All Worlds Emergency Maintenance (May 13)

[BerenTebogo]

I have never, ever, ever seen Retaliation proc on an anticipate and as I've already said, I full time Seigan + TE.

Yes Hasso could do more damage but unless I'm in a party with 5 other people I can reach out and punch in the face I'm going to enjoy my counters and not die.

[Cadsuane]

I'm not really a math major so it's hard for me to explain this concept. But try think about this for a second. We all know haste has increasing marginal returns. But by your logic it shouldn't. I's not like having more haste changes the way haste works (like how the TP formulas switch). The equation is constant. So why does haste have increasing marginal returns?

It's an artifact of math from how haste works as a divisor in the equations. The same sort of thing is in effect with your base delay. Reducing the top of the equation proportionally (i.e. by the same amount of haste) doesn't yield proportionally similar results when you process the rest of the damage equation. There turn out to be "sweet spots" in base delay to be where you'll get better effective results in reduction from your haste.

Oh for fuck's sake. I am perpetually surrounded by math morons.



The graph you showed described how the proportion of TP gained per delay is throughout the possible delay range. Calculating the TP per hit based on delay, or the proportion of TP per delay, is described by different linear equations depending on what delay range your weapon falls into. for example:

Perdu Voulge TP per hit: 13.0 + 1.5/50(504-480) = 13.7
TP/delay: 504/13.7= 0.0272

Maneater TP per hit: 5.0 + 6.5/270(276-180) = 7.3
TP/delay: 276/7.3 = 0.0264

The graph you showed relative TP gaining speeds in proportion to the speed you get at 999 delay. According to that graph, 190 delay is about 1.5 times as fast as 999 delay. All of this is completely irrespective of haste.

If you add dual wield or something like a sword strap, you will lose delay which means you also lose TP per hit. Because of the non linear scaling of delay/tp, from around 450-520 delay you will LOSE TP gain speed by adding a sword strap and swinging faster (in a non-discreet model). Similarly by adding 5% dual wield you will not increase your TP gain as if you were adding 5% haste, since you will lose TP per hit in the process. Also the interaction between haste and delay reduction is multiplicative and not additive, obviously, DWers can't reach 100% haste with DW.

Basically, haste gives equal benefits to all delays. 50% haste doubles your attack rate. THAT'S IT. It completely circumvents the issue brought to light with TP per hit calculations. All that graph illustrates is that certain delay ranges have an inherent advantage with the proportion of TP to delay. It is only applicable when evaluating your base delay, and any reductions to delay, which does not include haste. It says that, right on the fucking graph.

[Gredival]

Wow, it looks like higher delays lose more delay for a set amount of haste, math works apparently. (A set percentage subtracted from two numbers will remove a larger amount from the larger number, obviously)

In the actual way the game works, both delays gain roughly the same amount of total damage dealt from a set amount of haste, but strictly speaking you can NOT say that lower delays benefit more mathematically, because the opposite is true.

What's I'm trying to point out is that there a mathematical effect similar to haste's increasing marginal returns.

We all agree haste has increasing marginal returns right?

So let's look at these numbers for instance.

200 Delay w/ 60% Haste = 80 Delay
200 Delay w/ 59% Haste = 82 Delay
200 Delay w/ 58% Haste = 84 Delay
200 Delay w/ 57% Haste = 86 Delay
200 Delay w/ 56% Haste = 88 Delay

When you look at it this way, every single percent of haste reduces delay by the same amount, 2 delay. When you get more and more haste, it's not like each point of Haste is taking away more delay.

So why do we say that haste has increasing marginal returns?

What we mean is that when you add 4% Haste moving from 56 to 60 Haste there is a ~9% reduction in delay when you compare 80 to 88 when you look at initial delay - reduced delay divided by the initial delay.

Thus a 4% increase in Haste is actually a 9%~ decrease in delay marginally, that is to say, with reference to the previous delay not the base delay.

Now as an artifact of how FFXI calculates math and when you look at it from the marginal perspective, Haste returns vs. Delay turns out to be a slightly non linear equation which tends to benefit lower delays.

For some reason I guess I took this for granted and I wanted to avoid doing the math. But let's look at some numbers. The first set, if you don't recognize it, is the delay a set of Thief daggers (chosen to emphasize speed comparative to both Ridill and Gaxe)

----

252.5 at 23% Haste = 194.4
252.5 at 25% Haste = 189.3

2% Haste actually gives 2.62% Marginal Haste

412.8 at 23% Haste = 317.8
412.8 at 25% Haste = 309.6

2% Haste actually gives 2.58% Marginal Haste.

