Sortie +2 Earrings Do Not Exist

Fenrir.Aladeus

Offline

Server: Fenrir

Game: FFXI

user: Aladeus

Posts: 350

By Fenrir.Aladeus 2023-06-29 22:56:02

Link | Quote | Reply

i've gotten 4 +2s so far. 2 close to cap whm, 1 mid drk, and a capped run. i dont play run. lol. meanwhile, i have yet to get +1s for other jobs i play as others here have said.

Ragnarok.Primex

Offline

Server: Ragnarok

Game: FFXI

user: Blazed1979

Posts: 229

By Ragnarok.Primex 2023-06-30 03:05:57

Link | Quote | Reply

Odin.Moonja said: »

Any +2 earrings I have on each of my three accounts, are for jobs that they don’t play.

Pretty sure I have a BLM+2 earring on each, I do not play BLM anymore. I have a RUN+2 on Moonja, who doesn’t play RUN.

I would still be happy just getting some +1 earrings I would use.

I would still like a +1 PLD earring for some things. Only NQs on my account that has PLD.

Spoiler Alert! Click to view.

You villain.
I just noticed my BRD alt has 80 old cases.
What was it, 3% on +2?
Doubt its that high but lets see with pouches.

Valefor.Worlace

Offline

Server: Valefor

Game: FFXI

user: Worlace

Posts: 139

By Valefor.Worlace 2023-06-30 06:39:22

Link | Quote | Reply

Have probably got 400 old cases and yet to see a +2. Any +1 case I get is always PUP +1.

Dodik

Online

By Dodik 2023-06-30 06:50:50

Link | Quote | Reply

NQ cases can't give +2 earrings.

[+]

Asura.Clintbeastwood

Offline

Server: Asura

Game: FFXI

user: Raytheon

Posts: 491

By Asura.Clintbeastwood 2023-06-30 11:41:35

Link | Quote | Reply

I feel like I've been pretty casual at Sortie since its release and I've got +1's for every job except BLM.

Am I just super lucky?

Bismarck.Yvan

Offline

Server: Bismarck

Game: FFXI

user: yvan

Posts: 46

By Bismarck.Yvan 2023-06-30 11:59:56

Link | Quote | Reply

Sounds like it to me, no idea how many cases I've actually gotten but based off my runs I've got.

Earring +1 for all jobs (3 perfect)
Earring +2 SAM (lowest rolled)

Total gali earned: 4,719,727

[+]

Asura.Clintbeastwood

Offline

Server: Asura

Game: FFXI

user: Raytheon

Posts: 491

By Asura.Clintbeastwood 2023-06-30 12:16:27

Link | Quote | Reply

Bismarck.Yvan said: »

Sounds like it to me, no idea how many cases I've actually gotten but based off my runs I've got.

Earring +1 for all jobs (3 perfect)
Earring +2 SAM (lowest rolled)

Total gali earned: 4,719,727

Yeah I have no idea about how many cases I've popped either, but my split:

Earring +1 for all jobs -BLM, -GEO (1 perfect, SAM)
Earring +2 RDM (perfect), WHM (13/19)

Total gali earned: 2,440,000

Offline

Posts:

By 2023-07-04 12:27:44

Undelete | Edit | Link | Quote | Reply

Post deleted by User.

[+]

Asura.Illuminate

Offline

Server: Asura

Game: FFXI

user: ThordainSylph

Posts: 145

By Asura.Illuminate 2023-07-08 02:15:09

Link | Quote | Reply

I think what we need is some solid math to get the basic statistics out of the way. I started saving up NQ boxes until I had 148. Of those, 141 were NQ earrings, and 7 were +1 earrings. So the odds of getting a +1 earring from a NQ case is 4.965% or ~5%. I also looked at the tier of augment received on each earring using the NQ earrings only. They only have one augment, and it can 6,7,8,9, or 10. Here's the breakdown:

Total NQ cases: 148
Total NQ earrings: 141
Total +1 earrings: 7

Augments
"+6": 78 of the 141, or 55.3%
"+7": 28 of the 141, or 19.9%
"+8": 15 of the 141, or 10.6%
"+9": 10 of the 141, or 7.1%
"+10": 10 of the 141, or 7.1%

TLDR: What this means is the probability of just getting a max augmented +1 earring from a NQ case, (irrespective of its job) is = (0.05)(0.071)
so equals 3.5%! Dear lord....

