30 replies to this topic

[Duke Nedo]

Hey guys & gals,

I've toyed a bit lately with something one could call a mech rating model or whatever, http://mwomercs.com/...e-with-science/

This was for fun, and an attempt to somewhat systematically break down which factors contribute to the over all perceived performance of a mech, and to visualize the distribution of mech performance between all our variants.

Now, the big weakness of this model is that I have awarded sub-scores for all variants, and people pointed out correctly that the model is biased and subjective. Will you guys help me do something about that?

What I look for is 10-20 people who would be willing to fill in a spreadsheet, giving scores to certain mech characteristics. I would also ask for a completely subjective over all mech tier score for each variant for comparison.

If you have mastered ~80 variants or more, have a reasonably analytic mind and would be willing to give your opinion in this format, please drop me a forum PM with your email and I'll send you a spreadsheet to fill in. It would add credibility if some of you guys that are known to post insightful opinions here on the boards would contribute. If I mention three, I am thinking Mcgral, Deathlike and Greenjulius for example. I could name 20 but then someone would feel left out.

If I could get a dataset like that I could present two lists with average and standard deviation:

• Subjective mech Tier for every variant
  
• Modelled mech Tier based on subjective input of broken down characteristics

Could actually be an interesting and useful document to discuss further... who's in?

Edited by Duke Nedo, 16 April 2015 - 02:27 AM.

[CDLord HHGD]

Well, would there be any use for in-game data not variant specific, but loadout specific. Take the following screenshot for example...



I used this back when I posted in a BV thread. Instead of using arbitrary scores for what may be an effective weapon over another (which is also affected by quirks too).

So here we have a 50 ton mech. There's 50 points.
It has 320 armor points. There's 320 points.
It has a heat management of 1.28. There's 128 points.
Jump (hover) distance is a figure (not represented on this screenshot).
Firepower is 45. There's 45 points.

Basically what I am getting at, why not use data already in-game available on every custom mech and tracked in the database already? Maybe I missed the point though....

[Cion]

Please post the results of this poll/study.

Keep it up OP!

It will be interesting to put new mechs in your calculations after they come out.

Edit: PS: also, sadly, I've only mastered 79 variants

Edited by Cion, 16 April 2015 - 04:54 AM.

[dario03]

cdlord, on 16 April 2015 - 04:44 AM, said:

Well, would there be any use for in-game data not variant specific, but loadout specific. Take the following screenshot for example...



I used this back when I posted in a BV thread. Instead of using arbitrary scores for what may be an effective weapon over another (which is also affected by quirks too).

So here we have a 50 ton mech. There's 50 points.
It has 320 armor points. There's 320 points.
It has a heat management of 1.28. There's 128 points.
Jump (hover) distance is a figure (not represented on this screenshot).
Firepower is 45. There's 45 points.

Basically what I am getting at, why not use data already in-game available on every custom mech and tracked in the database already? Maybe I missed the point though....

That wouldn't account for a lot of things like hitboxes, weapon mount locations, quirks, etc, etc. Plus with that system wouldn't something like a LRM Atlas score really high?

[CDLord HHGD]

dario03, on 16 April 2015 - 04:51 AM, said:

That wouldn't account for a lot of things like hitboxes, weapon mount locations, quirks, etc, etc. Plus with that system wouldn't something like a LRM Atlas score really high?

True, it won't account for everything, but it's hard data versus arbitrary opinion. For instance, if properly played and supported, a LRM Atlas is very viable.

[Nazar24]

cdlord, on 16 April 2015 - 04:54 AM, said:

For instance, if properly played and supported, a LRM Atlas is very viable.

NO, it is not, never.
Lrm in general are not viable against any decent player.

[CDLord HHGD]

Nazar24, on 16 April 2015 - 05:01 AM, said:

NO, it is not, never.
Lrm in general are not viable against any decent player.

Like I said, opinion. So hard data available in-game is necessary.

[dario03]

cdlord, on 16 April 2015 - 04:54 AM, said:

True, it won't account for everything, but it's hard data versus arbitrary opinion. For instance, if properly played and supported, a LRM Atlas is very viable.

