Arcanaville

[ QUOTE ]
in sumation you may have confirmed the formula used for defiace, but have not convinced me at least that defiance actually works. i still believe and will continue to lobby for defiance to be changed.

[/ QUOTE ]

It was not my intention to convince anyone of that, thus the title of the thread.

What I know is that defiance can be *made* to work by an expert player: namely me. But that's actually totally irrelevant to the question of whether or not its a reasonable inherent for the blaster archetype, because blasters aren't specifically designed for me, and even if they were, just because I can make it work, doesn't mean its designed properly *to* work. But that's a different issue from discussing what it does mathematically, and what its in-game effects are statistically.


The information is useful to both sides of the issue. If you like defiance, this tells you exactly what it does. If you don't like defiance, this at least tells you exactly what it does. If you don't like it, and include in your reasons for not liking it literal falsehoods, like "it offers no benefit at all until you are almost dead, and I know because I have watched my combat chat and can prove it" your opinions can be easily - and justifiably - dismissed by players and especially the developers alike. In the absence of facts, people tend to exaggerate, and the devs know it, and probably filter it. At least now the question of whether defiance is a good thing or not can be based on reproducible facts, not mythology.

[ QUOTE ]
[ QUOTE ]
This, as has been mentioned before, is the inverse of what was requested and stated by the majority of the player base. That the early and mid game was fine, it was the late game that the Blaster Archtype needed a boost in.

[/ QUOTE ]

If Arcannaville was right about what factors the developers look at when decideing what to change, it doesn't matter what we, the players, asked for. I am not criticizing what you wrote Red, especially since I happen to agree with your post in toto, but throwing this out as a little side point to stay on topic.

[/ QUOTE ]

This is almost certainly how the devs process what we ask for, and to preface this I'll say that in principle I actually agree 100% with the approach:

The players ask for something, like say "blasters need more damage." The devs hear that, and ask themselves "why are blasters asking for more damage: does it appear to us that they need an offensive boost of some kind?"

Then they datamine, or whatever they do, to determine that for themselves. If they see, upon closer review, that there is a deficiency, they then ask themselves "based on the deficiency we see what's the best way to rectify it?" That might be the original player suggestion, or it might not. The problem intrinsic in the suggestion might not be the problem the devs see.

The devs might see something different. They might look at statistics and say "blasters are putting out a lot of damage relative to other ATs, so a damage boost doesn't seem waranted. But they seem to die more often; maybe we should address that." If they boost damage across the board, they increase survivability, but also increase damage output in places other than where the blaster was actually in jeopardy of dying. So they brainstorm the idea of *only* boosting damage when the blaster is in trouble. Thus: desparation/defiance.


*Intrinsicly* there's nothing wrong with that mode of thought. My main problem with it is in its actual implementation. I don't think the devs are looking at the right numbers when they decide what's wrong, and I don't think they always have a firm idea of what their changes are supposed to do on a specific level. If the only design goal for defiance is "boost damage when you're about to die" well it certainly does that. But I would want a much stricter design goal for the ability myself. Something like "the kill speed increase due to defiance exceeds the health regeneration speed penalty by at least 20%" or something extremely specific.

[ QUOTE ]
Also the above equations probably doesn't accurately describe what happens when your health is between 0 and 1 percent. (Although I will note the poster did set a boundary input value of 1, AND both will cap at 400 I'm sure so... no biggie).

[/ QUOTE ]

Actually, both expressions are well defined for the full range of 0% to 100% health. X ^ 0 is defined to be 1, and X ^ N where N is fractional is also well defined: for example, X ^ (0.5) is the square root of X (at least for the case where X > 0).

[ QUOTE ]
Simplifying the above expressions yields:

D = 565.7% / (1.072 ^ H)
T = 28.28% / (1.072 ^ H)

[/ QUOTE ]

Actually, if you count keypresses, those aren't that much shorter, although I know what you meant. The equations I gave have two advantages over reduced ones:

1. They actually are what the game calculates, so they are exact.

2. They give a hint as to how they were designed. Clearly, the damage buff was "centered" around having a 25% damage buff at 45% health. Slightly less obvious, it was designed to hit the blaster damage buff cap (+400%) at 5% health, which was probably considered "just about the bottom."