I will be starting to save up my +1 old cases and repeat this to see how skewed the augment tier really is. Less than 1 in 10 odds to get a max augment on a NQ earring from NQ case and 3.5% chance of getting a +1 earring with max augments from a NQ case. Gimme a break.

[+]

warmech

Offline

Posts: 73

By warmech 2023-07-08 03:05:21

Link | Quote | Reply

just opened 175 and only saw 6 +1. all trash too :/

[+]

Onimaru

Offline

Posts: 81

By Onimaru 2023-07-08 08:46:48

Link | Quote | Reply

We are not the same.

[+]

Izanami

Offline

Posts: 308

By Izanami 2023-07-08 09:02:58

Link | Quote | Reply

Since each case opening is an independent event, we can collect unbiased data of case openings from all players by simply being near the player when they open their case. If we use Windower's logger, then we can analyze data collected since the release of Sortie.

I've convinced Andromida (Asura) to let me post his data here, which he conveniently provided to me in a fancy table format already (thanks). About half of this data was obtained with a character AFK near the Sortie entrance/exit around September 2022, which provided a quick way to obtain results since many players wait until immediately after their run to open cases.

Code

Old Box 15489   NQ 14524   HQ 965

        | NQ Old Box                    | HQ Old Box
Job     | NQ#   NQ%     NQ+#    NQ+%    | HQ#   HQ%     HQ1+#   HQ1+%
BLM     |   772  5.3      38     0.3    |   58   6.0       3     0.3
BLU     |   597  4.1      22     0.2    |   34   3.5       1     0.1
BRD     |   594  4.1      28     0.2    |   45   4.7       2     0.2
BST     |   585  4.0      25     0.2    |   31   3.2       1     0.1
COR     |   556  3.8      28     0.2    |   45   4.7       4     0.4
DNC     |   579  4.0      21     0.1    |   35   3.6       4     0.4
DRG     |   576  4.0      22     0.2    |   37   3.8       1     0.1
DRK     |   516  3.6      21     0.1    |   36   3.7       2     0.2
GEO     |   560  3.9      23     0.2    |   28   2.9       0     0.0
MNK     |   770  5.3      24     0.2    |   53   5.5       2     0.2
NIN     |   615  4.2      19     0.1    |   42   4.4       1     0.1
PLD     |   610  4.2      27     0.2    |   42   4.4       1     0.1
PUP     |   587  4.0      21     0.1    |   40   4.1       2     0.2
RDM     |   787  5.4      29     0.2    |   54   5.6       2     0.2
RNG     |   581  4.0      23     0.2    |   37   3.8       1     0.1
RUN     |   556  3.8      24     0.2    |   37   3.8       1     0.1
SAM     |   611  4.2      27     0.2    |   42   4.4       1     0.1
SCH     |   574  4.0      32     0.2    |   43   4.5       1     0.1
SMN     |   559  3.8      22     0.2    |   27   2.8       1     0.1
THF     |   764  5.3      42     0.3    |   45   4.7       2     0.2
WAR     |   783  5.4      38     0.3    |   51   5.3       3     0.3
WHM     |   805  5.5      31     0.2    |   61   6.3       6     0.6

Total   | 13937 96.0     587     4.0    |  923  95.6      42     4.4

How do I read the data table?

What does this data tell us? (Results)

The rate of obtaining an upgraded earring from a case is roughly 4.0%. This applies for obtaining "+1" earrings from NQ cases and "+2" earrings from "+1" cases.
No "+2" earrings were obtained from NQ cases
The job distribution is roughly evenly spread, but some jobs appear to be programmatically more likely. As a first assumption, we expect each job to have a 1/22 (4.55%) chance of showing up, but the observed data deviates from simulations with this assumption (see below). In other words: there's a chance that some job earrings are actually more common than others.
An interesting note from Andromida's analysis of his data: "I believe SE are using a random number between 0 and 255 which each job being given 11 numbers with 6 numbers remaining which appear to be allotted to WHM BLM MNK WAR RDM THF so most jobs have 11 in 255 chance but those jobs have 12 in 255 chance. This is speculation but apears to match the numbers". Which suggests that the original 6 jobs have a higher probability of occurring and is exactly what we're seeing in the data.
The chance of obtaining a specific "Job +2" earring from a "+1" case is roughly (4.0%)*(4.55%) = 0.18%.
Assuming a Binomial distribution and an equal chance for obtaining each job (which is clearly incorrect): If you opened 50 "+1" cases, then you have roughly a 90% chance to see at least one specific "Job +1" earring and roughly 9% chance to see at least one specific "Job +2" earring.

