General class balance observations

"

ScrotieMcB wrote:

"

RagnarokChu wrote:

She can statistically be balanced but broken in the fact that majority of popular builds is best with her.

Although I'm not really sold on the idea that she can be statistically balanced but still imbalanced in practice - which is basically saying statistics are shit at measuring such things, when I think the right statistics could - it's really a moot point. She's not statistically balanced at the moment, or even close to it.

Popularity can be proportional to accessibility, not necessarily power.

I find it interesting that none of the INT classes except Witch have Energy Shield node starts. ALL the other classes have life starts, Templar seems confused and Shadow has neither :/
Life nodes seem far too prevalent in class starts.

"

ScrotieMcB wrote:

"

RagnarokChu wrote:

She can statistically be balanced but broken in the fact that majority of popular builds is best with her.

Although I'm not really sold on the idea that she can be statistically balanced but still imbalanced in practice - which is basically saying statistics are shit at measuring such things, when I think the right statistics could - it's really a moot point. She's not statistically balanced at the moment, or even close to it.

Statistically balanced would mean.

Let's say balance wise

5% life node = 12% damage node

Scion might have 5% life nodes around her start.

Shadow has 12% damage nodes around his start.

If they have 10 of each they might be statistically BALANCED to one another, with equal amounts of allotted power to one another, but not equal in play 50/50 due to other factors such as builds and so on. Scion would be more popular because stacking life is good, or Shadow might be popular if you had a legacy kaom's heart to make up the HP difference or leads to easy access to other HP clusers + other reasons in that area.


Statistics are just the half of it, you still need to Analysis it afterwards.

Another example would be poison arrow being ridiculously popular, even if you nerf scion people move to the 2nd best class to play poison arrow. Moving the population there thus skewing the population again, thus would it require a nerf to poison arrow or the skill tree?

Templar was beyond popular because of searing bond builds, they haven't really changed outside of that since searing bond isn't FOTM anymore.

this kind of 'analysis' is.. well..

im pretty sure that you can find significant correlation between nickname length and ladder position. given small enough sample and strong enough will one can 'prove' anything and everything using 'statistics' in the most broad and amateur form

also, im pretty sure that players with experience are the ones entitled to draw conclusions and provide explanations. and experience means several high-level chars from several different 'branches'


arm-chair generals spawn in each game given enough time and are at least +20 to fun when encountered, sadly they also mean 50% less experience earned

"

The additional data cast some serious doubt on pretty much every opinion I held in my "nitpick" section of the original OP, with the possible exception of Shadows being under powered (which still has some support from the data, but surprisingly weak support - I wouldn't be very confident providing large boosts to Shadows at this point).

I'll give you kudos for taking another look at the data and re-thinking your position. I'm going to offer a theory as to why looking at the ladder data is the wrong data set for looking at class balance. #15000 on the ladder is something like level 84 in Tempest and 88 in Warbands. Any build that reaches that level is probably powerful. It's at least more powerful than anything I've ever played. (I'm a horrible alt-a-holic; I keep starting new characters.) I would argue that the difference between #1 and #15000 have much more to do with gear, play style and persistence than class balance.

However, as a whole, any underrepresented class in the whole ladder is probably indicative of something. Duelist, Shadow & Templar are all played at a much lower percentage than the ~14% you might expect if all classes were played equally. I could offer several possible explanations:

1 - Hybrid classes are, in fact, weaker than the pure classes & Scion.
2 - There is less build diversity for hybrid classes leading to under representation.
3 - The powerful builds for the hybrid classes aren't as much fun to play as those of other classes.
4 - The leveling process might be faster starting as a pure class than a hybrid.
5 - There are a handful of popular builds being imitated leading to a skewed class distribution.

I'll ask my question again, since you ignored it the first time Scrotie, why does the difference in representation among top 1-5% compared to overall population indicate class power?

The only thing that your numbers show is that as the average character level increases, the class distribution changes.

How are you then concluding that the class that gains the most is overpowered? What is your definition of overpowered, and why does that definition matter?

----------

Also your suggestions like increasing leech for shadow are way too build specific, and hence not useful. Since builds near to each other often visit both starting locations, the only thing that matters are the 5 initial nodes. In 2.0 we are forced to take more defense. As a result people will always choose witch/ranger over shadow, mara/witch over templar, and mara/ranger over duelist. Scion is good for anything that wants to go to shadow/ranger and templar/mara, or duelist and witch. At the highest positions on ladder people legitimately choose one class over the other because it saves them 1 skill point. This makes sense when you consider the grind between L99 and L100.