The tohit buff is one twentieth of the damage buff. Its possible that the thinking there was that when the buff reaches the blaster damage cap of +400%, the tohit buff should hit the tohit ceiling for an even minion (where tohit is normally calibrated), which would be +20% (75% + 20% = 95%).

At least, that is my best guess as to the basis for the design of the numbers.

[ QUOTE ]
[ QUOTE ]
Even assuming defiance is working as intended (which it's not, I constantly have a persistent 10 on my defiance bar in Talos even if I'm at full health, screenshots have been posted before showing this)

[/ QUOTE ]

Didn't the OP show that the BAR was bugged, but Defiance was actually kicking in immediately, regardless of what the bar might currently be displaying?

[/ QUOTE ]

Under test conditions, yes. Its obviously much more difficult to test this under realistically complex lagged conditions, although lag is usually going to affect client visual elements more than server-side effects anyway.

[ QUOTE ]
I submit that any strategy which can be consistently used to increase your rate of xp gain isn't really worth a reward, no matter how complex. What I consider worthy of increased xp gain is playing on the edge and using your abilities to the limit. A pre I5 tank herding was practicing a known and complex skill, but by no means pushing the limits.

[/ QUOTE ]

Separate from the specific issue of how this thought might be applied to defiance, I do not agree with the general principle. For example, there are several ways to attack a group with single target attacks: fire every attack as it recharges, or fire the most efficient attack, which sometimes requires waiting for the right attack to recharge (you can also spread attacks among many targets in a pattern which allows you to fire most attacks as they become available).

I would say someone that uses more skill in deploying their attacks will gain XP faster on average than someone who doesn't, and that increase in XP speed is reasonable for that use of skill. There are lots of other examples where XP rate can be consistently increased, with no measurable increase in risk of death, where the commensurate reward increase is nevertheless appropriate.

In fact, I'm not sure its even laudable to attempt to control this form of skill reward.


So I think that the main issue I have with your statement is that "and" should really be "or."

"What I consider worthy of increased xp gain is playing on the edge OR using your abilities to the limit." Using your abilities to the limit doesn't necessarily presume "playing on the edge." It generally does, in the specific case of defiance, but it doesn't have to.

[ QUOTE ]
But isn't that three more brain cells than is needed to turn on your defense toggles and click Elude when you want to own?

[/ QUOTE ]

The blue bar under the green bar. Most of us have to watch that one, too.

[ QUOTE ]
Wow. You stated in one sentence something I took around 3 pages of verbiage to express...

[/ QUOTE ]

I'm sorry. I'll try not to let it happen again.

[ QUOTE ]
Arcanaville, you are getting very close to my Complexity vs Reward Rate balance paradigm.

[/ QUOTE ]

Actually, its as simple as (within certain limits) I'm in favor of rewarding skill.

[ QUOTE ]
Like on your herding list. One should be able to agro unlimited amounts of mobs. Because to tell you the truth I find it really stupid for me to fight a mob and 5 of their friends are just sitting there doing nothing. The only thing agro cap created is that I wont be getting hurt as much because the agro cap simple prevents the extra mobs from attacking me then and there. The moment their buddies drop they all jump into the fire and die. It actually helps the survivability of a herder. One should also be able to hit as many targets as those in range and line of sight as a given power indicate. The problem was collition checks. If mobs had good collition checks then they wouldn't stack. That would automatically prevent said AoE from hitting 100 targets because only a certain few would be in range and or in line of sight.

[/ QUOTE ]

My solution to this problem is to do what a lot of mobs already do: every mob aggroed after #17 refuses to enter melee range, and fires at range from some random distance greater than five feet. That way, aggro is partially self-regulating. Tankers can grab all the aggro they want, but beyond a certain point, the benefit is all defensive (they stop shooting at your team mates) and not offensive (they do not all stand on a bullseye and say "shoot me").

[ QUOTE ]
I disagree. I think your idea ignores the possibility that the maximum potential of one powerset might be harder to obtain than the maximum potential of another powerset.

