90 replies to this topic

[Y E O N N E]

Tristan Winter, on 16 October 2016 - 04:06 PM, said:

At least I learned a new term. "Orthogonal argument". I have no idea what it means though.

I've also learned that a mech can be the most overpowered mech in its class, without being the best mech in its class.

It means the points are not on on the same plane; they don't counter each other, rather they say something different about the topic.

[Pariah Devalis]

TL;DR:

If you can't aim, light mechs OP, dude!





Unless you bring streaks.

Or stick with your team.

Or git gud.

[Tahawus]

Yeonne Greene, on 16 October 2016 - 04:05 PM, said:

Well, by nature it won't. A Light makes precise shots and downs 'Mechs efficiently while Assault just hammer away at things until they die. But it should take about the same level of effort for me to maneuver in and get those precise shots as it takes a Kodiak to hammer away.

But that's not what is being measured in the leader board stats. The locust, arctic cheetah, kodiak... all dish out disproportionately high amounts of damage, though proportionately to a baseline adjusted based on their tonnage. Looking back at the model, the model's slope is not particularly steep with respect to tons (i.e. the intercept term is comparatively large compared to the slope's marginal value.)


A more theoretical discussion.

Let's assume that there should be parallel functions, tricky though they might be to formulate, expressing the value of other roles played by mechs other than just dealing damage such as tanking, scouting, harassing and similar values that are poorly represented in the current metric. The sum of those values should be near constant to establish mech to mech parity across all mechs and across roles.

The damage function should look something a lot like my my modeled response. A general increase in damage output per ton, with variations up and down based on the role that the mech is configured for (i.e. a light harasser should be able to do more damage than a mech of the same weight configured for scouting).

So what we're really missing is an effective treatment in the game of information warfare and a way to capture the value of harassment (other than damage).

To sum up, I approached the analysis with a limited scope, trying to idealize the the damage function while not adequately representing the failure of the game to represent (and reward) the other roles that mechs should be playing. I still believe that my efforts provide a good and valuable quantitative indication of what the expected damage a mech should be doing (as measured in the leader boards) and provides valuable insight into which mechs over and under perform compared to the rest of the mechs (again based on the leader boards).

I'd clean this up, but one of my kids just gave the other a bloody nose while roughhousing, and I've got other things to clean up....

[C E Dwyer]

I think you missed the Hunchback IIC off your list, it clearly stands out as the best Medium mech, by far. High mounts vicious damage output, with it's Uac10 dakka build if it's not focused down fast, a concentrated splat with the srm build.

The triple Uac5 build, the one I think is its 'weakest' is the same build that on the Illy phract with it's vulnerable XL engine people used to scream OMG its OP P2W a few years ago, in much better hard points.

Edited by Cathy, 16 October 2016 - 04:55 PM.

[Y E O N N E]

Tahawus, on 16 October 2016 - 04:47 PM, said:

But that's not what is being measured in the leader board stats. The locust, arctic cheetah, kodiak... all dish out disproportionately high amounts of damage, though proportionately to a baseline adjusted based on their tonnage. Looking back at the model, the model's slope is not particularly steep with respect to tons (i.e. the intercept term is comparatively large compared to the slope's marginal value.)


A more theoretical discussion.

Let's assume that there should be parallel functions, tricky though they might be to formulate, expressing the value of other roles played by mechs other than just dealing damage such as tanking, scouting, harassing and similar values that are poorly represented in the current metric. The sum of those values should be near constant to establish mech to mech parity across all mechs and across roles.

The damage function should look something a lot like my my modeled response. A general increase in damage output per ton, with variations up and down based on the role that the mech is configured for (i.e. a light harasser should be able to do more damage than a mech of the same weight configured for scouting).

So what we're really missing is an effective treatment in the game of information warfare and a way to capture the value of harassment (other than damage).

To sum up, I approached the analysis with a limited scope, trying to idealize the the damage function while not adequately representing the failure of the game to represent (and reward) the other roles that mechs should be playing. I still believe that my efforts provide a good and valuable quantitative indication of what the expected damage a mech should be doing (as measured in the leader boards) and provides valuable insight into which mechs over and under perform compared to the rest of the mechs (again based on the leader boards).

I'd clean this up, but one of my kids just gave the other a bloody nose while roughhousing, and I've got other things to clean up....

Also of worthwhile note, your data is collected in Quick Play. If you were to compile the data from MWOWC, you would very likely find that damage does go up with tonnage. Just from the OCE matches I saw today, that seemed to be the case. Lights hung out around 200 damage, Mediums up to around 450, Heavies to around 500 (a lot of variation here due to a lot of 'Mech variety), and Assaults doing 500+ if they didn't get focused down early. There seems to be a damage cap around 500 or so if one 'Mech isn't allowed to dominate, though I suspect this also has a lot to do with it being 8 v 8 instead of 12 v 12.

Edited by Yeonne Greene, 16 October 2016 - 04:56 PM.

