91 replies to this topic

[Widowmaker1981]

MrMadguy, on 08 February 2016 - 02:47 AM, said:

I'm against convergence removal, but I don't see problem here. As convergence is function of 'Mech positions you have to track 'Mech positions only - convergence may be calculated form position, when needed.

No, you have to track what the crosshair was aimed at, in order to know the current convergence level, and as such has far it needs to move to converge on your current aim point.

Delayed convergence USED TO BE IN THE GAME. PGI removed it because it wouldn't work with HSR. they didnt make that up, why would they remove a feature they spent time implenting?

[Troutmonkey]

MeiSooHaityu, on 08 February 2016 - 03:04 AM, said:

then manual set fixed convergence would be my other choice (convergence manually set per weapon group). No convergence really never made much sense to me and still doesn't.

Try the demo. I think you might change your mind. Fixed convergence is actually worse than no convergence because if you set your convergence point to short your shots will always miss. I made this demo to visually demonstrate just how bad no convergence and fixed convergence are.

[DovisKhan]

This looks amazing

[MeiSooHaityu]

Troutmonkey, on 08 February 2016 - 03:19 AM, said:

Try the demo. I think you might change your mind. Fixed convergence is actually worse than no convergence because if you set your convergence point to short your shots will always miss. I made this demo to visually demonstrate just how bad no convergence and fixed convergence are.

See I think that is a flaw in your video.

Why would no convergence be better? If the lasers start at the same point in both scenarios, then fixed convergence and no connvergence should still hit the target at short ranges (just with the pattern being a bit tighter with fixed convergence.

There should be no time where fixed convergence should be worse then none except at extreme ranges (after the convergence has crossed and now is scattered).

P.S. I think I get what you are saying now, but that is if you set your convergence at a really short range (which I wouldn't do). Even if your weapon (say SLaser) has an optimal range of 90m, I wouldn't set convergence to 90m, but maybe approx it's max of 180m (or farther). Yes it is outside it's optimal, but like you pointed out, you don't want a crazy cross pattern going. Overall, when you set a fixed convergence, you are trying to set your weapons (and pattern) to best suite your engagement ranges and what spread you can live with.

This is why I said manually set fixed convergence. If the convergence is automatically placed by the game at a weapon's optimal range by default, you are going to get some crazy patterns happening. You need to be able to focus them for what works for your build even if some weapons are crossing before or after their ideal optimal damage point.

Edited by MeiSooHaityu, 08 February 2016 - 03:37 AM.

[Widowmaker1981]

Personally, i think the solution is for players to take responsibility for twisting damage themselves, and not expect to be able to stare at targets and have the damage spread for them, but maybe thats just me.

When i die with pristine STs and a missing CT, i tend to blame myself, not the game. (unless its an EBJ lol)

[TexAce]

Widowmaker1981, on 08 February 2016 - 03:32 AM, said:

Personally, i think the solution is for players to take responsibility for twisting damage themselves, and not expect to be able to stare at targets and have the damage spread for them, but maybe thats just me.

When i die with pristine STs and a missing CT, i tend to blame myself, not the game. (unless its an EBJ lol)

Yeah show me how that twisting works for you in a Stalker, or a STD KGC, or an STD300 Atlas.

Edited by TexAce, 08 February 2016 - 03:37 AM.

[Widowmaker1981]

TexAce, on 08 February 2016 - 03:36 AM, said:

Yeah show me how that twisting works for you in a Stalker, or a STD KGC, or an STD300 Atlas.

Im confused, you just suggested 2 of the mechs that are best at twisting dmg in the game, stalker and atlas. (Don't run an Atlas or a KGC with a 300 engine anyway, a 3.0 agility rating is below usable levels since skill tree nerfs)

If you remove all requirement for the defending mech to twist in order to spread damage, and force all dmg to be spread by mechanics, doesn't that cater to bads who like to stare down their targets and not twist?

Edited by Widowmaker1981, 08 February 2016 - 03:58 AM.

[MrMadguy]

Widowmaker1981, on 08 February 2016 - 03:07 AM, said:

No, you have to track what the crosshair was aimed at, in order to know the current convergence level, and as such has far it needs to move to converge on your current aim point.

Delayed convergence USED TO BE IN THE GAME. PGI removed it because it wouldn't work with HSR. they didnt make that up, why would they remove a feature they spent time implenting?