[/ QUOTE ]

I don't ignore that. I just wouldn't balance the powers themselves around that directly under ordinary circumstances. And if I decided it *was* necessary, I would still develop a quantitative model to do it.

Why? Because I am not of the design school that says "throw it all up on the wall and see what sticks." I am of the school that says "you won't always do the right thing, but you should always at least know what the heck you're doing." So suppose I design set A to be harder to use than set B. How *much* harder did I do that? Hard to say. But what I'm *supposed* to know is how hard did I *intend* to make set A relative to set B. If I *intended* set A to be twice as hard to play as set B, I can decide how that should factor into the strength of set A, and set B.

I might be wrong, though, and datamining might tell me my judgement of difficulty is in error. There's two solutions: accept that even though I wanted set A to be twice as difficult as set B, since its actually three times harder, just go with that and tweak the power set so it makes sense when set A is three times harder. Or, say "well, I didn't hit my target of twice as hard, lets change the difficulty of the set so I actually hit the target I set for myself."

I'm of the school that demands the second option is the *only* option worthy of being put on a resume.


The best way to describe this philosophy is to say its goal-oriented. You set a goal for the design, and you try to hit it; you don't design in the blind, and change your goals to try to represent it.

If I want set A to be 30% harder, and I decide to give it 10% more performance for that 30% harder, I expect to hit those marks. And if I miss one, I fix it: I don't revolve everything else around the miss.


[ QUOTE ]
Now I know you disagree, but I think it's perfectly appropriate that Regen has more potential protective ability than SR since you have to decide when to use it.

[/ QUOTE ]

You don't know that, because I don't disagree. In fact, you'll see statements precisely to that fact for both Regen and DA in my scrapper secondary comparisons, with specific warnings about making a difference between knowing what the performance actually is (which the analysis shows), and what the desired performance actually should be, which is a separate issue.

The question, though, is how much better. For the cost of having to have three brain cells wired together in the right way to click a button, regen gets about twice the performance under ordinary conditions. For most players, those three brain cells are probably not a very high cost relative to the return, provided the average player doesn't play extremely intoxicated.

Actually, on my server, that's not entirely a safe assumption, but I assume it nonetheless.

[ QUOTE ]
So how do you 'eliminate the circumstance'? I don't see how that's materially different that 'balancing'.

[/ QUOTE ]

You need to separate balancing powers and effects, and balancing tactical mechanics. When the devs decided that travel power jousting was allowing things like blasters to fight with a lot more safety than they desired, did they:

a) Give all blasters a resistance debuff to make them take more damage.

or

b) Institute travel power suppression.

I make a distinction between designing the powers, and managing playstyle mechanics. If you want to manage playstyle, you can predict a lot of it if you are smart enough, but some things just require watching your players, and seeing what they come up with. But that is ordinarily a completely separate problem from powers strengths.

If the devs suddenly discovered that all scrappers were using taunt to single pull, and that was causing them to fight in total safety, would you consider that a case of needing to alter pulling mechanics, or reducing the strength of their defenses to make single pulling more of a challenge? I'm of the opinion that 99.9% of the time, its the former, not the latter.

And that is why datamining for performance is, in my opinion, a bad idea. Data mining behaviors to find out behaviors makes sense. Data mining performance intertwines two separate components of performance - power strength and player tactics - that should never be balanced together, almost always addressed separately.

[ QUOTE ]
This is where math fails you Arcana. Kiting as druids did(do) it in EQ is not something that simply looking at the druid class you could determine by plugging it into a formula.

[/ QUOTE ]

I didn't say it did. What I said was that no matter what it did, it couldn't have allowed you to do more than what your powers theoretically allow you to do. The whole point to balancing around potential instead of actual performance is that I really don't care if the design happens to create an unforseen and unpredictable circumstance like that described, because if the unforseen circumstance is tolerable, the sets don't need to change, and if its intolerable, the circumstance gets eliminated. But in no case is it "balanced" around: that's the whole point.

There's no way to look at a bunch of numbers and equations and predict, lets say, superjump jousting in PvP. But it *is* eminantly predictable that regeneration scrappers and stalkers would try hard to create a jousting strategy, because the mechanics of regeneration so highly favor it. That might have been useful information to a game designer, if there was someone around willing and able to tell them that, and they were willing to listen.

