So does every hate new PLD now?

[Lyth]

??????

I think that rDPS parity is the cornerstone of job balance. If jobs don't have similar rDPS values, then they're not balanced. If you want to create a second order model to look at dps(t) distributions and push them to similar burst profiles, that's fine. The data exists but it's not tabulated in a way that makes it easy to compare.

By curve, I'm referring to the distribution of your damage output over time. You can look at actual damage output if you like, or damage per second. In simple terms, if your burst profiles are similar, then the unmeasured benefit you provide to buff providers will be similar. But so too will your rDPS. What you're arguing for doesn't work, because this unmeasured benefit isn't consistent. Let's say every buff provider gains 20 rDPS extra from having you around as opposed to another tank, so I give you an 80 rDPS disadvantage. Well, then groups run a comp without buff providers and then just take the tank with higher rDPS. If you want to be obsessive about that 20 rDPS gain, then you have to just make the burst profiles identical as well. But that doesn't change the need for rDPS parity. That's why the first order model exists.

Also, I'm not sure why you would offer up singular data points on Crit and DH rate buffs. How can you even begin to guess what the benefit is off a single value? If you Crit on every attack for the next 15 seconds, is that because of the buff? What if you would have rolled the exact same string of Crits even without the buff? What if we compare a data point where players get no Crits across the entire group? It's incredibly easy to cherry pick data on this type of buff when there is no compiled data set on it.

Next you're going to try to offer me numbers on Everburn, I can just sense it.

[Shurrikhan]

I think that rDPS parity is the cornerstone of job balance.

It is for buffers, because it then includes what they provide towards team synergy, thus giving a sense of all that they provide. It is not the cornerstone for non-buffers, for whom it excludes all that they provide for team synergy, thus only giving an incomplete picture of their merit. It's not that hard.

By curve, I'm referring to the distribution of your damage output over time.

Then no, rDPS parity does not mean an equal damage curve. If I did 3,600,000 damage in 1 second and then again every 6 minutes, and the fight is only 5:59 long, I'll still have the same ~10,000 rDPS as someone who deals a flat rate of 10,000 damage per second. Will those damage curves anything close? No.

In simple terms, if your burst profiles are similar, then the unmeasured benefit you provide to buff providers will be similar. But so too will your rDPS.

Yes, because rDPS is the value without buffs. But unless purposely limiting ourselves to then-weaker comps, we do not run without buffs.

Let's say every buff provider gains 20 rDPS extra from having you around as opposed to another tank, so I give you an 80 rDPS disadvantage. Well, then groups run a comp without buff providers and then just take the tank with higher rDPS.

No, there'd be no reason to. Taking no buffers is still going to cost the party more than that amount of dps, regardless. You'd be taking the single individually "better" job by sticking your head in the sand to ignore the actual total contribution the job can bring... at cost to the party's actual total dps. And that's assuming anyone cares about comp synergy beyond the former likes of "PLD/MCH bad. Me no take you."

What you're arguing for doesn't work, because this unmeasured benefit isn't consistent.

How the hell do you think rDPS works on any job with at least a single buff? They're just as inconsistent, because team synergy (which rDPS rewards only to buffers) is inconsistent, because teams are inconsistent, and yet we can still make real comparisons because when those choices in composition are averaged out over hundreds of thousands of parses with no Stormblood-esque set comps required of different jobs, yeah, they stop being relevant in job comparisons (outside of maybe, maybe whether single-target buffers had decent exploiters with them, since then the relevant exploiter's skill levels only have a single layer of averaging).

Every job in every parse is going to luck out or be screwed over in some way by their actual Crit/DHit chance that run relative to what their gear says they should average. That doesn't make their data useless.

It's incredibly easy to cherry pick data on this type of buff when there is no compiled data set on it.

But there is: the averages. And if the gap in rDPS to aDPS then matches what is seen in the most self-similar parses (same clear times, same comps, same percentiles) that can be compared to between two tanks, yes, that proves out.

