WS Quest

[hexagram23]

Does your assault rank really have a say in it? I'm at floor 30 now, have 309 points, including captain rank and I'm already at "I can feel a marked increase in energy compared to when I first held it. Keep up the good work."

Scholar skilling up weapon sucks donkey dingo.

I have no idea whether or not Assault rank affects anything related to this, but the space between the first and second messages is extremely short. I doubt anything outside of spamming WSs/doing Nyzul Isle influences how fast you get it.

Also, the space between the second and third message is long as fuck. To put it into perspective, I'm on Floor 25, so I need about 6100 WSs. I got the 2nd message after 300 or so WSs, right now according to the WS counter program I'm 4200/6100 and I still haven't gotten the third message yet.

[Enedin]

I have no idea whether or not Assault rank affects anything related to this, but the space between the first and second messages is extremely short. I doubt anything outside of spamming WSs/doing Nyzul Isle influences how fast you get it.

Also, the space between the second and third message is long as fuck. To put it into perspective, I'm on Floor 25, so I need about 6100 WSs. I got the 2nd message after 300 or so WSs, right now according to the WS counter program I'm 4200/6100 and I still haven't gotten the third message yet.

Just noticed an error, know what you meant though

[iceynutz69]

someone posted a ws calculator as well as the weaponchecker on windower forums http://forums.windower.net/index.php...dpost&p=173320

[TummieGaruda]

I have no idea whether or not Assault rank affects anything related to this, but the space between the first and second messages is extremely short. I doubt anything outside of spamming WSs/doing Nyzul Isle influences how fast you get it.

Also, the space between the second and third message is long as fuck. To put it into perspective, I'm on Floor 25, so I need about 6100 WSs. I got the 2nd message after 300 or so WSs, right now according to the WS counter program I'm 4200/6100 and I still haven't gotten the third message yet.

Third message seems to either come up when you are 100 points away, or when you are at 95% completion which in your case is 5795/6100, but im not 100% sure. My friend is breaking her Inferno Claws, I'll have her check at 4712/4960

[miokomioko]

If Assault rank mattered, the NPC would most likely have mentioned it just like he went out of his way to specifically say Nyzul progress affected it.

[LD]

Assault Rank matters if you intend on getting the WS then going to the Mythic.

[Guitarman]

Most likely the differences we are seeing in people's counts is simply human error. If you notice, people claiming different numbers at the same floors are only differing by a dozen or less points. Either the ws calculator missed a few, or they miscounted if they didnt use it. Just round off the numbers they give out and assume it to be the number. 2500 is far more likely than an odd number like 2497.

[miokomioko]

Also, don't do gear swaps for WS points, let it all register in the log if you can... might account for some of the inconsistancies.

[Hadanelith]

Been tracking this forum for a while; just registered a moment ago to make a contribution I hope will be helpful.

Quoting Enygma55:

Spoiler: show

100 250
95 500
90 625 (680, 680) -- 8.8% off
85 750 (700) -- 6.7% off
80 875 (859, 900, 1111) -- 1.8%, 2.9%, and 27% off
75 1093
70 1312
65 1531
60 1750 (1715) -- 2% off
55 2187 (2250, 2320) -- 2.9%, 6.1% off
50 2625
45 3062 (2625, 2645) -- 14.3%, 13.6% off
40 3500 (2800, 3525, 3521, < 3600--floor jump) -- 20%, 0.7%, 0.6% off
35 4375 (4640) -- 6.1% off
30 5250 (5316, < 5511--floor jump) -- 1.3% off
25 6125 (6103) -- 0.4% off
20 7000 (7045) -- 0.6% off
15 8750 ( > 8596)
10 10500
5 12250 (12000) -- 2% off
0 14000



From using the numbers I've deduced a pattern in the calculations.
From floor 0-20 it starts at 14000 and declines a steady 1750.
Once it hits floor 20 the decline value cuts in half effectively 875.

This holds true until the next big floor aka 40, where again the declining value cuts in half once more 437.5.
Next decline value = 218.75.
This seems to hold true all the way up to floor 80. After that they probably just spaced it out accordingly.

Building on Enygma55's data (built on various other data itself), there are further distinct patterns in the data. Namely...


1) Points-per-floor reductions are, as noted, in 20-floor sets (1-20, 21-40, 41-60, 61-80) up until Floor 81.
2) The number of points required HALVES for every set of 20 floors cleared, up to Floor 80. (14k @ 0, 7k @ 20, 3500 @ 40, 1750 @ 60, 875 @ 80)
3) At Floor 81, a unique 15-floor set is created (81-95); otherwise this holds the previous pattern
4) The per-floor point reduction is either locked (read: nonexistant) or unkown for the four floors 96-99 (96, 97, 98, 99)
5) The commonly-known fact that clearing Floor 100 gives ~250point requirement, the lowest of the set.