There are limits to what the mathematical models can do. But I don't think most people are informed about what those limits are. They presume there is a limit, and assume its somewhere near what they themselves believe they are capable of. Its actually probably several orders of magnitude higher than you are estimating.

To the extent that mathematical models might fail to take certain things into account, in this particular game those things aren't balanced anyway. I cannot think of one thing *actually* balanced in this game, that could not have been with reasonable mathematical design.

[ QUOTE ]
So you tell me, how do you account for that? How do you account for a situation where a number of variables, none of which are problematic in isolation, combine to create a problematic situation?

[/ QUOTE ]

Exactly as I previously described.

[ QUOTE ]
Woo hoo! Castle actually read something I wrote!

And I'm the first to post after it!

On my birthday!

<does the Cabbage Patch>

[/ QUOTE ]

Possibly the most joy I've ever seen in a thread I've started.

[ QUOTE ]
The problems with mathematic models, though, is that they rarely take into consideration a number of variables that are very ephemeral and difficult to quantify. The aforementioned "quad-kiting" and "charm-kiting" are great examples -- both are player behaviors that were never accounted for in any of the math models used to create the game. All it takes is one aberrant behavior to throw a seemingly well-balanced system out of whack. For our game, adding in external effects like Temp Powers and Inspirations and things really jump in complexity.

[/ QUOTE ]

This brings up a philosophical question: should power sets be balanced around what they *can* do, or what the players *make* them do. If no one takes RPD, does that mean Invuln is weaker than it should be? If Invuln is boosted because of that, and *then* players start taking RPD above the boost, is Invuln *now* too strong, and has to be nerfed?

Do you balance sets factoring *in* player skill, or factoring *out* player skill? In my opinion, you balance factoring *out* player behavior in general, and change things only when by some freak occurance you design a set that has such a skill-significant issue, that skill doesn't become a reasonable reward unto itself (you do better because you are better), but becomes out of whack to skill level (learn one little thing, and suddenly you're twenty times better).

Otherwise, you're actually saying, in essence, the better you (the players) play your sets, the worse we're going to make them. We're penalizing good play. Contrawise, suck, and we'll boost to compensate. I'm sure there's an argument to be made that fundamentally, this is a good idea, but there isn't enough alcohol in the universe to convince me of that.


Now, I'm not an EQ player, but what I can say is that its not true that "math there didn't tell the entire story." Whatever you can do with weird kiting strategies is no more than the maximum you are calculated as being able to do. *Most* people do *less*, but no amount of kiting can allow you to do more. Which, if you balance around the philosophy of balancing around potential, and not actual performance, is not a problem.

That might mean the kiting itself is a problem, but that is exactly what datamining is good at: finding trends in player behavior. It says nothing, though, about what you should do about it, or even *if* you should do anything about it. And when player behavior allows a group to gain better leverage over the game environment than another, and its inextricable from the game, then it just becomes another factor in the mathematical balancing models.

If, hypothetically, I had a power that was designed in a weird way that allowed anyone with it to kite a group of critters in a way that prevented them from ever shooting back, ever, then in fact the damage mitigation potential of that power is 100%. Mathematical models can handle that. If it requires tremendous skill to pull off consistently, that power's effective damage mitigation potential can be said to be some fraction of 100%, based on average player skill. Mathematical models can handle that also. If the difference between the best players and the worst players is too high given the constraints of the game, such that no ratio properly represents the average effect of that power very well, then the mathematical models are screaming that that power shouldn't even be in the game at all. But in no case is that power beyond the ability for a mathematical balancing model to handle.


On the subject of time: we both know why there's often not enough time to construct good mathematical balancing models: its because mathematical skill isn't highly valued in the game design industry, and mathematical work isn't highly valued in game implementation projects.


On the subject of being berated about mathematical models, tell Weirdbeard I said hi.

[ QUOTE ]
[ QUOTE ]

Arcana, if I am reading between your lines correctly are you saying that much of the game engine is/was designed primarily by trial and error and not by creating/using a mathematical model to represent some particular plan or behavior?