And that's not merely a hypothetical, in this case. It's the reality. 6.28 DRK having more than double 6.28 PLD's (10.3% of its total value coming from buff synergy vs. PLD's 4.8%) over the average matches quite neatly with what we'd already seen specifically in the closest comparisons that can be drawn in individual parses.

And neither of those (~5% or ~10% of total throughput) are negligible, yet they're completely left out by rDPS.

Next you're going to try to offer me numbers on Everburn, I can just sense it.

Everburn isn't even in Phase 1. And that is the constraint you brought to these comparisons.

[Lyth]

...

I'm not sure why you keep misreading my statements about rDPS parity despite me reiterating this for the nth time in a row. If two jobs have equal damage curves, then by definition, they have rDPS parity. The converse is not necessarily true. Yet you keep quoting it backwards. Are you doing this deliberately? If you actually need me to explain how if-then statements work then I can, no worries. But otherwise this is a needless discussion.

There's really no distinction between whether you run a group of buffers or non-buffers because - guess what - both types of jobs are balanced on the basis of rDPS, which already accounts for this. The instant you start messing with rDPS to skew for that invisible 20 rDPS gain from your tank's burst distribution, you create inequalities between the two sets of jobs. Now I have a tank job that does extra rDPS at baseline, and only breaks even with other tank jobs when we run a group of buff providers. So we just run non-buff providers and take the rDPS gain. This is why it's a bad strategy to privilege a job with an rDPS advantage, period. If you want to account for the difference due to burst, you just have to make their burst profiles identical.

If you want to analyze data from Crit/DH rate buffs, then you need to compile it as percentile-based data. Ideally, for any buff, you should be able to pick a buff and look at the damage taken under that buff by percentile for each fight (i.e. for PLD, let's say the 96th offers 175 dps under Arcane Circle, while DRK offers 190.) You can look at a few individual runs and try to get a ballpark sense of what the numbers are for flat damage buffs. But trying to do this with Crit/DH rate buffs is impossible. You could cherry pick a run where there's no Crits at all under the buff. You could cherry pick a run where they're all Crits. These buffs are actually the most difficult to quantify the benefit from specifically because they're so variable, and you can't actually tell if the increase in probability was specifically responsible for a given crit. That's why you need to look at the entire dataset to find the expected value.

And Everburn is a confounder, which is why I looked at Phase 1 in the first place.

[Shurrikhan]

If two jobs have equal damage curves, then by definition, they have rDPS parity. The converse is not necessarily true. Yet you keep quoting it backwards.

Because

(A) you've more than implied the converse before whenever the differences between rDPS and non-buffers' total contribution have come up -- over which you've repeatedly dismissed any value a job may bring that is not included directly in rDPS, despite that being up to 14% of a tank's total contribution to a given party, and

(B) your statement otherwise has no bearing on the conversation nor in rebuttal to the points you started this whole back-and-forth off by contesting (that the changes were to make PLD's damage curve more similar to that of other tanks, reducing the degree to which it was an outlier that made exceptionally poor use of raid buffs, and therefore warranted by an underperformance unable to be accounted for in PLD's rDPS metric that would otherwise take overpowering them in light-party content to correct).

There's really no distinction between whether you run a group of buffers or non-buffers because - guess what - both types of jobs are balanced on the basis of rDPS, which already accounts for this.

So, does raw DPS, you realize? The party's total damage always will be the sum of its parts, regardless of whether you give the credit for synergy to exploiters / buff-usage (DPS) or buffers / buff-giving (rDPS). (aDPS is merely DPS made more usefully readable by removing the otherwise conflating gap between a given first and second place due to single-target buffs all being given just to that best user.)

That's not relevant, though, to which metric best account for a given job's typical total value averaged across whatever parties they may end up in. Which is, again, whichever one actually gives credit for what they contribute towards that team synergy. For anyone with more rDPS than aDPS, that's going to be rDPS. For anyone with more aDPS than rDPS, that's going to be aDPS.