This being said, attempts to determine an equation thus far have failed, but given the data, it's clear that SE tried (and succeeded) to prevent us from discovering a single linear or quadratic regression line by splitting the data into multiple linear regressions. Before I elaborate on the points-in-case, I'm going to give some assumptions:

A) The required point-total for a "True Zero" disc (Obtained Runic Disc, never even cleared Floor 1) is 14,000
B) Based on posted data, clearing the "n1-st" floor (21st, 41st, etc) of each set gives a FLAT point-reduction bonus, NOT the "n0-th" floor (20th, 40th, etc) as previously assumed; this "n1-st" bonus EXCLUDES Floor 0
C) The required point-values for 20-interval floors (20, 40, 60, 80) set the Base Requirement for the FOLLOWING Tier

With Assumptions A-C... the data actually breaks down very clearly into sets of Linear Regression; this makes attempts at finding an overall quadratic regression line futile, as any such equation will provide a completely inaccurate data for ANY floor other than a "Zero" Disc or a Runic Key (14,000 points & 250 points, respectively).

Please note that for all equations, X = The number of floors CLEARED since the PREVIOUS SET of 20 floors.

The sets - and their respective linear-regression equations - break down as follows:

Floor 0: 14,000 points
Floors 1-20: (14,000 - 350x) points
Floors 21-40: (7,000 - 175x) points
Floors 41-60: {floor}(3,500 - 87.5x) points
Floors 61-80: {floor}(1750 - 43.75x) points
Floors 81-95: (875 - 25x) points
Floors 96-99: ??? NEEDS DATA
Floor 100: 250 points

Using values resultant from these equations, you should ALWAYS get your exact value needed, +/- 5% error as an assumption of POTENTIAL diffences based on Job. Being percentile based, these margins are exceedingly high at low-floor discs and practically nonexistant for high-floor discs; this should account for why we're seeing differences in the ~50's area sometimes, in the multiple-hundreds other times, and sometimes in the thousands.

I'll provide a Margin of Error calculation for the Floor 100 data simply to illustrate the minor variants from 250 points we've seen; the rest are in Spoiler Boxes below.

Floor 100 Jobclass Error (5%)
Resultant Margin of Error Range: 262 points - 237 points
25-point spread (High - Low)

(High: [1.05 * 250] = 262 points {floored from 262.5}; Low: [0.95 * 250] = 237 points {floored from 237.5})


And for those who care, explanation/illustration of the odd spreads we've seen based on hypothetical 5% differences by class...
Example 1 - Using above equation for a Floor 27 Disc:

Spoiler: show

Formula: 7,000 - 175x; x = 7
(7 floors cleared & recorded since first getting the ability to start at Floor 21)
[7,000 - (175 x 7)] >>> [7,000 - 1,225] >>> 5,775 points required

Resultant Margin of Error Range: 6,063 points - 5,486 points
577-point spread (High - Low)
(Please note: 6,063 was floored from 6,063.75; 5,486 was floored from 5,486.25)

Example 2 - Using above equation for a Floor 92 Disc:

Spoiler: show

Formula: 875 - 25x; x = 12
(12 floors cleared & recorded since first getting the ability to start at Floor 81)
[875 - (25 x 12)] >>> [875 - 300] >>> 575 points

Resultant Margin of Error Range: 603 points - 546 points
57-point spread (High - Low)
(Please note: 603 was floored from 603.75; 546 was floored from 546.25)

I've also included an example to show how a Quadratic Regression formula based on data from floors 0-80 (the most stable sets, since the interval changes at floors 81+ and the Runic Key data destabilize things) provide entirely useless numbers. See the spoiler below.

Spoiler: show

Running a Quadratic Regression analysis on the data from floors 0-80, using matched-pair datapoints of (0; 14,000), (20; 7,000), (40; 3500), (60; 1750), & (80; 875) yields the following equation:

Floors 0-80
y= {[2.5 * (x^2)] + (-357.5 * x) + 13,725}

For 20-floor increments...
For x=20 we get 7,575 (75 ABOVE confirmed values)
For x=40 we get 3,425 (75 BELOW confirmed values)
For x=60 we get 1,275 (475 ABOVE confirmed values)
For x=80 we get 1,125 (250 ABVOE confirmed values)

For 10-floor increments NOT divisible by 20...
For x=10 we get 10,400 (100 BELOW formula value of 10,500)
For x=30 we get 5,250 (PRECISELY formula value of 5,250)
For x=50 we get 2,100 (525 BELOW formula value of 2,625)
For x=70 we get 950 (362 BELOW formula value of 1,312 {floored from 1,312.5})

Note that since the Floor 81-95 data follow a clear linear relationship, there is absolutely no point in even attempting to project a Quadratic Regression equation to them. However, if you're interested for OCD-related reasons, the equation is...