[/ QUOTE ]

This is a game where, (I hear- wasn't there), Tanks started out able to reach 100% damage resistance and they didn't realize it would be a problem.

Math came to this game somewhat late.

[/ QUOTE ]

This would be just one among many w-t-f moments in the game design and power balancing efforts. But that wasn't because Invuln's numbers were set randomly: more critically they were set according to a system that judged the raw numbers, and not the survivability those numbers represented.

This is evident from remembering descriptions of the situation when scrappers were set to "75%" of tankers, by instituting the 0.075 AT modifier for scrappers, essentially making scrapper invuln 0.75 times tanker invuln. This was described as making scrappers "75% as strong as tankers." This is a mathematical distortion that is impossible to overstate. And my guess is that it comes down to Geko being good with numbers, but not good with math. This cropped up again in discussions about Kheldians, and Kheldian health: the effects of +health get radically misstated by the devs there.

I think an underappreciated fact is that the devs don't ignore math when they design things, but they often misapply it. This is important, because many players think math "doesn't tell you the whole story." Those people would be surprised, I think, to learn just how the game *is* designed and balanced by the devs. Its not balanced by "feel" its balanced by numbers: but not in a consistently logical way.

Consider: in I6, thunder kick was lowered from 2.78 BI to 2.33 BI. Was this because it "felt" too powerful? Was this because the set was datamined to be too damaging? No. It was set to 2.33 BI because thats what their attack formula said it was supposed to be, given its recharge and all this time it was set wrong. That's it. The change was made to adhere to a mathematical rule, period.

I keep this, and many other things, in mind whenever someone basically tells me "math's great, but math isn't the final answer." How little they know how often math is in fact the final answer. Just not always in a good way. The problem isn't that the devs don't listen to math, its that they only listen to their math, most of the time. The challenge to me, and others with similar goals, is to make my math as compatible with their math as possible, because they aren't taking a bunch of math classes to hop the fence over to me.

[ QUOTE ]
Does Defiance do what is advertised in the perception of the gamer? From the discussions I think not. Maybe our perceptions and preconceptions should be handled better by NCSoft.

[/ QUOTE ]

On that note I've suggested directly to Castle that defiance damage have its own set of floating numbers, specifically to improve defiance's "image." Which I understand would be difficult, because unlike criticals, which are a separately calculated damage component, defiance is a damage buff, like assault, and doesn't have an easy way for the game engine to tease back out what it was.

But perhaps defiance should be critical or scourge-like anyway: that way it would be easier to represent, and it would be bonus damage that was separate from the damage cap. After all, why should a highly buffed blaster's damage not benefit from defiance if that blaster is at low health and in trouble: isn't he just as much in trouble whether damage buffed or not?

[ QUOTE ]
While it is possible to build and design a game whose balance is based on such even multiples and fractions, it has never seemed remotely likely that CoH actually was. The performance discrepancies in things like DR vs Defense vs. Regen scream for less "whole" numbers to crop up somewhere. Real mathematical modeling and its attendant balancing would, in a system as complex as CoH, produce a "fudge factor" that wasn't a nice number like "2"

[/ QUOTE ]

Its not possible to balance exactly with whole numbers. But because even the definition of "balance exactly" is vague, its possible to get close enough with reasonable numbers.

Ever since the concept of the AT scalers was made generally known, I've tended to make balancing suggestions that take that into account, suggesting numbers that are relatively "simple" when expressing their base values. Its possible to do, and even the devs have tweaked things to have moderately "odd" numbers: the SR toggle defenses have a base intrinsic value of 1.85 (which when multiplied by the scrapper defense AT scaler of 0.075 gives the actual defense value of 0.13875, or 13.875%).

There are other reasons for me thinking that the balancing mathematics, such as they exist, are totally borked. Castle gets an earful at least monthly.

[ QUOTE ]
Before you start on Vigilance, if you really wanted to start something, do an analysis of which of the Defender primaries offer the most protection to a team.

[/ QUOTE ]

I've thought about it. I'm still working, as time permits, on looking at blaster secondary effects, which is a very complex problem. Also, working on badge #374 and #375 has been taking up a bit of time the last couple weeks.

