Why is everyone anxious about a PLD Rework?

[Lyth]

When aDPS is looked at from the standpoint of 'percentiles', there is no consideration given to what group compositions were used to attain said percentiles. The result is completely meaningless. The number has no value whatsoever unless you're making a comparison of two runs with the same group.

Unless you're looking at percentile data in the context of the damage generated under individual raid buffs organized by percentile data (Mug, Arcane Circle, and so on), you cannot comment on a job doing 'X' more damage due to contributed DPS.

[Shurrikhan]

rDPS provides a poor comparison between exploiters, always, because it literally leaves their essential teamplay component out of the equation. aDPS provides a poor comparison of buffers, because it leaves their essential teamplay component out of their equation.

If two jobs who always have more aDPS than rDPS have the same rDPS (others are more dependent on them than they are on others for rDPS -- so, an exploiter), probe further at their aDPS; you're not done yet.

Then you can say a NIN benefits by X rDPS by virtue of having a GNB present instead of a PLD, or a NIN benefits Y rDPS by virtue of having a SAM present instead of a DRG. It's a much more effective method of analysis rather than looking at a parameter that is completely meaningless without knowing the composition in which it was generated in.

The same result, from the opposite direction, is doable already, so long as one is cool with plugging numbers for a while. But both your suggested approach and that would require multiple steps until you find a way to fairly balance who gets credit for exploiting a buff; only then could you form a single metric.

In your case, though, by your logic, you'd need each buffers' run to have exploiters of equal or near-equal percentile for a fair test of exploitation; if you can't compare aDPS against (smaller / less impactful) deviations in compositions even when averaged over some 30,000 parses, why would you want data on ability exploitation that'd potential compare a 99th percentile GNB's exploitation of Chain Strategem among 5 other raid buffs to that of 9th percentile DRK's with only two other raid buffs?

Or, heck, if we're comparing theoretical bests anyways, just look at potency maps, where there is no deviation to be had at all.

When aDPS is looked at from the standpoint of 'percentiles', there is no consideration given to what group compositions were used to attain said percentiles. The result is completely meaningless.

When any form of DPS metric, even rDPS, is looked at from the standpoint of 'percentiles', there is no consideration of party mistakes (which badly punish buffers' rDPS), buff synergy, crit rates, etc. Why would the result not likewise be meaningless then?

If a dominant composition trends across a percentile, consider the rest outliers, and compare what you can within in. Then drop to the next percentile in which the pool of jobs expands and compare there. Repeat.

Regardless, you're going to find the same very easy means of comparisons: you can take the party logs (such that the composition is obviously identical) where the jobs you want to compare of similar percentile, and look at the difference between rDPS and aDPS of those jobs.

That is their degree of exploitation, and it shows a significant difference between DRK's and others' (especially PLD/WAR) contribution to the party's total (r)DPS. Whether that's worth doing anything about is a different story. I'm okay with DRK being the damage-focused mad-lad, so long as the other tanks offered something of reasonable equal value. I don't particularly care that PLD is not desirable for post-content speedruns. If that's to be the case, though, I want it to be at least as desirable as DRK (up to a small advantage over the likes of DRK, assuming the party doesn't specifically need its damage carried) in progression.

Now, that may require bringing additional depth to encounters. But there's been anything mutually exclusive between that, kit depth, kit parity, or anything else.