Well give us some examples of popular builds vs some LRM boats like the LRM Atlas. I can't do it right now but I think in that system you would come up with a higher score for the LRM Atlas than something like popular TimberWolf builds and I would not rank a LRM Atlas higher than those.
Plus those numbers are still arbitrary. Why does a mech get points for weight? Why is armor times 1? why is heat management times 10 and firepower times 1? Is speed being counted? you could probably come up with some 65 ton mechs that have a firepower of 45 and heat management of ~1.3 and I know you can come up with 65 ton mechs that have dual guass only which gives them heat management of what 2.0? Dual guass is good but I don't know if its 55 points better in all situations.
Hard fact numbers are great and all but it leaves out way to much. Its like if you came up with a tier list for a Street Fighter game and said Zangief is the best because he has the most health, the highest damage moves, and only gives up speed. Sounds great and all but it completely ignores combos, range, setups, counters, basically everything.

[Duke Nedo]

cdlord, on 16 April 2015 - 04:44 AM, said:

Well, would there be any use for in-game data not variant specific, but loadout specific. Take the following screenshot for example...

I used this back when I posted in a BV thread. Instead of using arbitrary scores for what may be an effective weapon over another (which is also affected by quirks too).

So here we have a 50 ton mech. There's 50 points.
It has 320 armor points. There's 320 points.
It has a heat management of 1.28. There's 128 points.
Jump (hover) distance is a figure (not represented on this screenshot).
Firepower is 45. There's 45 points.

Basically what I am getting at, why not use data already in-game available on every custom mech and tracked in the database already? Maybe I missed the point though....

I don't think any of the hard data are really hard except for the Armor value, sorry... All other values will change depending on loadout, and to be honest the parameters that affect the performance of a mech is not really the hard values at all.

In my experience the two important parameters are 1. Hardpoints (number, sort and z-coordinates) and 2. Hitboxes (size and distribution). In principle I guess someone who knows Cry engine could extract those as hard data (coordinates and areas), but that would take lots of work.

I think the only way to get descriptive input is to ask many experienced players and take an average. It's worth a shot and see what turns out so it would be cool if anyone would be cool and contribute with their experience here.

Edited by Duke Nedo, 16 April 2015 - 05:19 AM.

[Widowmaker1981]

cdlord, on 16 April 2015 - 04:44 AM, said:

Well, would there be any use for in-game data not variant specific, but loadout specific. Take the following screenshot for example...



I used this back when I posted in a BV thread. Instead of using arbitrary scores for what may be an effective weapon over another (which is also affected by quirks too).

So here we have a 50 ton mech. There's 50 points.
It has 320 armor points. There's 320 points.
It has a heat management of 1.28. There's 128 points.
Jump (hover) distance is a figure (not represented on this screenshot).
Firepower is 45. There's 45 points.

Basically what I am getting at, why not use data already in-game available on every custom mech and tracked in the database already? Maybe I missed the point though....

theres nothing about that rating system i like. It would rate an LRMS and SRMs way higher than their worth, due to their high firepower, because no number on that chart takes into account weapon delivery (Spread/PP/hitscan/etc). it also doesnt take range into account. It also has no means of judging hardpoint placement or hitboxes.. all these things would mean the data that comes out would be wildly wrong.

Whats actually better, a KGC with 2xGauss + 3xLL or a KGC with LRM60 + 4xML? That system would rate the LRMCrab higher than the GaussCrab by quite a bit.....

Additionally the differences between mechs of similar mass would seem tiny due to the HUGE score weighting toward armour points.

[Tombstoner]

cdlord, on 16 April 2015 - 05:02 AM, said:

Like I said, opinion. So hard data available in-game is necessary.

The presence of a single LRM's boat forces players to play in a very specific manner to counter the LRM's . This is tossed out as the end all be all counter that trivializes LRM's. People forget that type of counter forces players into a very predictable counter strategy. one that can be exploited easily. The only way to counter an LRM heavy force that will eventually angle you negating your terrain shield and rain death upon your rear armor, is to use LRM's yourself or massed ECM: Silly one dimensional thinking.

OP - i applaud your efforts but the game is heavily flawed in ways that will heavily affect you ranking system. The relation ship between mech speed, volume, weapon type and placement interaction on TTK, for example needs to be quantified, by PGI. IMO that iss the fundamental source of performance delta leading to "tiers" when mech of the same tonnage should be = but clearly are not.

[Mister Blastman]

I'm experienced. And I've tried. And it has gone nowhere. So, no. I'm just sitting back on the island and watching now.