[ QUOTE ]
When CoH came out it was City of Blasters and that had to be adjusted. We don't want to see that skewed dynamic again.

[/ QUOTE ]

Historical note: the very first "numbers post" I ever made on the forums was one of my first posts ever: having read the "City of Blasters" discussions that were going on at the time, and not quite seeing that for myself, I conducted census tracking of my server in various zones at various times of day. My conclusion was that blasters at the time made up between 30% and 40% of all logged in characters depending on zone, with the number rising to a peak in the mid 20s, and falling off above 30. My conclusion at the time was that blasters were levelling out of the single digits and teens very fast, which depleted blasters in the low levels slightly, and piled them up in the 20s. But then levelling speed levelled off, and from that point forward they tended towards their true percentage, which was about 30%-35% of all logged in characters. Which is high, in a game with 5 archetypes, but not ridiculously so, given the natural conceptual attraction of playing a blaster when the game was very new.

If I recall correctly, the ones much lower than the expected average were scrappers and (especially) controllers.

[ QUOTE ]
[ QUOTE ]
A comment on datamining. While I can't tell you exactly how and what we look at in datamining, I can say that we don't usually look at 'Defiance' in and of itself. Overall performance matters much more than any single aspect of a character.

[/ QUOTE ]

Doesn't that make it terribly difficult to understand what contribution any given power or aspect provides to overall performance?

[/ QUOTE ]

I believe what Castle would say if he were able to comment is that blasters prior to the addition of defiance were not, as an archetype on average, meeting some specific performance metric they data mine for, whatever that might be. After adding defiance, they appear to, in their own internal test servers, on the test server, and now on live. At least to the extent that they believe defiance was supposed to change things.


The devs, as a rule, use datamining to look at macroscopic things. They then sprinkle changes to powers, inherents, etc, in an attempt to alter those macroscopic things. As I mentioned previously, I do not believe that is a proper way to balance things: you balance based on mathematical models of the ATs, and you use data mining to inform those models.

And as I said previously: you do not use data mining to determine the performance of powers. Its the worst of all possible ways to do that, and for the most part, the devs don't even try to do it, as Castle indicates.


The Rosetta Stone for dev-balancing thought comes from the I3 scrapper tests the devs performed and posted about. Ignoring a lot of the weirdness in those tests, they demonstrated several critical facts about dev-balancing methodology:

1. They don't have good mathematical models of the scrapper secondaries' performance, and by extension probably everything. Otherwise, those tests would have been redundant.

2. They do not actually datamine to determine specifics of performance. There would be no point to conducting those tests if their datamining was even remotely precise. Its clear that what datamining tells them is less whether the car is on the right road, and whether or not it just crashed through a barn.

3. Lacking mathematical models, there's also no mathematical guidence as to what changes to make if there *is* a perceived problem.


By the way, the I3 scrapper tests had an interesting consequence: they were part of what inspired me to create the scrapper secondary mathematical models I did in I3 and I4: essentially, I conjectured that I could reproduce the results the devs got through playtesting. The I4 models get most of the way there, the I7 ones basically get all the way there, except for modelling CoF and OG. Essentially, with proper mathematical modelling, the devs could have had their I3 test results before launch day, with all that implies.

[ QUOTE ]
[ QUOTE ]
I have a thought about that, something related to vigilence. I can't quite express it fully yet, but when the thought finally settles I'll post it here.

[/ QUOTE ]

I would really look forward to an analysis of this sort for Vigilance.

[/ QUOTE ]

Vigilence is a little trickier to analyze because of its group dynamic. But what I was thinking about was more of a perspective on defiance and vigilence rather than their actual strength (although that is part of it).

One interesting complaint that is shared to a degree between both inherents is that in effect, they reward bad play. You can only get the damage boost from defiance if you get punched in the face a lot, you can only get a lot of it if you put yourself at significant risk of death (intentionally or not). The same is somewhat true for vigilence: you can only get a boost from it if, in effect, you let your team get beat up.