You have to track what the crosshair was aimed at anyway, even without convergence, otherwise you won't know, whether you hit target or not. Delayed convergence may have another problem - it's too unreliable. When convergence is instant, all you have to know - is whether other 'Mech was under your crosshair or not. But with delayed convergence you have to ensure, that 'Mech was under crosshair for the whole convergence duration - so any crosshair fluctuations due to desync and server time quantization - may kill convergence and make game unplayable.

Troutmonkey, on 08 February 2016 - 02:56 AM, said:

Added delayed convergence video to original post. Updated download to include new build of the test scene.

Delayed convergence works incorrectly in your demo. It shouldn't completely disappear, when you don't aim at something - it should converge at terrain, as it does now. The difference between instant and delayed convergence - isn't an absence of convergence, when you don't aim at something. This difference - is that with delayed convergence it will take some time for convergence to change from one distance to another.

Edited by MrMadguy, 08 February 2016 - 04:05 AM.

[Widowmaker1981]

MrMadguy, on 08 February 2016 - 04:00 AM, said:

You have to track what the crosshair was aimed at anyway, even without convergence, otherwise you won't know, whether you hit target or not. Delayed convergence may have another problem - it's too unreliable. When convergence is instant, all you have to know - is whether other 'Mech was under your crosshair or not. But with delayed convergence you have to ensure, that 'Mech was under crosshair for the whole convergence duration - so any crosshair fluctuations due to desync and server time quantization - may kill convergence and make game unplayable.

Currently, you have to track what the crosshair was aimed at at the instant of firing, and only at the instant of firing. With Delayed convergence you need historical data on where the crosshair has recently been pointed, so you can know what range the weapons are currently converged at.

[oldradagast]

As I've said for the billionth time, the goal is not to "make it so you can't hit what you're aiming at" but to end the out of theme and grossly unbalanced idiocy of having all your damage magically focus on a single pixel you're aiming at. That is at the heart of the pinpoint meta problem, which has dominated this game since somebody stuck 2 Gauss rifles or AC20's on a Catapult.

Delayed convergence won't with HSR, and zero convergence is supported by a very small percentage of the community. A Cone of Fire is the only logical solution. The game can already handle it - machine guns, MASC, and Jumpjets demonstrate this - so all we need is a small one in effect at all times that will reduce the brutal efficiency of long range, pinpoint alphas. The cone can also be influenced by heat, sensor lock, etc. giving the designers more space in which to add features.

MWO is about the only game out there that has such brutally effective, perfect accuracy long-range weapons, and this "feature" has been at the heart of nearly every major game balance issue since some wise-guy strapped a pile of PPC's or a pair of AC20's or Gauss to a mech while the game was still in Beta. Every solution presented so far has been an illogical band-aid compared to a Cone of Fire. Nearly every other major first-person shooter out there has something like this implemented, and MWO's engine can handle it, so it is just absurd - and the "skillz people" - that we haven't done it yet.

Widowmaker1981, on 08 February 2016 - 03:57 AM, said:

Im confused, you just suggested 2 of the mechs that are best at twisting dmg in the game, stalker and atlas. (Don't run an Atlas or a KGC with a 300 engine anyway, a 3.0 agility rating is below usable levels since skill tree nerfs)

If you remove all requirement for the defending mech to twist in order to spread damage, and force all dmg to be spread by mechanics, doesn't that cater to bads who like to stare down their targets and not twist?

Twisting does't do you a lick of good when a pair of Gauss rounds hit you for 30 damage, pinpoint to one location. Similarly, you cannot twist away the initial blast of damage from a pile of hit-scan lasers, also doing a pile of pinpoint damage. Long range pin-point alphas have been at the heart of nearly every balance issue in this game. They are unrealistic, out of theme for Battletech, and make for boring gameplay and strange balance decisions, like ghost heat.

[Wolfways]

So...what's the problem with fixed convergence? If it's fixed at the weapons optimal range won't that mean that weapons are better used at their intended range?
(Sorry if I'm being a bit thick. I'm tired and trying to stop smoking )

Also, I feel this is appropriate for MWO losing its perfect convergence.
"I felt a great disturbance in my mech, as if thousands of voices suddenly cried out in terror "But meh skillz!" and were suddenly silenced."

[Widowmaker1981]

oldradagast, on 08 February 2016 - 04:26 AM, said:

Twisting does't do you a lick of good when a pair of Gauss rounds hit you for 30 damage, pinpoint to one location. Similarly, you cannot twist away the initial blast of damage from a pile of hit-scan lasers, also doing a pile of pinpoint damage. Long range pin-point alphas have been at the heart of nearly every balance issue in this game. They are unrealistic, out of theme for Battletech, and make for boring gameplay and strange balance decisions, like ghost heat.