504 at 23% Haste = 388.0
504 at 25% Haste = 378.0

2% Haste here actually gives 2.57% Marginal Haste

-----

I'm actually surprised that it's so close between Gaxe and Ridill. So yes, each chunk of haste knocks off more actual delay when you're starting with a higher delay weapon. But when you look at the marginal returns, lower delay is gaining more marginal haste.

Of course in a real parse situation we're not talking about a buff of just 2%. What I mean by the faster DDs getting more benefit is that they will get better marginal haste with the extra 35% song/spell haste than the Gaxe. I'm really not interested in going back and crunching the numbers for 35% though.

Point is that marginal Haste benefits lower delay. This is seen most clearly comparing the pair of daggers versus either of the War sets. Granted now that I actually look at it, it's pretty small. I thought it was going to favor Ridill far more than it probably will.

Basically, haste gives equal benefits to all delays. 50% haste doubles your attack rate. THAT'S IT. It completely circumvents the issue brought to light with TP per hit calculations. All that graph illustrates is that certain delay ranges have an inherent advantage with the proportion of TP to delay. It is only applicable when evaluating your base delay, and any reductions to delay, which does not include haste. It says that, right on the fucking graph.

Math Above. Stop foaming at the mouth about your crusade against the evil Ridill empire, I already in the very same post you quoted that the chart was ill-deployed.

What I wanted to illustrate was there is mathematical artifact to how marginal haste is calculated such that lower delays have better marginal returns for equal amounts of haste increase. The math, again, is above.

[Max™]

Wait, what?

You're comparing the difference between a - (b%) = c, and a - (b+2%) = d.

While doing this, you're comparing c to d, and noting that for lower values of a, c - d is a larger percentage of a?

Let's take your first example, 252, giving 194 and 189, 5 delay off for 2% haste.
Great Axe, 504, giving 388 and 378, 10 delay off for 2% haste.

How you're turning 504 - 10, and 252 - 5 into 2.57% and 2.62% is still losing me, I honestly cannot determine where you're getting the importance of those numbers from.


I can however understand the benefits of haste, so here ya go.

504 - 10% = 453, 504/453 = 1.11~
504 - 25% = 378, 504/378 = 1.33~
504 - 80% = 100.8, 504/100.8 = 5~

252 - 10% = 226, 252/226 = 1.11~
252 - 25% = 189, 252/189 = 1.33~
252 - 80% = 50.4, 252/50.4 = 5~

412 - 10% = 370, 412/370 = 1.11~
412 - 25% = 309, 412/309 = 1.33~
412 - 80% = 82.4, 412/82.4 = 5~

I'll be damned, set amounts of haste give the same increase in hit rate when evaluated properly, what an amazing proof that math works.

1.11 to 1.33 is not a linear increase as the jump from 10% > 25% might suggest, 1.33 to 5 is not either, when you plot it, the curve shoots upwards towards infinity at 100% haste. Thus, increasing returns relative to each point you already had, regardless of your starting delay.

[Ruuz]

Even though Gredival's posts are maily tl;dr for me, he is right about the Haste (though I really dont give a shit about the ridill debate anymore). The lower the base delay, the more actual improvement you are getting. This is especially true for NIN and MNK (and why Haste is so important). Still, GAxe or nothing for WAR imo.

[Max™]

NO HE IS NOT RIGHT!

FUCKING GODDAMMIT!

1.11 = 1.11
1.11 - 1.11 = 0
1.11 x 1 = 1.11
1.11/1.11 = 1

That number does not care if you have one hundred billion delay, or one hundred delay.

Taking a percent off one hit rate which is taken to = 1, gives you a new hit rate, in the case of 10%, 1.11.

1.11 > 1
0.11 > 0.1

You get more from each point of haste than the one before it gave you, RELATIVELY SPEAKING, higher delays lose a more significant chunk, but the hit rate boost is the same.

504 to 100 is 404, 252 to 50 is 202~.

404 > 202, but 504/100 = 252/50, and 504/404 = 252/202.

Same benefit in the end, if you disagree you should take it up with the guy who invented math.

[Gredival]

Wait, what?

You're comparing the difference between a - (b%) = c, and a - (b+2%) = d.

While doing this, you're comparing c to d, and noting that for lower values of a, c - d is a larger percentage of a?

Let's take your first example, 252, giving 194 and 189, 5 delay off for 2% haste.
Great Axe, 504, giving 388 and 378, 10 delay off for 2% haste.

How you're turning 504 - 10, and 252 - 5 into 2.57% and 2.62% is still losing me, I honestly cannot determine where you're getting the importance of those numbers from.


I can however understand the benefits of haste, so here ya go.

