upcoming rdm changes?

[Gruntler]

you are comparing all percentiles.....it's better to look at 90+

Cringe. No. No no no no no.

For proper balance you look at the spread. You don't pick a percentile outside standard deviation and go 'Here, this is where you go for balance.' You look at how many are above standard deviation. You look at how many are below. You look at what the average is. You look at baddies bringing it down. You look at overperformers getting padded pulling it up. You look at the number of parses.

You. Look. At. It. All. There is no other analysis that makes sense. You don't cherry pick. Balance within a job is NOT exclusively 'the 95th percentile'. You do look at the top end but you also consider how much of the spread is close to that mark--a job where the 95th is 2% DPS away from the median but the 5th is 50% away from the median is a completely different scenario than one where the 95th is 50% away from the median but the 5th is 2% away from the median, as an example.

Picking the 95th percentile is, literally, selecting a single data point, and calling it an analysis. That's literally ONE parse. Show up to a science or statistics class with a single data point and you just failed your lab/analysis. If it's not good enough to get a C in an entry level course, it's not a terribly strong argument.

RDM is partically punished at lower percentiles.

Agreed! And yet, that's a part of balance! If RDM is lagging at low parses, that's a part of balance! The problem is that what we DON'T have, which is actually more important to balance for a job like RDM, is attempts->clear %. If the job has less DPS, but a higher than normal clear%, then it's saving runs.

The fallacy in the 'only 95%' argument is that it thinks that job balance is based around the idea of farm parties, and not around progression or clears; that would be like if Dark Souls devs decided a given weapon was to be nerfed because it was used in world-record speedruns--that'd be rediculous.

The reason the devs know RDM needs help, it's not because of fflogs data--it's because RDM is lagging behind in clears despite being a job without major clunk, after being a highly popular DPS in the previous expansion that wasn't gutted in its kit. I can guarantee, they aren't looking at a single percentile of data, and would be rigorous. If we wish to convince them of balance issues, it behooves us to be rigorous. 'just use 95th' is lazy.

[OneTrueMiqote]

You. Look. At. It. All. There is no other analysis that makes sense. You don't cherry pick.

Jobs should be balanced around the high end because that's when their kits are utilize to the fullest and has the potential to be broken. It's at these high percentiles where the imbalances show to their fullest extent as the element of skill variance is removed. The simple counterexample to your argument is BLM. BLM has always been low at median and on the lower quadrilles, but extremely powerful in the hands of skilled players. Following your logic, BLM should receive a buff.

Another issue in your argument is that you arguing about looking at more than just percentiles. While it's true that jobs shouldn't be based solely on dps percentiles, the post you responded to specifically referred to the context of just percentiles. They did not say you should only balance off percentiles, but rather if you do, you should be balancing off the high end.

Sidenote, I find it ironic how much you ridiculed the idea of "just 95th percentile" being a single point of data and the importance to look at a range of data, when they said "90+", which is a fine amount of data and not a single point.

[Archwizard]

You. Look. At. It. All. There is no other analysis that makes sense. You don't cherry pick. Balance within a job is NOT exclusively 'the 95th percentile'.

So... what, the bottom-end values given by players who used a skip potion and bought the bare-minimum gear to enter content (or recycle gear from other jobs) have an equal weighting for job balance as the top-end players who know the ins-and-outs?

It frankly doesn't make sense to look much lower than the top few percent of parses, since they show the job's maximum potential and can be used as a guideline for effort vs performance. "Use everything" has to account for all levels of skill and gearing, most of which would give the illusion that every job is underperforming because not everyone knows what they're doing.

Mathematically it's difficult to parse out the players who have no idea what they're doing with their job from the ones who have bad gear or the ones who don't keep up with fight mechanics. At least at the top end performers are expected to have a similar level of gearing and skill, both to each other and other parsers of the same content.

[Rilifane]

You. Look. At. It. All. There is no other analysis that makes sense. You don't cherry pick. Balance within a job is NOT exclusively 'the 95th percentile'. You do look at the top end but you also consider how much of the spread is close to that mark--a job where the 95th is 2% DPS away from the median but the 5th is 50% away from the median is a completely different scenario than one where the 95th is 50% away from the median but the 5th is 2% away from the median, as an example.