Floors 81-95
y= (-25x + 2875)

I have omitted the null-void [0 * (x^2)] statement from the above equation; the remaining equation is simply a linear equation in quadratic form, with a 100% fit to the data.

I hope this information is useful to those of you trying to crack the exact number of WS points you need for your Nyzul Isle WS's.


-Hadanelith of Ifrit


PS: I myself have had absolute CRAP for luck when entering Nyzul with a Static-in-Training, that has basically suspended due to people giving-up before they get our system down =/ Floor 4 cleared, no Radiant Lance or Wightslayer drops for my 75 DRG & BLU.

[Aselin]

Very nice work Hadane.

I've been wracking my brain out on an equation and I still can't come up with anything.

I've already tried quadratic, parabolic, and logarithmic equations. Nothing ever matched.

Yours is the closest so far but it still isn't exact. SE really did something to this that we can't figure out.

We have the values for several floors. We have the constants for a Floor 0 disc and a Floor 100 disc. We know one variable to work with thus far-- number of floors on a disc.

But, finding an equation to work for everything seems almost impossible.

[iceynutz69]

nvm

[Bozo]

Floor 70, 1,317 points. No WS.

Used the Weapon Checker.

>.<

Edit: Just checked, 3rd message.

[Enedin]

Seems that any floor number +20 will result in half the WS points. 30-50, 55-75, 15-35. Might not be super-accurate, but it's nice to know that if you climb 20 floors, your points needed (total) will be somewhat halved.

[Air]

A friend on Floor 15 got Drakesbane with 9280 points.
POIDH? Here we go.
http://img395.imageshack.us/img395/1...dywsks7.th.jpghttp://img395.imageshack.us/images/thpix.gif

He used the latest weapon checker.

[Fwahm]

Otherwiki has this setup for the numbers, supposedly gotten from the JP FFXI wiki.
100: 250

80-99 : 500+20*(99-x)

60-79 : 1000+40*(79-x)

40-59 : 2000+80*(59-x)

20-39 : 4000+160*(39-x)

00-19 : 8000+320*(19-x)

This was spot on with my floor 39 4000 point experience.

[Bozo]

Update: Floor 70, Leaden Salute @1,357

[Hadanelith]

Floor 70, 1,317 points. No WS.

Used the Weapon Checker.

>.<

Edit: Just checked, 3rd message.


Update: Floor 70, Leaden Salute @1,357

Using your numbers to check my equation(s):

Floor 70: Equation
{floor}[1,750 - (43.75x)]; x = 10
1,750 - (43.75 * 10) >>> 1,750 - 437.5 = {floored} 1,312
5% higher (MOE) = ~1,378
WS Obtained @ 1,357; well within a 5% error margin.

Going to make a separate post in a moment to compare the equations Fwahm found on the JP site to my own, using the more-or-less confirmed 20-floor interval numbers as grounding data.

-Hadanelith

[Hadanelith]

Otherwiki has this setup for the numbers, supposedly gotten from the JP FFXI wiki.

Spoiler: show

100: 250
80-99 : 500+20*(99-x)
60-79 : 1000+40*(79-x)
40-59 : 2000+80*(59-x)
20-39 : 4000+160*(39-x)
00-19 : 8000+320*(19-x)

This was spot on with my floor 39 4000 point experience.

That equation set provides the following numbers, juxtaposed against my own...

Spoiler: show

100: 250 (JP) vs 250 (mine)
95: 580 (JP) vs 500 (mine; difference outside MOE)
90: 680 (JP) vs 625 (mine; difference outside MOE)
80: 880 (JP) vs 875 (mine; very similar, difference within MOE)
70: 1,360 (JP) vs 950 (mine; difference outside MOE)
60: 1,760(JP) vs 1,750 (mine; very similar, difference within MOE)
50: 2,720 (JP) vs 2,100 (mine; difference outside MOE)
40: 3,520 (JP) vs 3,500 (mine; very similar, difference within MOE)
30: 5,440 (JP) vs 5,250 (mine; similar, difference within MOE)
20: 7,040 (JP) vs 7,000 (mine; very similar, difference within MOE)
10: 10,880 (JP) vs 10,500 (mine; similar, difference within MOE)
F0: 14,080 (JP) vs 14,000 ("max" not counting MOE; WELL within MOE)

MOE = Margin of Error, no way I was typing it that many times ^_^

Given the 5% MOE worked into my own equations, the JP model matches-up for all 10-floor intervals EXCEPT 90, 70, & 50; our Floor 95 data is also a mismatch.