504 - 10% = 453, 504/453 = 1.11~
504 - 25% = 378, 504/378 = 1.33~
504 - 80% = 100.8, 504/100.8 = 5~

252 - 10% = 226, 252/226 = 1.11~
252 - 25% = 189, 252/189 = 1.33~
252 - 80% = 50.4, 252/50.4 = 5~

412 - 10% = 370, 412/370 = 1.11~
412 - 25% = 309, 412/309 = 1.33~
412 - 80% = 82.4, 412/82.4 = 5~

I'll be damned, set amounts of haste give the same increase in hit rate when evaluated properly, what an amazing proof that math works.

1.11 to 1.33 is not a linear increase as the jump from 10% > 25% might suggest, 1.33 to 5 is not either, when you plot it, the curve shoots upwards towards infinity at 100% haste. Thus, increasing returns relative to each point you already had, regardless of your starting delay.

But haste is linear. Look.

200 Delay w/ 60% Haste = 80 Delay
200 Delay w/ 59% Haste = 82 Delay
200 Delay w/ 58% Haste = 84 Delay
200 Delay w/ 57% Haste = 86 Delay
200 Delay w/ 56% Haste = 88 Delay

Every point of haste in this case grants 2 less delay. If you plotted out a graph of haste to delay it'd a straight line with a slope of 1/2. For every 1 unit right you move for haste you move down 2 units for delay. That sounds linear to me.

So wtf does it mean to say Haste has increasing marginal returns?

When we say haste has increasing marginal returns we say it's marginal because the base point of comparison is the last amount of delay/haste, not the default delay. That is the definitional sense of marginal.


If you don't think that these numbers bear an impact, then you're also denying that Haste gets increasing marginal returns.

0% haste gives you a 1.00 multipler, 16% haste gives you a .84 multiplier. The difference between both these percents is ~16. 50% haste works by giving you a .5 multiplier, 66% haste gives you a .33 multiplier. The difference between percents and multipliers are both also ~16.

But we commonly hold, due to increasing marginal returns, that going from 50% to 66% is a bigger boost. That doesn't make sense unless you are looking at it from a marginal perspective, otherwise in both cases there is equal return (16%)

I'm probably not explaining this well at all but the numbers crunching backed me up, and I've always understood this as what is meant by "Haste has increasing marginal returns"

[Max™]

200/80 = 2.5
200/82 = 2.43~
200/84 = 2.39
200/86 = 2.32

You don't compare it to what each point of haste before it took off, you compare it to the original delay/modified delay, or it is MEANINGLESS.

The number crunching doesn't back you up in this case, because you're crunching the wrong numbers.

200 x .5 = 100, 200/100 = 2
200 x .33 = 66, 200/66 = 3.03~

[Cadsuane]

Haste has increasing returns with respect to your rate of attacks. Shut the hell up about this marginal returns bullshit, that's just a buzzword they use on alla. (psst, neither you or they know what it means)

Delay * (1-haste) = Hasted delay

This is a linear equation, so yes, the relation between haste and delay is linear. Since DoT is what you'd be interested in for evaluation purposes though, you need a rate of attack. You calculate rate like this:

1 / period = frequency or rate

So we'll find out what 504 delay is in hits per second, or it's rate of attack or frequency. Assume 60 delay = 1 second, 504/60=8.4s:

1 / 8.4 = 0.119

So 504 delay (or 8.4 seconds) attacks 0.119 times every second. A composite function that calculates attack rate after haste has been added looks like this:

(8.4(1-haste))^-1 = x

Which commutes to...

0.119/(1-haste) = x

Which should look pretty fucking familiar. It's a rational equation. The single root in the denominator indicates a vertical asymptote at x=1, which is where this function is undefined because f(x) will go to +- infinity. Graph it and it'll look like two sweeping lines that curve away from x=1 while getting infinitesimally closer to it.

Another way to explain it is haste takes off the same chunk of delay with respect to your original attack speed every time you add it, and every subsequent time that chunk represents a larger portion of your remaining delay. So, your attack rate rockets up. For example, you need 50% haste to halve delay once and double attack rate. you need 25% to do it again. You need 12.5% to do it again, etc.

Functions are something you should have learned in grade 10 and it's a much nicer way to describe relations between variables; please shut the fuck up with this english major semantical masturbation. If you can't understand the algebra here, just assume you're wrong and you're too poorly educated to figure it out. Save us all lots of time.

EDIT:

Math Above. Stop foaming at the mouth about your crusade against the evil Ridill empire, I already in the very same post you quoted that the chart was ill-deployed.

What I wanted to illustrate was there is mathematical artifact to how marginal haste is calculated such that lower delays have better marginal returns for equal amounts of haste increase. The math, again, is above.

O RLY? So I'm lead to believe that when you said this:

For the optimal haste situation for WAR's we're talking 23% haste in gear, 20% in BRD, and 15% in spell? 58% reduction then.