There's really no distinction between whether you run a group of buffers or non-buffers because - guess what - both types of jobs are balanced on the basis of rDPS, which already accounts for this.

No. Again, rDPS alone does not account for job balance, and is painfully insufficient to account for the total value brought by non-buffers.

An aDPS-poor job could have higher rDPS (essentially, as a non-buffer, solo DPS) and, paired into the very same composition, provide that party with less DPS than would another job, because the rDPS lost to its buffers will exceed its rDPS advantage over another non-buffer that has higher aDPS.

Yes, the rDPS loss caused by the change in non-buffers is accounted for... but unlike with aDPS, rather than being within the scope of comparison, that difference is rDPS is only shown elsewhere, not on the stats for the jobs you're balancing against each other. Here, we are comparing tanks, who are purely non-buffers, and yet you insist on balancing those non-buffers specifically on the basis of a metric that would say "X is better" even while X provides less DPS to its party than its alternatives. Why the hell would choose to use that metric, then, as your sole point for comparison in this context?

Every buffer's rDPS [their metric for personal+synergies] depends on their buffs' usage. Every exploiter's DPS [their metric for personal+synergies] depends on their buffs received. You do not look at DPS alone for buffers. You do not look at rDPS alone for exploiters.

* And yes, when single-target buffs are involved, the single-target buffer's rDPS gets even more dependent on having a great exploiter, while any optimal exploiter's DPS can greatly widen its lead based on whether it has single-target buffers (to the point of hiding how large the gap between it and second place would normally be). That volatility is why we look at AST/DNC rDPS far more carefully, and why we tend to look instead to aDPS rather than just DPS.

Tl;dr: No, you do not balance tanks --nor any other job that has more aDPS than rDPS-- around their rDPS alone. Such would balance their throughput only for solo encounters. If you want to balance them for party play, you must look also at their measures of what they contribute to party play (instead of specifically trying to remove that element just to, say, make the case that two jobs are wholly balanced while one has nearly 4% more contribution than the other).

[Lyth]

?????

You have directly misquoted me in a way that is false. I have corrected you repeatedly. If you want to claim otherwise, please provide a direct citation or stop making that claim. Thank you.

nDPS and rDPS is identical on tanks. I'm not really sure where you're going with your 'aDPS is a slightly more useful nDPS' tangent. Also, you do realize that nDPS is composition independent, right? By definition you're excluding all external buffs provided by other players.

If you want to see the impact of burst on buffs, we can directly look at the damage taken under any given buff. Nobody has tabulated this data as percentiles yet, but the information exists. If you genuinely have an interest in evaluating this rather than trying to win internet arguments, then this is the quantity to be looking at and offering discussion around. We can directly look at the parameter of interest in this discussion. You don't need to offer a proxy.

[Shurrikhan]

If you want to see the impact of burst on buffs, we can directly look at the damage taken under any given buff. Nobody has tabulated this data as percentiles yet, but the information exists. If you genuinely have an interest in evaluating this rather than trying to win internet arguments, then this is the quantity to be looking at and offering discussion around. We can directly look at the parameter of interest in this discussion. You don't need to offer a proxy.

I'm aware. I'm the one who laid that out for you months ago.

But why would the burden of proof lie with the idea that jobs' throughput should be balanced around all that a job offers to an average party (or that the closest proxy for that metric should be the indicator for this shouldn't be the one specifically designed to ignore a significant portion of its contribution), as compared to the idea that looking at only part of what a job brings would somehow be preferable?

Regardless, that still wouldn't be worth anyone's time nearly so much (as mentioned last time) as using simple relative-potency spreadsheets, as those do not rely on matching all of the various contexts of a given fight, let alone the Crit/DHit variation.

Note: rDPS is no less a "proxy" for the performance of every buffer. Will you be holding every buffer, then, to the same skepticism that'd require careful tabulation by clear time, composition, and the percentiles of the receiver, buffer, and average across the party? Why is the one metric that accounts for team synergy safe despite being equally susceptible to variation, but the other not?