Yes, it does. Because it makes it harder for the firer to get the rounds onto the component that he wants to. Twisting does not have to be reactive, it can be preventative. Some people are better at spreading damage and staying alive than others, which implies to me that there is defensive skill involved.

I'm not completely against a CoF type mechanic, assuming it was controllable (i.e. when not moving and at zero heat it had better be possible to hit the same component with my weapons), but i do accept that it does remove an element of player skill - its much harder to track an enemy with your weapons accurately while you are moving as opposed to standing still, and being able to do so is a large defensive advantage. Make it impossible, and you just took away one of the advantages of being a better player.

[SQW]

MWO just needs something like how FPS such as the original counter strike dealt with bullet accuracy.

You have a small area where your shots will scatter within a certain cone. The larger the movement such as running, twisting etc, the bigger the cone. You'll have to be very lucky to hit the same torso at medium to long range with an alpha so TTK goes way down and laser boating suddenly loses its 'charm'. That's pretty much how the book describes target lock anyway - the computer makes a best guess effort instead of giving you pinpoint accuracy.

There's no need to have manual convergence or auto convergence or whatever people's trying to mess with over the years.

Edited by SQW, 08 February 2016 - 05:14 AM.

[VirtualSmitty]

Well done sir!

[TexAce]

SQW, on 08 February 2016 - 05:14 AM, said:

MWO just needs something like how FPS such as the original counter strike dealt with bullet accuracy.

You have a small area where your shots will scatter within a certain cone. The larger the movement such as running, twisting etc, the bigger the cone. You'll have to be very lucky to hit the same torso at medium to long range with an alpha so TTK goes way down and laser boating suddenly loses its 'charm'. That's pretty much how the book describes target lock anyway - the computer makes a best guess effort instead of giving you pinpoint accuracy.

There's no need to have manual convergence or auto convergence or whatever people's trying to mess with over the years.

I just posted a thread about CoF for exactly that, and because I didnt want it to mix up with this thread, which discusses convergence techniques too.

Take a look: http://mwomercs.com/...-with-pictures/

[Troutmonkey]

Wolfways, on 08 February 2016 - 04:29 AM, said:

So...what's the problem with fixed convergence? If it's fixed at the weapons optimal range won't that mean that weapons are better used at their intended range?
(Sorry if I'm being a bit thick. I'm tired and trying to stop smoking )

Also, I feel this is appropriate for MWO losing its perfect convergence.
"I felt a great disturbance in my mech, as if thousands of voices suddenly cried out in terror "But meh skillz!" and were suddenly silenced."

Did you see the video's or try the demo? With fixed convergence you're shots with fly off wildly to the side if your convergence distance is closer than the target. As for manual convergence just download the game files and see how quickly you can move around while setting the convergence point and still hit stuff.

MrMadguy, on 08 February 2016 - 04:00 AM, said:

You have to track what the crosshair was aimed at anyway, even without convergence, otherwise you won't know, whether you hit target or not. Delayed convergence may have another problem - it's too unreliable. When convergence is instant, all you have to know - is whether other 'Mech was under your crosshair or not. But with delayed convergence you have to ensure, that 'Mech was under crosshair for the whole convergence duration - so any crosshair fluctuations due to desync and server time quantization - may kill convergence and make game unplayable.


Delayed convergence works incorrectly in your demo. It shouldn't completely disappear, when you don't aim at something - it should converge at terrain, as it does now. The difference between instant and delayed convergence - isn't an absence of convergence, when you don't aim at something. This difference - is that with delayed convergence it will take some time for convergence to change from one distance to another.

My implementation is hastily done, as I noted. It's all magic number to get something working with minimal effort but it's close enough to make a point.