.42 x 409 = 171.78 (Ridill)
.42 x 504 = 211.68 (Gaxe)

As you can see, haste gets the Ridill just past the very sharp spike drop, putting it better off, relatively, to where it was before. Gaxe turns out to be roughly where it was before.

For only one march we're talking 49% reduction

.51 x 409 = 208.59
.51 x 504 = 257.04

Here again Ridill turns out better as 257 places at a decreasing part of the slope.

I believe it's the same sort of effect in Dual Wield (DW naturally favors weapons with less combined delay) which makes Koggelmander/Perdu so attractive to BLUs -- they can get into a "sweet spot" for delay that optimizes how reduction trades off with TP.

You weren't trying to say that because haste puts your ridill and axe equivalent delay on the sharp drop or "sweet spot" of the TP/delay graph, you get the TP/delay benefit of having a weapon with that actual delay? As if Haste has a single solitary fucking thing to do with delay/tp? Or are you just backpedalling to not look like quite as much of an idiot?

[Max™]

I think he honestly believes there is some magical delay range where you get more from haste.

Even with it being beaten into his head repeatedly, he keeps insisting on it.

[Burnsy]

Even with it being beaten into his head repeatedly, he keeps insisting on it.

Alla trolls have naturally high defense Max, we better bust out GAxeWAR/Hasso posts.


Did he say somewhere that a healer can't keep up with /sam 2H, and /war MNKs? Do the bards and red mages on this guy's server not know what ballad/refresh does?

[Tempyst]

Alla trolls have naturally high defense Max, we better bust out GAxeWAR/Hasso posts.


Did he say somewhere that a healer can't keep up with /sam 2H, and /war MNKs? Do the bards and red mages on this guy's server not know what ballad/refresh does?

Look, 3 a tick from the red mage, 3-7 a tick from the bard, and 3-4 a tick from the corsair isn't enough to keep up with the wheelings and dealings of a melee who isn't subbing ninja. Maybe when they add a fourth job that can do a different refresh effect it can be considered. As you can see from my fifty pages of math and misinterpreted charts, melees can effectively sub SAM when the mages have 35 mp/tick.

[Xno Kappa]

I have never, ever, ever seen Retaliation proc on an anticipate and as I've already said, I full time Seigan + TE.

Yes Hasso could do more damage but unless I'm in a party with 5 other people I can reach out and punch in the face I'm going to enjoy my counters and not die.

I have absolutely nothing to contribute to this thread except to say your sig gave me a hardy lol. I thank you.

[Limlight]

oyi not this again. ridill /nin is nice for certain areas in einherjar and maybe main blink tanking stuff in limbus. But yea gaxe or go home.

[Cadsuane]

What the fuck did he mean by a "mathematical artifact" anyway? Besides something that was totally there and proved the point or would if he could only explain it but can't because he's not a math major. Goddamn half-wit.

[Max™]

Supposedly he was conflating the points where Delay/TP is ideal with the relationship between Delay/Haste, even after it was pointed out that Haste doesn't alter TP/Hit.

Then he started with that weird ass backwards math that you get an increasing margin from each bit of haste for a smaller delay.

I mean, I can reverse the math and see a relation, I just don't know where the 2% haste = 2.67% at 252, 2% haste = 2.57% at 504 shit came from in the first place.

I can't logically take anything regarding the haste calculation and arrive at those numbers.

[Cadsuane]

Oh, lol. I understand what he did.

252.5 at 23% Haste = 194.4
252.5 at 25% Haste = 189.3

2% Haste actually gives 2.62% Marginal Haste

412.8 at 23% Haste = 317.8
412.8 at 25% Haste = 309.6

2% Haste actually gives 2.58% Marginal Haste.

504 at 23% Haste = 388.0
504 at 25% Haste = 378.0

2% Haste here actually gives 2.57% Marginal Haste

He calculated 1-309.6/317.8 to get 2.58%, and 1-378.0/388.0 to get 2.57%, and so forth. What a jackass. He thinks the proportion of period lost is the metric by which haste is evaluated. Evaluate frequency, not period! 1/p=f!

I can't believe how basic concepts like the measure of frequency or recently the concept of constant/diminishing returns escapes some people, who will turn around and argue endlessly about stupid shit they don't understand. I liked it when someone on this forum replied to another cock-waving pseudo mathematician with an irreverent "crunch the numbers for us bro"; I mean, math has been abused and fucked beyond all recognition in regards to stupid FFXI mechanics by really dumb people who wouldn't have passed grade 11 functions in the high school I went to. Sers, open message to the FFXI community: fuck off with the math. You suck at it.