Diminishing Return on Stats

[Viridiana]

Marginal diminishing value is the general result where stacking a stat becomes poorer from a value perspective. This is the basic concept of +10% crit when you have 0 base crit is worth +5% DPS. +10% crit when you have 60% base crit is worth +3.7% DPS. The more crit you stack, the less useful stacking crit is. This effect has a slow decay (each +1% decays the value by 1%).

While I see how this could be construed as diminishing returns, in another sense, it isn't really. If all your other stats stayed the same and you only increased Crit, then each % increase would give you the same increase in total DPS.

[Gamemako]

One thing oddly absent from this discussion: abilities also diminish the relative returns of certain stats. Automatic critical hits and other flat critical hit bonuses (Internal Release, for example) reduce the relative returns from the critical hit rate stat. Also worth noting that Determination appears to also suffer diminishing relative returns to strength; i.e. it is at least partially additive with strength rather than multiplicative.

It's not. That's only a reflection of the UI truncating numbers. A guy on reddit showed the detailed tests of Rez (8s base spell) and the cast time appeared to be reduced in 0.1s increments at the same ratio as a 2.5s base spell was reduced. E.g. the SS formula is probably a straight % reduction in cast time, with the game UI truncating.

As a note, this reduction is applied after most other bonuses and is additive with them; for example, Greased Lightning reduces your GCD natively to 2.125s, and skill speed still reduces by 0.00069s per point thereafter (rather than 0.00059, which it would do if it were applied beforehand and then reduced by the 85% factor from GL3). I believe Fey Light at the very least is an exception, acting directly on the Skill Speed stat itself.

However, the benefit of the acceleration requires a LOT of the stat to be stacked -- much more than is reasonable without killing your TP bar, or mana bar for most classes.

This is true, though not really the reason everyone hates skill speed. The damning thing about skill speed is that it fails to affect autoattacks, off-GCDs, and DoTs (barring situations where reapplication is naturally delayed by things like Jump), which in total comprise upwards of half of total DPS for all DPS jobs but BLM. Accelerating returns can't touch that one way or the other. Until they fix it so the stat affects autoattacks at least, the only job deliberately taking it will continue to be a tank. If it did affect at least autoattacks, it would be nearly as valuable as any other stat for many jobs (especially MNK).

[EasymodeX]

no we shouldn't understand "increase of DPS in %".

That's the context of the discussion of MMORPG damage optimization with incremental stats/gear. So yeah.

If I ask you, durng an exam "is DPS linear in crit" and your answer is no, you failed the question.

This isn't math class.

If anything, this is economics.

The damning thing about skill speed is that it fails to affect autoattacks, off-GCDs, and DoTs [...] Accelerating returns can't touch that one way or the other.

Well, the fact that it doesn't affect AAs and oGCDs simply sinks the initial value of SS for classes that have those. The acceleration still happens and is still relevant. Specifically, Monks, due to the GL stacking as you noted earlier (increasing the initial value for SS), get just about / almost as much value from SS as they get from crit (crit gets chunked down with diminishing marginal returns on normalized value due to the BS autocrits and traited IR that you also mentioned).

[Gamemako]

Well, the fact that it doesn't affect AAs and oGCDs simply sinks the initial value of SS for classes that have those. The acceleration still happens and is still relevant.

I'm just saying that the TP really isn't too much of an issue in practice, but the other disadvantages pretty well relegate the stat to niche status. Adding autoattacks would at least give MNK another stat to consider; right now, DTR is twice as valuable as either alternative.

[EasymodeX]

While I see how this could be construed as diminishing returns, in another sense, it isn't really.

Yeah. The confusion comes down to context: return on <insert stat> for what?

Return on crit rating for crit% is completely irrelevant. What players want to know is what their return on X stat for DPS is.

Please, don't throw stupid statements like "diminish by stacking a single stat". Spell Speed has an increasing return.

L2r before clicking the reply button.

[Alcyon1]

EasymodeX : if you don't change the other stats, X CRIT = Y DPS. This is the definition of LINEAR in terms of DPS.
What you are talking about is X CRIT = ? %DPS. There, it depends of your base crit. Example ?

My base DPS (with 0% crit) is 100.
At 0% crit : I increase my crit by A, giving me 0+X% crit. Conclusion : my DPS increases of X/2 or X/2%.
At, let's say 50% crit (my DPS is 125 now) : I increase my crit by A, giving me 0+X% crit. Conclusion, my DPS increases of X/2 or X/2.5%.

SPS is theorically the best stat for BLM's, but many people like to see big numbers. Some statistical tests showed how full SPS > full crit. Since SPS has an increasing return .....
PS increasing return = better than linear by definition.

[Judge_Xero]

2. Spell Speed / Skill Speed has accelerating returns. Technically they're not exponential, like mathetmatically exponential. However, SS actually has accelerating (better than linear) returns. More SS makes SS more valuable. Pretty nuts.

I'm curious if there is more to this that we will see with a level cap raise.

I never checked what the base stats were at level 1 and I don't have a level 1 DoW/DoM to check now but....

If the GCD is 2.5s with say 20 SKP @ Lvl 1 (base stat) and the GDC is 2.5s @ lvl 50 with 341 SKP (base stat) then
either the floor for the GCD is 2.5s and we only started to see the effect @ 341 SKP for it to start going down, or it is relative to
Base Stat + Character Level. (Greater returns the higher level we get)

Has anyone done any low level skill speed testing to see the amount of SKP required to get the GCD to budge a little?

[EasymodeX]

Anecdotal evidence points to low level = doesn't need much SS to get improvement. Which means if the level cap is raised, then it would take more than 10.5ss to get 0.1s off the GCD.

[EasymodeX]

EasymodeX : if you don't change the other stats, X CRIT = Y DPS. This is the definition of LINEAR in terms of DPS.

It's not linear in terms of +% DPS, which is what you want to know. You want to know this to compare it against other alternatives, like DTR or SS or w/e.

At, let's say 50% crit (my DPS is 125 now) : I increase my crit by A, giving me 0+X% crit. Conclusion, my DPS increases of X/2 or X/2.5%.

Your DPS increase is [(X/2)/(125)]% which is less than [(X/2)/(100)]%.

The second increase is decayed because you already have 50% base crit.


Edit: Why am I even bothering? You either get it or you don't. This conversation has been repeated a hundred thousand times on MMO mechanics discussions across the internet.

[Alcyon1]

I get it, I just say :
THE INCREASE IS LINEAR IN TERM OF FLAT DPS (DAMAGE PER SECOND) and NOT LINEAR IN TERM OF %DPS. Stop forgetting to say EXACTLY what you're cheking. If your DPs is 250 and after some changes in your stuff, your new DPs is 275, you can say about your DPS "the increase is 10%" or "the increase is 25", both are correct. When you are testing/calculating your DPS, you are just checking the final outsome and not the increase in %. You are taking the highest DPS as the BiS.