Spoiler

//delayed
		else if (convergeMode == 4)
		{
			ray = cam.ScreenPointToRay(new Vector3(Screen.width / 2, Screen.height / 2, 0));
			if (Physics.Raycast(ray, out hit))
			{
				//delayedDistance = hit.distance;
				if (hit.distance > delayedDistance)
					delayedDistance += Time.deltaTime * delaySpeed;
				else
					delayedDistance -= Time.deltaTime * delaySpeed;
					//Mathf.Lerp(delayedDistance, Vector3.Distance(transform.position, hit.point), Time.deltaTime * delaySpeed);
				Debug.Log("dis " + delayedDistance + "- " + hit.distance);
			}
			Quaternion dir = transform.rotation;
			Vector3 point = ray.GetPoint(delayedDistance);
			transform.LookAt(point);
			float dif = Mathf.Abs(delayedDistance - hit.distance);
			float lerp = Mathf.Clamp01((5 - dif) / 5);
			ray.direction = Vector3.Slerp(transform.forward, ray.direction, lerp);
			transform.rotation = dir;
		}
 
if(Physics.Raycast(ray,out hit)) {
			laser.gameObject.SetActive(Input.GetMouseButton(0));
			laser.transform.position = Vector3.Lerp(transform.position, hit.point, .5f);
			laser.LookAt(transform);
			laser.transform.localScale = new Vector3(.01f, .01f, hit.distance);
 
			uiElement.transform.position = cam.WorldToScreenPoint(hit.point);
 
			if (hit.rigidbody != null && hit.rigidbody.name.Contains("Plane"))
			{
				uiElement.color = Color.red;
			}
			else
			{
				uiElement.color = Color.green;
			}
		}

Edited by Troutmonkey, 08 February 2016 - 05:33 AM.

[Lightfoot]

Just make the mechs tougher and hitboxes smaller or overlapping. Hit-boxes over-lapping does the same thing as imperfect convergence mostly. BTW we had delayed convergence in Closed Beta and they said it was producing too much lag as I remember.

[Troutmonkey]

oldradagast, on 08 February 2016 - 04:26 AM, said:

Delayed convergence won't with HSR, and zero convergence is supported by a very small percentage of the community. A Cone of Fire is the only logical solution. The game can already handle it - machine guns, MASC, and Jumpjets demonstrate this - so all we need is a small one in effect at all times that will reduce the brutal efficiency of long range, pinpoint alphas. The cone can also be influenced by heat, sensor lock, etc. giving the designers more space in which to add features.

I'm very much in the Cone of Fire camp, but believe it should be done as per Homeless Bill's suggestion where you only get it if your breach your "Targetting Computer Limit" by firing too many weapons at once. That it's a choice between accuracy and instant damage.

Lightfoot, on 08 February 2016 - 05:34 AM, said:

Just make the mechs tougher and hitboxes smaller or overlapping. Hit-boxes over-lapping does the same thing as imperfect convergence mostly. BTW we had delayed convergence in Closed Beta and they said it was producing too much lag as I remember.

Pinpoint requires 1 raycast.
Fixed requires 1 raycast.
Manual requires 2 rays and a 1 raycast
Delayed requires 2 rays, 2 raycast and historical data on all previous aim points

[Bobzilla]

Widowmaker1981, on 08 February 2016 - 04:54 AM, said:

Yes, it does. Because it makes it harder for the firer to get the rounds onto the component that he wants to. Twisting does not have to be reactive, it can be preventative. Some people are better at spreading damage and staying alive than others, which implies to me that there is defensive skill involved.

I'm not completely against a CoF type mechanic, assuming it was controllable (i.e. when not moving and at zero heat it had better be possible to hit the same component with my weapons), but i do accept that it does remove an element of player skill - its much harder to track an enemy with your weapons accurately while you are moving as opposed to standing still, and being able to do so is a large defensive advantage. Make it impossible, and you just took away one of the advantages of being a better player.

I'd have to really disagree with you. The skill set for this game isn't solely point and click twitch (tho is seems like it with perfect convergence). Changing perfect convergence would just put more weight on all the other skills involved like twisting, positioning and map awareness.

I don't think anyone thinks that aim=skill only, if this was the case, adding a CoF wouldn't make a lick of difference anyways as poor aim is still going to lose to good aim, even if both are stationary.

[Product9]

Troutmonkey, on 08 February 2016 - 05:37 AM, said:

Pinpoint requires 1 raycast.
Fixed requires 1 raycast.
Manual requires 2 rays and a 1 raycast
Delayed requires 2 rays, 2 raycast and historical data on all previous aim points

Instead of raycasting for each arm or whatever, why not just use a single raycast and then fire weapons based on some previously computed quantity? For example, you could adjust the fire of the weapon based on an angle/time value that is already known by both the server and client. Assuming each laser weapon uses its own raycast anyway (with projectiles being instanced objects with their own firing angles themselves), it wouldn't add any extra overhead apart from a simple table of values.

I guess that isn't all that different from a cone of fire mechanic, only with a more predictable result.