14 replies to this topic

[Antares102]

I've created a spread sheet which ranks (almost) all weapons according to effectiveness.
The formulas I used are inside the spread sheet (https://ufile.io/jnm5r).
If somebody thinks the formulas I used are rubish please propose an improvement.



Remarks:

• Clan AC have duration because cooldown does not start before all volleys are fired
  
• "Spread" (1=best) is subjective based on my experience
  
• "Ammo" is subjective and reflects the amount of ammunition I would bring
  
• "100/dmg" should reflect the face time i.e. how often you have to expose yourself to do 100 damage
  
• Range is not calculated linear, it is based on a sigmoid function with 400m optimal range as the center point which represents a factor of 1.
  
• Max range is not considered
  
• Min range is not considered
  
• The Gauss charge up is simply added to the cooldown
  
• MG / Flamers are not considered

Edited by Antares102, 02 July 2017 - 07:12 AM.

[El Bandito]

Looks well thoughtout. I assume ASRM6 didnt make it due to spread?

[Kiiyor]

That DPS on AC10's is something I haven't really paid attention to before.

I'm going to put 4 in an Annihilator.

Nice.

Here's the calculation from the spreadsheet with heading names instead of column references:

=(10000*DPS*Range factor*IF(Velocity<>0,Velocity/1000,1))/(Heat*(Slots+Tons+Ammo*2)*(Slots+Tons+Ammo*2)*100/dmg*100/dmg*(1+Duration)*(1+Duration)*Spread*Spread)

Heat seems to be a comparatively enormous factor when compared to other stats like DPS and damage, which is why the gauss scores so much higher than anything else, I guess. I may have weighted it a little differently, but it's very interesting indeed.

If you don't mind me asking, what was your thought process when arriving with the 4.39 range factor value in your parameters sheet?

[Tarogato]

I have my own spreadsheet that does the same thing as this, but I don't dare share it until it actually poops out numbers that are consistent with how I and the competitive community as a whole rates weapons.

Ie., before the energy rebalance, cSPL was the best weapon in the game, hands down, absolutely by a long shot. If you put in the old numbers for that weapon, it should be rated as #1, if your algorithm is accurate, nothing should beat the cSPL. After that, the cLPL is good, the cERSL, the LPL, the cERPPC, the cLB10... the worst weapon should be the SL by a considerable long shot.

Being as I can't get my algorithm to do this yet... it's worthless atm.

[LordNothing]

Kiiyor, on 02 July 2017 - 08:42 PM, said:

That DPS on AC10's is something I haven't really paid attention to before.

I'm going to put 4 in an Annihilator.

Nice.

Here's the calculation from the spreadsheet with heading names instead of column references:

=(10000*DPS*Range factor*IF(Velocity<>0,Velocity/1000,1))/(Heat*(Slots+Tons+Ammo*2)*(Slots+Tons+Ammo*2)*100/dmg*100/dmg*(1+Duration)*(1+Duration)*Spread*Spread)

Heat seems to be a comparatively enormous factor when compared to other stats like DPS and damage, which is why the gauss scores so much higher than anything else, I guess. I may have weighted it a little differently, but it's very interesting indeed.

If you don't mind me asking, what was your thought process when arriving with the 4.39 range factor value in your parameters sheet?

ive been exploiting the ac10's cooldown for some time. seems ever since it got the ammo buff. even before then it could net you some good dps if the quirks were there. ctf-im could always run 3, but the ammo wasnt always there, the ammo buff directly made those 3 ac10s op. by some miracle it somehow managed to fly under the radar for a very long time. its pretty weak now since the skill trees hit its mobility. part of its charm was the shock and awe factor. if you encountered a single mech in a choke point, you could usually take it out before he has a chance to determine that you are running an xl. the boom hammer is kind of taking its niche though.

Edited by LordNothing, 02 July 2017 - 10:34 PM.

[Paigan]

Antares102, on 02 July 2017 - 07:02 AM, said:

I've created a spread sheet which ranks (almost) all weapons according to effectiveness.
The formulas I used are inside the spread sheet (https://ufile.io/jnm5r).
If somebody thinks the formulas I used are rubish please propose an improvement.



Remarks:

• Clan AC have duration because cooldown does not start before all volleys are fired
  
• "Spread" (1=best) is subjective based on my experience
  
• "Ammo" is subjective and reflects the amount of ammunition I would bring
  
• "100/dmg" should reflect the face time i.e. how often you have to expose yourself to do 100 damage
  
• Range is not calculated linear, it is based on a sigmoid function with 400m optimal range as the center point which represents a factor of 1.
  
• Max range is not considered
  
• Min range is not considered
  
• The Gauss charge up is simply added to the cooldown
  
• MG / Flamers are not considered

Nice idea, but obviously not matured (to phrase it diplomatically).
Because:

1.)
If one weapon has almost 4.0 and (almost) all other weapon are below 1.0, something can't be right.
There is obviously some predjudice involved, over-weighing some variables and ignoring others.
Either to glorify or to blame a certain weapon. Which one applies is more less irrelevant.

2.)
Is the C-Gauss really 59 times more effective than a CLPL?
Absolutely ridiculously not.

