Help me understand why dps can't be closer in damage output

[aiqa]

It's 5.5%, when I checked yesterday it was closer to 6% (keep in mind imbalance is done from the median where everyone would ideally be, not from top to bottom).

It's a bit high, but it's also still quite close, in particular given the comparatively high number of damage classes this game has. Other games have specs and the like, but they generally do not balance except for a "target spec", GW2 in particular is extreme about this with millions of possible specs for a single class but the devs only look at a handful of particular traits and abilities to even balance at all, nevermind balance against other classes.

WoW for a long time aimed for a 5% imbalance or less, in which case yeah, this is too much. I agree in that regard. OTOH in any game before WoW and its massive success of modern-day inclusive raiding where all classes can participate instead of some being intentionally designed for different content, anything below a 50%++ difference would be "eh, close enough!". There was in fact a funny moment somewhere early WoW Cataclysm where the community started focusing on <3% imbalance instead of <5%, and the devs were caught off-guard by it.

Explain how you got to 6%. People generally use upper quartile to get a "when it's played properly, but not optimized to the extreme" score. And 75th percentile is also the default when looking looking at the "entire tier" scores. So no idea why you claim median should be the default, but median is about the same difference anyway so it doesn't even really matter.

[Carighan]

No I meant 6% from the median between the different jobs. In an idealized world you'd nerf the top performing class by 5.5% (in your numbers) and boost the lowest performing class by 5.5% and everyone else according to their divergence from the point where those two outliers have now met. Making everyone have the exact same DPS output, in the middle of what was formerly the spread.

Although in the real world companies need to pick their battles so most (non-SQEX, who hates easy solutions) would pick the top and the bottom performer each patch and nerf the former while boosting the latter. This'll leave the former 2nd best and 2nd worst as the biggest outliers for next patch, but it also means you only got two calsses to focus on each patch. And if you somehow cause secondary effects making for an unplanned extreme outlier, it'll automatically get selected next patch.

To apply this to FFXIV, next patch would buff Dancer by ~1600 rDPS, while nerfing Dragoon by the same amount.

[aiqa]

Ah I see, but even then it's not 6%. Keep in mind we are talking about median (middle value), not the mean (sum divided by n). There are 13 dps jobs, making the 7th job the median score. Which is RPR.

RPR at "all bosses", upper quartile = 32595
DNC at "all bosses", upper quartile = 29810

29810/32595=0.9145

In that sense DRG is fine.
DRG at "all bosses", upper quartile = 33242

33242/32595=1.0198

[Carighan]

Aaah, sorry, I totally missed that part. Yeah I should not have used median, I keep confusing the two. >.<

[Valence]

You're comparing apples to oranges. XIV doesn't have many variables beyond damage, and predictable scripted models of encounters and deterministic rotations for most of them without any customizable talents or builds. Everybody plays with the same buttons without any rng beyond a minority of jobs. This 5% or whatever the number is, is not there by margin of error, but by design.

[ShaneDawn]

Various reasons have been given in the past as to why they decided to start pushing this balancing philosophy, but ultimately it probably boils down to making sure players are encouraged to bring two melee dps over anything else. Justifications such as "difficulty" or "friction" are hard to take seriously when jobs like Viper and the current iteration of Black Mage are performing strong, or when jobs like Bard are fairly weak compared to the effort to get the most out of it.

[aiqa]

Aaah, sorry, I totally missed that part. Yeah I should not have used median, I keep confusing the two. >.<

I don't see why you should not have used median. Median is a fine way to look at these things.
And when looking at the mean, that is still not "5.5%, or closer to 6%".

DRK 33237
MNK 33125
BLM 33077
VPR 33014
NIN 32917
SAM 32863
RPR 32593
PCT 31799
RDM 31246
SMN 31044
MCH 30619
BRD 30038
DNC 29813

Mean 415385/13 = 31952.69231
DNC deviation from the mean = 29813 / 31952.69 = 0.933

So DNC deviates -6.7% from than mean, or -8.5% from the median.