Balance is important for endgame content and gameplay level.
Because that's where it shows and matters and comparing it to lab science is a poor example. You will hardly feel the difference between jobs in DF content because outside factors have a much larger influence than job balance. You can't do much as a melee if other people keep dropping their aoes in melee range and forcing you away but that isn't a primary concern when it comes to balance, so the upper percentiles are much more interesting and valuable for analysis as most of those external factors are eliminated.

Balance (or the lack of) shows in endgame content and gameplay level because that's where everyone utilizes the whole toolkit to its fullest or close to it plus you can look at synergies between jobs; knowing your class is not optional nor is playing it well when it comes to balance.
The 90+ percentiles is the range where the playing field is as even as it can get in praxis, meaning: the people have BiS gear, play very well and understood the mechanics including how to play them correctly with minimal loss. Once players hit the ceiling or close in on it you can truly see how well a job is able to perform.
But balancing around bad gear, poor gameplay, inattentiveness, cluelessness about mechanics, poor connection etc. is a bad idea and those are the most common reasons for low percentiles. Paying as much attention to low percentiles as to high percentiles means the devs would include factors into job balance that have nothing to do with the core of the problem.

If another player kills you and your dps drops it's because another player killed you, not because of poor balance and it shouldn't be considered for future changes.
If you learn a new fight, your dps usually absolutely sucks on BLM compared to RDM. That does not mean that RDM is in any way stronger than BLM, it only means that it is much easier to wing a fight on RDM.
This is something that can be looked into so one job isn't considered too difficult or punished over others and thus rarely played. But the dps range of jobs on the various ex and savage bosses are generally more an indicator for how easy it is to master one job compared to others.

[Gruntler]

[QUOTE=Rilifane;5273275]

This is something that can be looked into so one job isn't considered too difficult or punished over others and thus rarely played. But the dps range of jobs on the various ex and savage bosses are generally more an indicator for how easy it is to master one job compared to others.

Which is a vector of balance; ergo; should be looked at. Describing something that is completely a part of job balance, and saying 'No, we shouldn't look at that' is lazy data-use.

I had a longer explanation, but really, this is the meat--you bring up points of job balance and then say we should not look at them, and then wonder why I am criticizing 'look at 95' as lazy. It's lazy. It's lazy thinking, lazy hypothesis, asking us to look lazily at data because you're afraid 'the bads' will influence job balance for some undefined reason.

Fact: Difficulty is based around prog therefore that is where balance is most important. Looking at post-prog mastery only shows how well a job can do on shit it's already conquered. This is important for balance, of course, but it is not the only vector, and 'look at 95' ignores those vectors at the expense of real balance.

If another player kills you and your dps drops it's because another player killed you, not because of poor balance and it shouldn't be considered for future changes.

is as strong an argument for not looking at a single data point as I could make, so I don't know why you could make this argument, not realize there's a similar argument in padding ('my dance partner was a god' 'my buff was physical only so we excluded magic classes') and that therefore forces one to conclude that looking at a -single parse- (which is exactly what 'look at 95 is') is bad.

[Mansion]

Fact: Difficulty is based around prog therefore that is where balance is most important. Looking at post-prog mastery only shows how well a job can do on shit it's already conquered. This is important for balance, of course, but it is not the only vector, and 'look at 95' ignores those vectors at the expense of real balance.

I don't get that. Prog is prog, meaning that it's likely that kits are not used at their most optimal potential. Specifically a RDM can see its DPS lowered by raising and MP challenges linked to that. And prog is finite, it's a matter of a few days / weeks. I agree that it is part of the overall balance and has to be considered, but it's always better to look mainly at optimal play on high end content because it simply trickles down to every use of said job. If a job performs well in Savage, there's no real reason for it to not perform in lower difficulty content. While a job that does well in dungeon does not necesseraly work well in savage (see 4.X WHM for instance)

The only "problem" here is that we tend to not have AoE at all in Savage / Ultimate so you can't really look at that kind of data. But jobs tend to be focused on single target rotation anyway.