3.)
Are almost ALL other weapons from CERML to C-LB10-X really so close together?
If the C-Gauss is 59 times more effective then the CLPL, is the CLPL really only 1.5 times as effective as the LBX?

That formula must be so immensly off that it would be more accurate to estimate every effectiveness subjectively by hand.


You didn't factor in a lot of stuff.
For example:
- Every Mech has built-in HS, meaning the first few heat producers are "free", while the Gauss does not benefit from this at all.
- Ammo and Gauss can explode.
- Bigger weapons are much more prone to crits than Energy weapons are
- Ammo weapons waste total damage potential beyond optimal range. Energy weapons do not.
- Laser have perfect accuracy. Although that admittedly kind of counts for gauss as well. But not for the rest.
- Due to being hard-capped at 2 Gauss being able to fire simultaneously, the maximum Gauss alpha can be 30. With 4 LPL, it can be 52 if you're properly fitted and skilled to take the ghost heat. At comparable range. With rather short duration.
- probably dozens of other points, but you'll get the idea.


Too bad you have blocked me and can't even read this post.
But everyone else will.

Edited by Paigan, 02 July 2017 - 11:48 PM.

[ToXikyogHurt]

TL;DR response:

I don't think your range formula has correct parameters.
SRM maximum range is an important break point.
I think the difference in value between (Clan) ER-Medium and (Clan) Large Pulse range (400m-600m) isn't as pronounced as ~80%.



Full-fat response:

Antares102, on 02 July 2017 - 07:02 AM, said:

If somebody thinks the formulas I used are rubish[sic] please propose an improvement.

Range is not calculated linear, it is based on a sigmoid function with 400m optimal range as the center point which represents a factor of 1.
Max range is not considered
Min range is not considered

I'm not sure I fully agree with the use of a simple sigmoid to model the value of range, at least not with the parameters you've used. I think there are 'plateaus of effectiveness' created by things like SRMs and map design - the distances between commonly used pieces of cover play a large role. What would also be good would be range of charts, to account for how easy it is to close into firing distance depending on the speed of your 'Mech.


I certainly think that some consideration needs to be applied to min/max ranges. Particularly for LRMs, SRMs & PPCs, which have hard cutoffs and therefore 0 value outside of their optimal ranges.


In general, flattening the curve to the right seems fine, but there are times (looking at you, Frozen City) when having ER-PPCs means you can fire effectively for several minutes and take 0 return damage. It's a mistake for the opposing team to let you do this, but it happens and it's not reflected in your formula. The value difference between the 600m Clan LPL and 810m ER-PPC is only 1.98/1.8 = ~10% where in actual play the LPL cannot touch the ERPPC in response.

Flattening to the left seems a reasonable thing to do for Light 'Mechs but not for Assaults - and averaging across 'Mech weights/speeds is a little misleading. For a slow brawler, increasing your range from 110m (SPL) to 165m (cSPL) could indeed be the ~80% value increase given by your numbers. For fast light 'Mechs those 40m are likely to be crossed almost instantly and thus irrelevant. Of course sometimes it's impossible to get closer (because there's no cover) and then 80% isn't a large enough figure, but perhaps (probably?) 40m range also isn't enough to make a difference either.


Being able to exceed SRM range gives a notable advantage I don't think you quite capture. It's sort of modelled by your range formula having a steep slope in the 300-500m region, but I'm not convinced that 450m-550m should provide the same value increase as 250m-350m. The latter takes you safely outside SRM range where the former does not.

I think you'd need to measure the average distance between major pieces of cover in areas where fights often happen and apply some heuristic to reflect these plateaus where a minimum value is needed but grossly exceeding this doesn't provide much benefit. The top of HPG is ~270m across, the caldera of Caustic is ~450m, the centre of Mining is ~300m. Going from 200m to 250m really gains you nothing in any of these circumstances, having 700m range is of no benefit over 500m.



I've plotted some of your range values below for comparison.




I would, in any case, use the mid-point of a weapon's optimal and maximum range as the basis for range value calculations - not just its optimal range. This accounts better for the (current) behaviour of clan ER lasers, and represents the ability to poke for useful damage slightly beyond optimal range (with lasers but not SRMs). See below for plots of those values.

Edited by ToXikyogHurt, 03 July 2017 - 05:17 AM.

[Davegt27]

I am sure we have a lot of engineers on MWO that could come up with formulas

Quote

• Range is not calculated linear, it is based on a sigmoid function with 400m optimal range as the center point which represents a factor of 1.
  
• Max range is not considered
  
• Min range is not considered
  
• The Gauss charge up is simply added to the cooldown
  
• MG / Flamers are not considered

• 
  
• you start with a wrong definition of effectiveness (why because your only using 400 meters)
  
• 
  
• so a weapons combat effectiveness can be defined as
  
• 
  
• :ability to deliver destructive power to a target over time
  
• 
  
• range has to be consider as long as destructive pwr is being delivered to a target
  
• we could exchange power for the phrase "hit points" (more game oriented) as long as hit points is defined
  
• so an example is ER LL can start delivering destructive pwr/hit points against a target at its max range
  
• and it will continue to deliver that power for the beam duration

JMTCW

[Antares102]

Great thank you for your feeback, keep it coming

El Bandito, on 02 July 2017 - 06:01 PM, said:

Looks well thoughtout. I assume ASRM6 didnt make it due to spread?

