16 replies to this topic

[Maximum Overkill]

I tested several laser weapon from different distances up to 100 m and with all zoom levels. The outcome is pretty surprising as damage varied heavily.

Recorded on PTS4

Edited by Maximum Overkill, 19 November 2015 - 11:36 AM.

[Dave Forsey]

Damn. Latest patch?

The health gets reset shortly after the damage number is reported - I'll try this myself, but do you get the same behaviour if you let the health reset to 100% before taking the next shot?

[Maximum Overkill]

PublicTest - 17/cl181002

Didn't try it with let them regenerate. But it also seems a bit mech chassis dependent. The MPL tested on the locust 1E does something about 6.3 damage and in the HBK-4P 7.3 damage

Edited by Maximum Overkill, 19 November 2015 - 12:00 PM.

[Raubwurst]

Some days ago I heard someone via VoIP asking the exact same question ("Do you noticed anything fishy with the lasers after last patch?")
Thus may be outside Academy, too!

[Maximum Overkill]

Something I forgot: I tested it also with spl's on anartic cheeta. They did about 5-5.5 damage instead of 6. I think you should test this

[Dave Forsey]

I'm not seeing this - but thanks to your video I'm pretty sure you're not using the Nov 17th patch where this was addressed.

Please let me know if this is not the case.

The problem stems from the original inability to reset the health on the turrets from the flowgraph- so they were set to high to stop you rotters :-) from destroying them too easily and screwing something else up... But I didn't realize the values were single precision floating point."

Here's the full explanation for the curious:

The issue is that turrets have an extremely large health pool (~1 million health) and floating point numbers at this size become inaccurate.

When firing a laser at a turret with a million health, the specific fraction of damage dealt to the turret can't be stored precisely so it rounds it up or down to the nearest representable value.

That causes the turret to take slightly more/less damage than it should from an individual fraction of laser damage, which adds up over the duration of fire.</p>

Note that with a health pool in the millions, a single precision floating point number (like the one we use to store health) breaks down after 2 decimal points, which is enough to throw off the precision of laser damage over time

With floating points around the 100's or 1000's, we reliably have about 5-6 digits of accuracy, which is plenty to keep the issue from showing.

The health is now set to 1000, and the health is reset after the damage is reported.

[zudukai]

is this limited to the academy?

[Scout Derek]

Dave Forsey, on 19 November 2015 - 02:27 PM, said:

I'm not seeing this - but thanks to your video I'm pretty sure you're not using the Nov 17th patch where this was addressed.
Please let me know if this is not the case.

The problem stems from the original inability to reset the health on the turrets from the flowgraph- so they were set to high to stop you rotters :-) from destroying them too easily and screwing something else up... But I didn't realize the values were single precision floating point.

Here's the full explanation:

<div class="action-body flooded">
The issue is that turrets have an extremely large health pool (~1 million health) and floating point numbers at this size become inaccurate.

When firing a laser at a turret with a million health, the specific fraction of damage dealt to the turret can't be stored precisely so it rounds it up or down to the nearest representable value.

That causes the turret to take slightly more/less damage than it should from an individual fraction of laser damage, which adds up over the duration of fire.

Note that with a health pool in the millions, a single precision floating point number (like the one we use to store health) breaks down after 2 decimal points, which is enough to throw off the precision of laser damage over time.

With floating points around the 100's or 1000's, we reliably have about 5-6 digits of accuracy, which is plenty to keep the issue from showing.&quot;

The health is now set to 1000, and the health is reset after the damage is reported.

Cleaned it up for those who didn't like those <p> </p> on Mr. Forsey's reply.

[Maximum Overkill]

I used this client: PublicTest - 17/cl181002
The actual build from the test server. The video was recorded a few hours ago.

I was wondering how the MPL +heat sink changes will affect my 9MPL HBK-4P and was very surprised that I do alphas with 63 damage instead of 54.

Thanks for your explanation, it makes sense. Can this also explain why the SPL do 3 damage and the SL do over 5?

[Wildstreak]

Is this done on a specific range like Targeting Practice or do you just go out and shoot turrets?

[Maximum Overkill]

This is recorded in the "selection" stage. Where you can choose mechs and trainings.

[Dave Forsey]

The public test is a separate branch and may not have integrated the main stream as yet.

I've just tried it live using a wvr-7k with small, med, and large pulse lasers and get consistent damage numbers.
Same correct results with clan lasers.

I appreciate the time you put into making a video. I imagine it was very un-nerving to think this was happening in the live game.

And yes, it seems like the same mechanism would account for the damage difference between laser types

[Tarogato]

I just tested this with a pair of large lasers and pair of medium lasers on a CDA-3M on the live server. No issues at all whatsoever, even shooting while moving perpendicular angles. SRMs seem fine as well, so far.

[Surn]

Dec[8,2] ?

[Tarogato]

Dave Forsey, on 19 November 2015 - 11:54 AM, said:

.

Suggestion: Could the damage be reported sooner? It's kind of awkward firing a shot and waiting more than a second or two for it to tally the damage and report it back. It would be nice if the feedback was more immediate, like 0.5 second delay (you could see the counter going up as it added more damage while you hit it).

Also, is it possible to have this damage reporting mechanic added to the moving mechs in the newly added challenges?

[Dave Forsey]

The delay allows time for weapons to cool-down and re-fire, which lets people determine such things as total damage before over-heating, sustained dps and such. Which you've probably figured out already.

Unfortunately the Flowgraph lacks any interface that would display continuous values properly. The notification box takes 1.5 seconds to open and a second to close and doesn't dynamically update. So I'm a little stuck for options until this sort of feature (and there are other UI items on the wish list) becomes more important than a lot of the other, non-Academy, work going on.

I'd really like continuous reporting on the 'Mechs in the shooting challenges, which would provide good feedback on how well you're keeping the beam on target...

[Jay Z]

Idea: what about using a cockpit monitor to show damage done? That gives you the real estate to show damage done per volley and have a match tally on damage like we will with kills.