[Pariah Devalis]

Yeonne Greene, on 16 October 2016 - 04:56 PM, said:

Also of worthwhile note, your data is collected in Quick Play. If you were to compile the data from MWOWC, you would very likely find that damage does go up with tonnage. Just from the OCE matches I saw today, that seemed to be the case. Lights hung out around 200 damage, Mediums up to around 450, Heavies to around 500 (a lot of variation here due to a lot of 'Mech variety), and Assaults doing 500+ if they didn't get focused down early. There seems to be a damage cap around 500 or so if one 'Mech isn't allowed to dominate, though I suspect this also has a lot to do with it being 8 v 8 instead of 12 v 12.

If you watch the damage numbers even in 12v12 public queue, you'll find that if your teammates are reasonably well skilled and carry their own weight, that same 500 damage cap appears. A band between 300-500 damage, spread among the teammates fairly evenly usually indicates pretty solid and equal performance across the entire team. People doing excessively high damage beyond 500 usually indicates either they had to compensate for severe under-performers, or they themselves were extremely inefficient at outputting damage in a directly lethal way.

[Y E O N N E]

Pariah Devalis, on 16 October 2016 - 05:08 PM, said:

If you watch the damage numbers even in 12v12 public queue, you'll find that if your teammates are reasonably well skilled and carry their own weight, that same 500 damage cap appears. A band between 300-500 damage, spread among the teammates fairly evenly usually indicates pretty solid and equal performance across the entire team. People doing excessively high damage beyond 500 usually indicates either they had to compensate for severe under-performers, or they themselves were extremely inefficient at outputting damage in a directly lethal way.

Sounds about right, but that's still a QP value, that's not competitive. Competitive, Lights spend most of the match gathering intel, capping, or making feints. They don't really expose much until the late phase unless somebody messes up, because to do so has a high cost.

Really, the best way to describe it is that your damage is capped by your armor. If you have more armor, you can afford to expose more. More exposure means more trades means more damage. Ergo, you get that damage-versus-tonnage slope that Tahawus is looking for.

[Pariah Devalis]

Yeonne Greene, on 16 October 2016 - 05:13 PM, said:

Sounds about right, but that's still a QP value, that's not competitive. Competitive, Lights spend most of the match gathering intel, capping, or making feints. They don't really expose much until the late phase unless somebody messes up, because to do so has a high cost.

Really, the best way to describe it is that your damage is capped by your armor. If you have more armor, you can afford to expose more. More exposure means more trades means more damage. Ergo, you get that damage-versus-tonnage slope that Tahawus is looking for.

Absolutely. Still doesn't keep the "500 damage cap" from being an interesting phenomenon. The 300-500 damage band even mostly fits into the comp scenario, too. Come to think of it, it would be interesting to see what the "average damage per player" is on a winning team. Over a large sample size, that is.

[Y E O N N E]

Pariah Devalis, on 16 October 2016 - 05:17 PM, said:

Absolutely. Still doesn't keep the "500 damage cap" from being an interesting phenomenon. The 300-500 damage band even mostly fits into the comp scenario, too. Come to think of it, it would be interesting to see what the "average damage per player" is on a winning team. Over a large sample size, that is.

It's interesting only because I would expect the Assaults to out-damage the heavies, but they rarely do. The reason is because the Assaults get focused down, which mitigates the advantage their armor is supposed to give them over Heavies, thus capping their output at around the same point.

E: and it doesn't quite fit into the comp scenario. At least, not without chassis restrictions.

Edited by Yeonne Greene, 16 October 2016 - 05:23 PM.

[Pariah Devalis]

Yeonne Greene, on 16 October 2016 - 05:21 PM, said:

E: and it doesn't quite fit into the comp scenario. At least, not without chassis restrictions.

Well, no, not in the per chassis sense. I'm talking more in the general sense. There is only so much damage to go around, basically. If everyone puts forth maximum effort, no one person should be doing much better than 500 damage, and no one person should be doing much worse than 300 damage. In this, there are serious similarities between comp and public team damage values. Presumably related to damage saturation. There is only so much damage an enemy team can sustain before a domino effect.

[Y E O N N E]

Pariah Devalis, on 16 October 2016 - 05:25 PM, said:

Well, no, not in the per chassis sense. I'm talking more in the general sense. There is only so much damage to go around, basically. If everyone puts forth maximum effort, no one person should be doing much better than 500 damage, and no one person should be doing much worse than 300 damage. In this, there are serious similarities between comp and public team damage values. Presumably related to damage saturation. There is only so much damage an enemy team can sustain before a domino effect.

Well, yeah, I get that. I thought that was just understood.

I'm just trying to point out that what Tahawus is expecting to happen with damage over tonnage generally holds true when you remove pilot skill as the factor.

[Novakaine]

Clan mechs.

[WrathOfDeadguy]

I get scared when I see an ACH... when I'm also playing a light 'Mech. It isn't "fetch me my brown pants!" terrifying like a surprise streakcrow or streakdog is, but I know I'm better off running than fighting. In a medium or heavy? Only worrisome if I'm alone and it can circle me. If I'm getting circled in a medium, I'm doing something unforgivably wrong and deserve to die messily for it. If I'm getting circled in a heavy, then either everyone else is dead and I'm probably done for anyway, or I let myself get isolated and it's my own damn fault. Otherwise, it's just another striker light; dangerous if ignored or wolfpacked, but chased away easily enough if it's alone.