Both seem to penalize good play. A blaster that protects himself well cannot get much of a boost from defiance. A defender that protects her team well cannot get much of a boost from vigilence.

But I was thinking about how the SR passive resistances work in relation to this. A lot of people don't fully appreciate that those resistances are a lot stronger than they appear, specifically because they (often, not always) get stronger as they are needed most: that acts to skew their net effect higher, because they only apply power where necessary, and not anywhere else.

In a sense, defiance and vigilence can act in a manner similar to those resistances: as "bumpers" that tend to kick players back into normal realms of play. A blaster might *attempt* to play without getting damaged a lot, but defiance acts as a partial safety net for when they fail: and in degrees, the more they fail, the stronger it acts to kick them back, by increasing their kill speed and presumably reducing their incoming damage. Vigilence has a similar negative feedback effect: if you don't protect your team well enough that they start to take damage, your endurance goes up to give you more options to be profligate with your powers to keep them alive.

In effect, both defiance and vigilence are not buffs in the conventional sense, they are really gap reducers for the AT. They reduce the difference between perfect play and imperfect play: they reduce the penalty of errors.

Seen that way, both inherents are actually casual player tools of a sort. Training wheels for the AT, to reduce their difficulty: you can get the same benefit with less skill, and you cannot easily get lots more benefit with more skill.

This doesn't mean skilled players can't figure out ways to leverage both, but their overall intent might not be to boost those ATs, but rather to reduce the *spread* in their performance. And that *is* something that would show up in the datamining the devs probably do.


This says two things to me if true:

1. Defiance isn't actually comparable to something like criticals, and it never was. Criticals are a set-wide boost to scrappers: they make all scrappers better. Defiance and Vigilence aren't: they are learning-curve softeners. They reduce the gap between the best and worst blasters and defenders.

2. This reopens the question of whether or not blasters really do enough damage, because defiance isn't specifically a damage-boosting tool. That's secondary to its actual purpose. An interesting question would be to find out if the introduction of defiance, independent of other factors, improved the net average performance of all blasters, or if it primarily reduced the spread of performance around a similar average, or if the top performance stayed basically the same, the bottom moved up, and the average therefore moved up - but only by moving the worse performers upward, not because the blaster AT itself moved upward in performance overall.

[ QUOTE ]
[ QUOTE ]
Math Modeling

[/ QUOTE ]

I see this and can't help but think,

'Oh, yeah, work it. Derive that function!'

[/ QUOTE ]

Project: N-way. Next on Bravo.

[ QUOTE ]
Excellent work. So, yeah, Defiance is working according to its design. That's fine. But that doesn't make me loathe it any less.

[/ QUOTE ]

I have a thought about that, something related to vigilence. I can't quite express it fully yet, but when the thought finally settles I'll post it here.

[ QUOTE ]
You may think it "an abuse of data mining" to use it to determine actual performance, but I think you'd be fighting an uphill battle on that front with the Developers. We know from _Castle_'s posts that the Developers use datamining to determine quite a bit.

[/ QUOTE ]

More than you realize.


[ QUOTE ]
The problem with a straight numerical analysis is that it doesn't tell us anything, and all conclusions are pure speculation, pure theory.

[/ QUOTE ]

It tells me a lot. Numerical analysis is not pure speculation. For example, my scrapper analyses come with hundreds of hours of playtesting their conclusions, under a wide variety of circumstances. They make predictions that are testable. My analysis of the SR passive resistances in particular, make non-intuitive predictions of their behavior that can be seen if tested for. My survivability analysis rely on certain unbreakable constraints on performance the game adheres to. And my discrete calculator allows me to see precisely what the genuine effects of many conjectured effects are, such as the efficiency of dark regeneration is relative to its survival power.

In any case, the debate of whether numerical analysis is actually effective or not is immaterial in this case, because no number in CoH is actually *set* by data mining. They are all either computed, or arbitrarily set by a designer. Data mining might suggest to a designer that a number is too high, or too low, but it doesn't give any guidence as to what it actually ought to be. And when there is a total absence of any numerical understanding of what is being balanced, and you attempt to replace understanding with data mining, you end up flailing around randomly. cf: Regeneration. Its never a good circumstance to be in.