01 October 2023 - 08:54 AM.

[MechMaster059]

Widowmaker1981, on 01 October 2023 - 08:50 AM, said:

I just did the test using video recording and timestamps and.. you appear to be wrong....

I have double-checked this and I am indeed wrong.

I verified for sure with LRMs at 380m and I keep getting times of 2.00, 2.05, 2.10. If I were correct it would take 2.50 seconds for LRMs to make that trip and there's no way I'm off by 0.5 seconds pressing the button on my stop watch, therefore, the LRMs are indeed travelling at 190m/s.

My mistake with LPPCs was including the animation time of the flowery burst that happens when they hit. I should have timed to just before that burst animation happens. (Labelling the numbers in the UI m/s would have been helpful as well...)

(You should take no pride in proving me wrong Widowmaker. There is a prophecy that if I'm proven wrong 5 times when I feel certain I'm right the universe will collapse in upon itself. I've been wrong twice before in my life. Once when I was convinced Burning Wish could only pull cards from your sideboard instead of your whole card collection, and another time I was wrong about something else but I can't remember the exact details other than I feel convinced I was wrong about something another time.

That's 3 times now I've been proven wrong... 2 more and the universe ends.)