"

MatrixFactor wrote:

I'll ask my question again, since you ignored it the first time Scrotie, why does the difference in representation among top 1-5% compared to overall population indicate class power?

The only thing that your numbers show is that as the average character level increases, the class distribution changes.

How are you then concluding that the class that gains the most is overpowered? What is your definition of overpowered, and why does that definition matter?

I'll start by not defining overpowered here, because the evidence here is not definitive, as a mathematical proof might be. It does, however, establish a correlation. I can say, as a 100% factual, zero-opinion statement, that if you picked Scion in Warbands, you are over 50% more likely to have finished in the top 4% than if you didn't.

Admittedly, the "why" of that statement is multifaceted. To some extent, more skilled players could have a preference for Scion which causes them to choose it between two equally viable options. This is why sample size is important, because the population becomes less individually eccentric the larger it gets. However, I consider it flat-out unreasonable to attempt to explain a huge gulf, such as at least 25% increased/reduced performance in both leagues (Scion, Templar), entirely with claims of personal preference and randomness. There may be some random idiosyncrasies thrown in, but not enough to explain the stats. When I read Shakespeare, I do not assume it was the product of thousands of monkeys randomly pressing keys on typewriters; a pattern is a pattern.

Still, despite its strength, the evidence isn't a proof. You can tinfoil it away if that's your preference.

"

MatrixFactor wrote:

Also your suggestions like increasing leech for shadow are way too build specific, and hence not useful. Since builds near to each other often visit both starting locations, the only thing that matters are the 5 initial nodes. In 2.0 we are forced to take more defense. As a result people will always choose witch/ranger over shadow, mara/witch over templar, and mara/ranger over duelist. Scion is good for anything that wants to go to shadow/ranger and templar/mara, or duelist and witch. At the highest positions on ladder people legitimately choose one class over the other because it saves them 1 skill point. This makes sense when you consider the grind between L99 and L100.

Pretty good point. Adding or removing even a single node might correct the imbalances I've pointed to.

"

ScrotieMcB wrote:

if you picked Scion in Warbands, you are over 50% more likely to have finished in the top 4% than if you didn't.

We have P(Scion|Top 4%) > P(Scion). You're saying P(Top 4%|Scion) > P(Top 4%).

By Bayes' Theorem P(Top 4%|Scion) = P(Top 4%)P(Scion|Top 4%) / P(Scion). Then since P(Scion|Top 4%)>P(Scion), P(Scion|Top 4%)/P(Scion)>1 and your claim checks out.

Despite being factually correct, it's still kind of a perverse to thing to say that just by picking Scion you're increasing your likelihood to end up in the top 4%.

I mean it's the build that matters and we can probably conclude that scion does builds better, so I guess it makes sense to say she is overpowered.

Good job once again ScrotieMcB.

"

MatrixFactor wrote:

Despite being factually correct, it's still kind of a perverse to thing to say that just by picking Scion you're increasing your likelihood to end up in the top 4%.

It's not only "perverse", it's plain wrong.

The reasons that get you into top 4% are numerous.

Time spent.
Experience in the game, how to build up, what to trade for and when... read as "skill".
Sheer Luck.
Even more time spent.
...

If I had to weight all reasons why someone ended in top4%, I'd put "class choice" below 1%.

"

Peterlerock wrote:

The reasons that get [one] into top 4% are numerous.

I am not saying they aren't. It is worth noting that when I say Warbands Scions are over 50% more likely, I mean it like how the game does increased critical strike chance - a random Warbands Scion had a (just over) 6% chance to be top 4%, a random Tempest Templar a (just under) 2.5% chance. Clearly, there is something weeding out the other 94% of Scions, and of the 97.5% of the Templars weeded out, I figure 96% probably deserved it.

So if you want to get really technical about it, in this case I'd put a class choice of Scion in Warbands at slightly over 2%. Just over double your "under 1%" estimate. But considering that only 4% make it and that extra Scion 2% makes it due to the strength of the class, where another class of equal skill would not make it, well, that's not very fair, is it?

As we've discussed earlier, even a small imbalance, like a single passive skill point, will shift balance drastically, as long as it is clearly superior. You get 101 of them by level 78, so one passive point is under 1%, right? Turns out 2% is a hell of a lot in this situation. Too much, in fact.

Due to the frequent skill tree wipes (high-level) Scions make excellent respec dummies later on. I imagine this is not insignificant.