This isn't to say one model or the other - The JP or mine - is more or less correct, just illustrating the differences for those who want to test both equations.

Fwahm: 4,000 points @ Floor 39?
JP Equation: 4,000 points
My Equation: ~3,675 points, +/- 5% (3,858 - 3,491)

JP model seems to fit Floor 39 much better; perhaps the 20-floor linear set equations are running in 0-19/20-39/40-59 etc. sets rather than 1-20/21-40/41-60 etc. sets. Alternately, there could be 10-floor interval sets at work here, that happen to line-up with the "halving by 20's" rule we've seen so far.

I'll have to compare my equations & the JP equations against data posted here over the next couple days; will fill out a long report (with several spoiler boxes) when I'm done.

For now though, I'm enjoying my weekend ^_^

-Hadanelith



PS:

Further JP vs Hadan Equation Testing...

Spoiler: show

Data based on Bozo:
Floor 70, 1,357 points
JP Equation: 1,360 points
My Equation: 1,312 points, +/- 5% (1,378 - 1,246) {floored values}

Data based on Air's friend:
Floor 15, 9,280 Points
JP Equation: 9,280 points
My Equation: 8,750 points, +/- 5% (9,187 - 8,312) {floored values}

JP equation seems spot-on twice more if you count for a single 3-point error in Bozo's data. My model encompasses Bozo's data if he had 0 error whereas JP model does not (but is insanely close anyway); by contrast, my model does NOT encompass Air's friend's data where the JPmodel hits it precisely.

As mentioned above, will do more comparisons later this week to test both models against this forum's gathered data. The JP model seems more accurate at first glance, but we have to see how it compares against all the other counts in this forum as a test. I really can't see people being 100+ points off of a completely linear model, even if they're counting by hand, unless there is some sort of percentage adjustment for individual Job Classes.

[TummieGaruda]

Otherwiki has this setup for the numbers, supposedly gotten from the JP FFXI wiki.
100: 250

80-99 : 500+20*(99-x)

60-79 : 1000+40*(79-x)

40-59 : 2000+80*(59-x)

20-39 : 4000+160*(39-x)

00-19 : 8000+320*(19-x)

This was spot on with my floor 39 4000 point experience.

I can vouch for this because I posted my findings for floor 30-39 on page 45..

I am starting to see somewhat of a pattern:

Between Floors 30 and 39, the points per floor is looking close to 160 points, and then when going from 39 -> 40, it jumps by 500 points as a reward for clearing the previous tier, and the points are reduced per floor for the next tier.

From 30-40 it would look something like this:


40 3500 <------ Crebont: Page 42 (500 point jump to the next tier)
39 4000 <------ Fwahm: Page 42
38 4160
37 4320
36 4880
35 4640 <------ Daelan: Page 36, Dyterium: Page 42
34 4800
33 4960 < ----- Broke mine at this total
32 5120
31 5280
30 5440 <----- Vector: Page 36 total 5511

160 per floor seems to fit in with 30-39 instead of 175. Once you clear a Tier of floors (the boss floors) you are given a large "discount" off of points. From 39-40, 500 is taken off.

[Hadanelith]

I can vouch for this because I posted my findings for floor 30-39 on page 45..

... ... ...
... ... ...

160 per floor seems to fit in with 30-39 instead of 175. Once you clear a Tier of floors (the boss floors) you are given a large "discount" off of points. From 39-40, 500 is taken off.

Yeah, that does seem to be lining up; my equations were based off a very simple principle: Take the fairly well-recorded values at floor intervals at a set of 20 (including 0), subtract the value for the NEXT set of 20 (ie. Floor 20's points - Floor 40's points) and then divide by the 20-floor spread; this always got you from one set of 20 to another spot-on, and the 5% MOE covered for most variances noted for individual floors.

However, if the sets are constructed 0-19, 20-39, 40-59 etc. rather than 1-20, 21-40, 41-60 etc. (foundation of my model) then yes, my model falls apart completely & the JP model seems to take over. It's going to come down to which model (including any yet to be introduced) best-accounts for the rather extreme variances we're sometimes seeing between people on the same floor(s).

It could be as simple as the JP formula's values are the base and a 5% adjustment (as in my model) is applied to certain Jobs - perhaps Samurai & Ninja for example, given how much more quickly they can build TP for WSs - and that's the end of it. /shrug

-Hadanelith