[Summon3r]

cdlord, on 16 April 2015 - 04:54 AM, said:

a LRM Atlas is very viable.

lol and thats enough of this thread for me

[Kiiyor]

I've found that hard numbers based on firepower and heat dissipation alone are almost meaningless, as in MWO, it seems to be the mechs with the best overall hardpoints and hitboxes that make the biggest impact.

I've been toying with my own data, and have been looking at a usability rating. Put simply, it measures the distance between weapon hardpoints and the cockpit, and tries to account for the amount of real estate you have to expose to be able to employ your weaponry. So far, the fruits of my labor have produced some shiny half finished graphs, but not much else. I think it's a bust, BUT I WILL PERSEVERE.

IMHO, top 3 mechs in the game are the DireWolf, MadCat, and StormCrow. They're the best at what they do. There are other mechs out there that can match or possibly even out perform them at some particular aspect of their overall awesomeness (like the Stalker for mid-long range laser vomit) but they tend to have to specialize to the point where they suck at everything outside their narrow niche.

Also, the torn and dangerous badlands of the PUG killing fields are a completely different beast to the organized chaos of community warfare. I think that if you are attempting to make a meta scale, you should create one for both game modes.

Good luck with your SCIENCE! You should start by pestering Gman (of metamechs fame) if you want part of the ear of the competitive scene.

[CDLord HHGD]

Tombstoner, on 16 April 2015 - 05:43 AM, said:

The presence of a single LRM's boat forces players to play in a very specific manner to counter the LRM's . This is tossed out as the end all be all counter that trivializes LRM's. People forget that type of counter forces players into a very predictable counter strategy. one that can be exploited easily. The only way to counter an LRM heavy force that will eventually angle you negating your terrain shield and rain death upon your rear armor, is to use LRM's yourself or massed ECM: Silly one dimensional thinking.

OP - i applaud your efforts but the game is heavily flawed in ways that will heavily affect you ranking system. The relation ship between mech speed, volume, weapon type and placement interaction on TTK, for example needs to be quantified, by PGI. IMO that iss the fundamental source of performance delta leading to "tiers" when mech of the same tonnage should be = but clearly are not.

How is one LRM boat considered a "LRM heavy force"? I'm not talking about marching in a whole lance of these things. One well equipped boat though (and an atlas has the tonnage for ammo) would force the opfor to, as you say, counter it very specifically, which in my experience, usually exposes them to the non-lrm forces making up the rest of the team. You should hear the "LRMS ARE OP NOOBS" screams when we implement this tactic. Only thing one dimensional here is thinking your way is the only way.

BUT, this isn't a discussion for the viability of a mech build or tactic. I offered what I think should be a factor (not whole, but in part) of a comprehensive ranking system. Take it, leave it OP, but I'll bow out of this discussion to prevent further derailment.

[Duke Nedo]

Kiiyor, on 16 April 2015 - 06:07 AM, said:

I've found that hard numbers based on firepower and heat dissipation alone are almost meaningless, as in MWO, it seems to be the mechs with the best overall hardpoints and hitboxes that make the biggest impact.

I've been toying with my own data, and have been looking at a usability rating. Put simply, it measures the distance between weapon hardpoints and the cockpit, and tries to account for the amount of real estate you have to expose to be able to employ your weaponry. So far, the fruits of my labor have produced some shiny half finished graphs, but not much else. I think it's a bust, BUT I WILL PERSEVERE.

IMHO, top 3 mechs in the game are the DireWolf, MadCat, and StormCrow. They're the best at what they do. There are other mechs out there that can match or possibly even out perform them at some particular aspect of their overall awesomeness (like the Stalker for mid-long range laser vomit) but they tend to have to specialize to the point where they suck at everything outside their narrow niche.

Also, the torn and dangerous badlands of the PUG killing fields are a completely different beast to the organized chaos of community warfare. I think that if you are attempting to make a meta scale, you should create one for both game modes.

Good luck with your SCIENCE! You should start by pestering Gman (of metamechs fame) if you want part of the ear of the competitive scene.

Thanks! I agree there are all sort of things to consider making it in principle impossible for a single Tier value to be accurate, but PGI started that thinking and they based the quirkening on it, so in order to progress balance further I think it would be great if we could help to visualize where we stand with the Tiers now.

It doesn't have to be input from the competitive scene, just people with experience with enough chassi so that they are able to evaluate the mechs relative to each other.