Well, its one factor but the formulas obviously tried to consider all factors.
Just play around with the values if you disagree with what I put in.

Kiiyor, on 02 July 2017 - 08:42 PM, said:

If you don't mind me asking, what was your thought process when arriving with the 4.39 range factor value in your parameters sheet?

I wanted to have 200m at 20%, effectiveness 600m at 80% effectiveness (pareto principle inspired) and 400m at 50%.
In terms of range everything below 200m sucks hard IMO and above 600m only gives you an edge on a few long range maps

Paigan, on 02 July 2017 - 11:33 PM, said:

1.)
If one weapon has almost 4.0 and (almost) all other weapon are below 1.0, something can't be right.
There is obviously some predjudice involved, over-weighing some variables and ignoring others.
Either to glorify or to blame a certain weapon. Which one applies is more less irrelevant.

I already thought about normalizing the values but it is corrent that some variables are obviously overweight right now.

Paigan, on 02 July 2017 - 11:33 PM, said:

2.)
Is the C-Gauss really 59 times more effective than a CLPL?
Absolutely ridiculously not.

The effectivness calculation is only supposed to provide a rank not a direct comparison.

Paigan, on 02 July 2017 - 11:33 PM, said:

You didn't factor in a lot of stuff.
For example:
- Every Mech has built-in HS, meaning the first few heat producers are "free", while the Gauss does not benefit from this at all.
- Ammo and Gauss can explode.
- Bigger weapons are much more prone to crits than Energy weapons are
- Ammo weapons waste total damage potential beyond optimal range. Energy weapons do not.
- Laser have perfect accuracy. Although that admittedly kind of counts for gauss as well. But not for the rest.
- Due to being hard-capped at 2 Gauss being able to fire simultaneously, the maximum Gauss alpha can be 30. With 4 LPL, it can be 52 if you're properly fitted and skilled to take the ghost heat. At comparable range. With rather short duration.
- probably dozens of other points, but you'll get the idea.

Yes a lot of factors are not considered. I already thought about some variables but it will be more tricky to factor them in.
After all I only use my subjective judgement to rate the results if they are correct or not.
Thats's why I put up the WIP result here to redeive feedback.

ToXikyogHurt, on 03 July 2017 - 05:13 AM, said:

Full-fat response:

I am trying to consider your feedback with the next update.

Edited by Antares102, 03 July 2017 - 11:03 AM.

[Y E O N N E]

This seems a lot like what I did for my laser comparison tables...

[DerMaulwurf]

If I were you, I would take a hard look at your results . And then do an honest appraisal, if it matches your experience with in-game performance. An obvious example is that you rate the isAC/10 5 times higher then the cUAC/10.

One important factor you're missing is compatibility with poking/twisting. The AC/10 has mighty stats on paper. But in practice you either pay in dps if you twist; or pay in survivability if you don't.

I applaud the attempt to get some mathematical approach onto weapon stats, but I think that finding the proper weighting between all factors that are relevant in live performance is going to be too tough.

[Antares102]

DerMaulwurf, on 03 July 2017 - 11:22 AM, said:

If I were you, I would take a hard look at your results . And then do an honest appraisal, if it matches your experience with in-game performance. An obvious example is that you rate the isAC/10 5 times higher then the cUAC/10.

As I said earlier the effectivness values should only provide a rank and should not be used for comparing them as factors.

DerMaulwurf, on 03 July 2017 - 11:22 AM, said:

One important factor you're missing is compatibility with poking/twisting. The AC/10 has mighty stats on paper. But in practice you either pay in dps if you twist; or pay in survivability if you don't.

This is factored in with the shots for 100 damage because less shots means less face time thus weapons with high alpha are better.

DerMaulwurf, on 03 July 2017 - 11:22 AM, said:

I applaud the attempt to get some mathematical approach onto weapon stats, but I think that finding the proper weighting between all factors that are relevant in live performance is going to be too tough.

I agree on the toughness but its fun to try.

Edited by Antares102, 03 July 2017 - 11:27 AM.

[Davegt27]

another way to approach the problem would be to use player stats

gather top player stats on there weapons could give a picture of the weapons effectiveness

Edited by Davegt27, 03 July 2017 - 11:36 AM.

[Humpday]

This is a very educated approach to a videogame, geez lol
I'm all for it, but man, most the time, I'm like...hey this fits, k lemmi go shoot someone and see what happens.

Keeping going, its interesting.

[Antares102]

Davegt27, on 03 July 2017 - 11:35 AM, said:

another way to approach the problem would be to use player stats

gather top player stats on there weapons could give a picture of the weapons effectiveness

Yes I thought of that.
What I actually want to do is creating a model that reflects a current game state in terms of weapon balance to predict changes to weapons or even better balance weapons.
However, I wont get my hands on PGIs server statistics therefore I have to use my (and your) judgement