I get scared when I see a LCT if and only if I know the player driving it is scary good with it. Otherwise, it's a pest of about the same level of threat as the mosquito buzzing around my desk. Can it be very irritating and distracting, and possibly get me killed? Sure. But if I let it do that, it's entirely my own damn fault. Sure, the LCT is in a better place now than it's ever been, and is enjoying a surge of popularity... but the poor bug's a sideways glance from death at any given moment, and most of my configs even down to my light 'Mechs have enough firepower to do the job in one solid hit. I'm not seeing any balance issue here.

Warhammer... no, not really. It's no more devastating than any other 70-tonner, which is to say that I'm not going to ignore it (it is a heavy after all), but at the same time the Night Gyr or Timberwolf standing next to it is a way bigger threat. It's a good 'Mech, but it doesn't even get close to OP. It's gotten better since PPFLD climbed back on top, and it was never bad to begin with thanks to good quirks and hitboxes, but still... not the boogeyman.

I'm with everyone else who'd rather point their fingers at the Bear. That's a 'Mech to be genuinely concerned about. KDK has the tonnage, the hardpoints, the armor, and enough speed to put the most damage where it will do the most harm- it's a perfect storm 'Mech in the same way that the TBR and SCR (and to a lesser extent the ACH) were when they first came out, but for the assault class this time. Games are won or lost because one team had more of that specific chassis than the OpFor... something that can't really be said about any other 'Mech right now.

[Requiemking]

WrathOfDeadguy, on 16 October 2016 - 10:31 PM, said:

I'm with everyone else who'd rather point their fingers at the Bear. That's a 'Mech to be genuinely concerned about. KDK has the tonnage, the hardpoints, the armor, and enough speed to put the most damage where it will do the most harm- it's a perfect storm 'Mech in the same way that the TBR and SCR (and to a lesser extent the ACH) were when they first came out, but for the assault class this time. Games are won or lost because one team had more of that specific chassis than the OpFor... something that can't really be said about any other 'Mech right now.

Except, that really only applies to the 3, which is severely OP. The 1 and 5 are out-done as Laserboats by the Banshee and Battlemaster, the 2 is out-done as a jumping Assault by the Victor, the 4 is bested by the Direwolf and King Crab as an all big guns Assault, and the Spirit Bear is out-done as a Brawler by the Atlas.

Edited by Requiemking, 16 October 2016 - 10:47 PM.

[Duke Nedo]

Yeonne Greene, on 16 October 2016 - 05:28 PM, said:

Well, yeah, I get that. I thought that was just understood.

I'm just trying to point out that what Tahawus is expecting to happen with damage over tonnage generally holds true when you remove pilot skill as the factor.

Wasn't there also some other graph that had an optimum at ~70 tons? I PUG play I would imagine that to be true on average since you need some speed to not be left behind and good agility should generally > armor in disorganized play.

That's another thing to consider about the statistical analysis, it's based on top scores where only the top 10 games out of an undefined number of total games count, right? That would make it a measure of mech potential rather than mech power, thereby over estimating the score of "feast-or-famine" mechs over reliable performers. I am guessing that is why you see more lights, medium bombers than you would expect in the top, for example Locust, Nova, Stormcrow, Cicada, Viper etc. For the same reason I'd expect brawlers to be under estimated since they are much more likely to always take return fire when they deal their damage thereby reducing their feast potential.

[The6thMessenger]

I wish the worst for all Locusts -- get legged early in the game, and die last and lose the game. But i wouldn't call them op, just annoying as one mech can ever be.

[Duke Nedo]

My top-3 list would be:

1. Kodiak-3
2. Arctic cheetah
3. None

As for the rest, I feel IS has again lagged a bit behind due to the last round of rescale and structure quirk normalization per tonnage that double-nerfed some chassi. IS mechs could use a minor buff across the line to put the best ones closer to the performance of the best clan bots NTG, TBR, EBJ, HBK-IIc, SCR, ACH, JNR-IIc, KDK, etc, and the worst ones closer to the worst clan mechs. As it is now most of the best mechs are clan and most of the worst mechs are IS, I am sure everyone would agree.

[DrxAbstract]

Requiemking, on 16 October 2016 - 10:47 PM, said:

King Crab as an all big guns Assault

KGC's claim to fame is 4xUAC5 despite the initial showcasing of 2xAC20... i.e. Dakkacrab, e.g. Not big guns.

There are virtually no other builds as universally effective as that for a KGC.

[The6thMessenger]

DrxAbstract, on 17 October 2016 - 12:28 AM, said:

KGC's claim to fame is 4xUAC5 despite the initial showcasing of 2xAC20... i.e. Dakkacrab, e.g. Not big guns.

There are virtually no other builds as universally effective as that for a KGC.

I had moar success with 2x Gauss + 2x PPC though.

[kesmai]

This thread is an example of a flawed psr.


(come on that was a nice diversion, wasn't it?)