Tombstoner, on 16 April 2015 - 05:43 AM, said:

The presence of a single LRM's boat forces players to play in a very specific manner to counter the LRM's . This is tossed out as the end all be all counter that trivializes LRM's. People forget that type of counter forces players into a very predictable counter strategy. one that can be exploited easily. The only way to counter an LRM heavy force that will eventually angle you negating your terrain shield and rain death upon your rear armor, is to use LRM's yourself or massed ECM: Silly one dimensional thinking.

OP - i applaud your efforts but the game is heavily flawed in ways that will heavily affect you ranking system. The relation ship between mech speed, volume, weapon type and placement interaction on TTK, for example needs to be quantified, by PGI. IMO that iss the fundamental source of performance delta leading to "tiers" when mech of the same tonnage should be = but clearly are not.

I also think it's hard to impossible to only use hard values to determine a mechs Tier, but the input I choose for that model includes the things you mention. There is input for hardpoint capacity and placement/asymmetry as well as hitbox-input which includes both hitbox size, distribution and agility. If enough perceptive people would care to provide their input it would be really interesting to see the result!

[Ghogiel]

Tombstoner, on 16 April 2015 - 05:43 AM, said:

The presence of a single LRM's boat forces players to play in a very specific manner to counter the LRM's . This is tossed out as the end all be all counter that trivializes LRM's. People forget that type of counter forces players into a very predictable counter strategy. one that can be exploited easily. The only way to counter an LRM heavy force that will eventually angle you negating your terrain shield and rain death upon your rear armor, is to use LRM's yourself or massed ECM: Silly one dimensional thinking.

Take lurms yourself, lol no. Best way to counter any lurm heavy teams is direct fire. Brawl or laser vomit range. Don't even need ECM just radar derp. only 1 map, caustic, can possibly be problematic. Every other map, ggclose.

Edited by Ghogiel, 16 April 2015 - 11:02 AM.

[Deathlike]

It is literally impossible to use strict mathematical formulas.

What we can use is "relative" usefulness, depending on the situation.

If you have a 4 ERLL Timberwolf compared to a 4 ERLL Hellbringer, they will only excel in sniping... the Hellbringer will have more indirect durability (ECM) whereas the Timberwolf would be better for better cooling and obvious armor improvements.

You can run a 4 ERMED + 4 ASRM6 Timberwolf and compare that to a 4 ERMED + 4 ASRM6 Maddog, and see that the Maddog is better if you're trying to save on tonnage, but it also comes with poor mobility (no JJ option) and low hardpoints (for the energy weapons) and far less durable (not just because it is a 60 tonner).

You can compare, but numbers don't tell the entire story.

It's easier to sort by role and relative effectiveness in that role. In theory, there are some sniper builds that allow you to brawl to a degree, but of course all of this is related to player skill in being able to execute it (most conjecture involve players of equal skill).

Mech balance requires actual understanding... and not stupidly strict min-maxing, but actual practical usage in real games.

[FupDup]

Deathlike, on 16 April 2015 - 11:22 AM, said:

It is literally impossible to use strict mathematical formulas.

What we can use is "relative" usefulness, depending on the situation.

If you have a 4 ERLL Timberwolf compared to a 4 ERLL Hellbringer, they will only excel in sniping... the Hellbringer will have more indirect durability (ECM) whereas the Timberwolf would be better for better cooling and obvious armor improvements.

You can run a 4 ERMED + 4 ASRM6 Timberwolf and compare that to a 4 ERMED + 4 ASRM6 Maddog, and see that the Maddog is better if you're trying to save on tonnage, but it also comes with poor mobility (no JJ option) and low hardpoints (for the energy weapons) and far less durable (not just because it is a 60 tonner).

You can compare, but numbers don't tell the entire story.

It's easier to sort by role and relative effectiveness in that role. In theory, there are some sniper builds that allow you to brawl to a degree, but of course all of this is related to player skill in being able to execute it (most conjecture involve players of equal skill).

Mech balance requires actual understanding... and not stupidly strict min-maxing, but actual practical usage in real games.

TL;DR: Balance is often more qualitative than quantitative.

Edited by FupDup, 16 April 2015 - 11:35 AM.

[Deathlike]

FupDup, on 16 April 2015 - 11:34 AM, said:

TL;DR: Balance is often more qualitative than quantitative.

But... we gotta balance by "normalizing everything" bro. AC2s need to have 3DPS man because "numbers". LPLs need heat increase for no particular reason, other than "normalization".

I mean, why field test stuff? It's so Lostech man.