[Gruntler]

I don't get that.

It's very obvious. In a PvE game, the object is to overcome the environment. One vector of balance must therefore be in the capability of overcoming said environment. If it's based purely on dps, but a job has inherent problems overcoming environments, there's a problem in balance that you cannot see looking at one single parse. If a job is supposed to take a hit to dps in exchange for having less problems overcoming environments, but it is not overcoming the environments more, there's a problem in balance that you cannot see looking at a single parse.

If you have balance in multiple vectors, and you don't look, you don't see balance issues. This shouldn't be a hard concept.

Prog is prog, meaning that it's likely that kits are not used at their most optimal potential.

Optimal. Potential. Is. Only. One. Vector. Of. Balance.

So one someone says 'Okay, look at that, but let's look at other things too, because those also matter regarding balance' what's the resistance or difficulty, if not laziness?

Specifically a RDM can see its DPS lowered by raising and MP challenges linked to that. And prog is finite

Prog is also the primary expression of difficulty in a PvE game. Again--when you look at balance in say, Dark Souls, do you look at speed runs, or do you look at how people struggle? There is FAR more data in the struggle than there is in mastery, in PvE.

, it's a matter of a few days / weeks. I agree that it is part of the overall balance and has to be considered, but it's always better to look mainly at optimal play on high end content because it simply trickles down to every use of said job. If a job performs well in Savage, there's no real reason for it to not perform in lower difficulty content. While a job that does well in dungeon does not necesseraly work well in savage (see 4.X WHM for instance)

Strawman. I'm not saying that we look at dungeons and shit, because the jobs, as you say, are not tested there.

What I'm saying is that, when looking at Savage, you look at more than one, single, solitary parse. Which is what you're doing when you look at 95th percentile--that's not a spread of parses, even at mastery level. It's literally just the one parse that happens to fall into some arbitrary mathematical criteria. That's the point: The answer to your concerns is not to eschew data for datum. Datum is useless. Data is useful. And one parse is -Datum.-

[Kabooa]

What I'm saying is that, when looking at Savage, you look at more than one, single, solitary parse. Which is what you're doing when you look at 95th percentile--that's not a spread of parses, even at mastery level.

95th percentile is not a single parse. It's the 95th percentile. It's taking the aggregate of data and saying "Out of a scale of 1 to 100, those in this range did better than 94% of people and worse than 5%".

It's not a singular parse.

We choose 95th-99th for the following reasons.

A) These are people who are playing the game better than 94-98% of the rest of those that publicly provide data, which is already much less than those who play. These are players who may not understand the underlying gearworks and theory but understand its practice.

B) We prefer it to max because Max tends to be examples that aren't easily replicated - like critting 100% of your big potency dumps despite having 20-25% crit rates. The vast majority of 'good' runs will fall in the 90-95th range.

C) The players on the left end of the bell curve are generally making significant mechanical mistakes that would convey across any job, not just the job in question - Which is why we basically exclude just about everything under 75th. A job's relative strength cannot be accurately measured by those who haven't figured out the bare essentials.

[Adventica6]

It pays way too much for it's utility for the damage that it deals, plain and simple. Either remove vercure and increase it's damage or remove verraise and increase it's damage, we don't need both.

Rdm's in FFXI were way, way, superior to what we have here, by leaps and bounds.

A jack of all trades does not mean do everything poorly, it means do everything decent, please follow your own job identity that has been in place since 1987.

[Archwizard]

It pays way too much for it's utility for the damage that it deals, plain and simple. Either remove vercure and increase it's damage or remove verraise and increase it's damage, we don't need both.

In fairness, Vercure does have a mechanical benefit to RDM as a means to proc Dualcast on-demand when targets aren't available.

But if your point is already that its damage cost for its utility is too high, wouldn't another solution simply be... reducing the cost of its utility, thereby bypassing any need to remove anything at all? Your argument hinges on the damage costs being disproportionate, not on the utilities themselves being worthless.