I wrote a code to open 14524 cases (the number of NQ cases in the sample above) 5000 times to check the expected per-job numbers, assuming each job is equally likely to occur (but see Andromida's note about the 6 original jobs potentially being more probable). Using the 84th and 16th percentiles as the 1-sigma 68% confidence region, I find that each job should see 633+-25 (with 3-sigma 99.9% confidence range of 567-708) NQ earrings and 26+6-5 (with 3-sigma 99.9% confidence range of 13-43) "+1" earrings, which is not what we're seeing in the sample of 14524 actual case openings. Some jobs are significantly outside of this range, such as WHM with 805 NQ earrings and DRK with only 516. The HQ job distribution agrees well with the 3-sigma expectation, but the sample size is much smaller on a per-job basis.

The simulations also suggest an upgraded earring probability of 3.9% +- 0.7%, which agrees well with the 4.0% we're seeing in the data, but also allows for a 4.5% chance of obtaining an upgraded earring. This would be an interesting situation if we consider how the FFXI code might handle providing upgraded earrings. Perhaps each case rolls two jobs with a 4.5% chance of any one job occuring. The resulting earring is then the first/second roll's job, which is upgraded if both rolls are the same job. If it had three job rolls, then we could argue that the chance of a +2 earring from a NQ case should be (1/22)^3 = 4.5% * 4.5% * 4.5% = 0.009%. This is just interesting speculation and was not observed in the sample of 14524 NQ cases.

What does this data not tell us?

How can you check your own log files?

[+]

Bismarck.Zattano

Offline

Server: Bismarck

Game: FFXI

user: nvasquez89502

Posts: 20

By Bismarck.Zattano 2023-07-08 12:50:02

Link | Quote | Reply

Izanami said: »

Since each case opening is an independent event, we can collect unbiased data of case openings from all players by simply being near the player when they open their case. If we use Windower's logger, then we can analyze data collected since the release of Sortie.

I've convinced Andromida (Asura) to let me post his data here, which he conveniently provided to me in a fancy table format already (thanks). About half of this data was obtained with a character AFK near the Sortie entrance/exit around September 2022, which provided a quick way to obtain results since many players wait until immediately after their run to open cases.

Code

Old Box 15489   NQ 14524   HQ 965

        | NQ Old Box                    | HQ Old Box
Job     | NQ#   NQ%     NQ+#    NQ+%    | HQ#   HQ%     HQ1+#   HQ1+%
BLM     |   772  5.3      38     0.3    |   58   6.0       3     0.3
BLU     |   597  4.1      22     0.2    |   34   3.5       1     0.1
BRD     |   594  4.1      28     0.2    |   45   4.7       2     0.2
BST     |   585  4.0      25     0.2    |   31   3.2       1     0.1
COR     |   556  3.8      28     0.2    |   45   4.7       4     0.4
DNC     |   579  4.0      21     0.1    |   35   3.6       4     0.4
DRG     |   576  4.0      22     0.2    |   37   3.8       1     0.1
DRK     |   516  3.6      21     0.1    |   36   3.7       2     0.2
GEO     |   560  3.9      23     0.2    |   28   2.9       0     0.0
MNK     |   770  5.3      24     0.2    |   53   5.5       2     0.2
NIN     |   615  4.2      19     0.1    |   42   4.4       1     0.1
PLD     |   610  4.2      27     0.2    |   42   4.4       1     0.1
PUP     |   587  4.0      21     0.1    |   40   4.1       2     0.2
RDM     |   787  5.4      29     0.2    |   54   5.6       2     0.2
RNG     |   581  4.0      23     0.2    |   37   3.8       1     0.1
RUN     |   556  3.8      24     0.2    |   37   3.8       1     0.1
SAM     |   611  4.2      27     0.2    |   42   4.4       1     0.1
SCH     |   574  4.0      32     0.2    |   43   4.5       1     0.1
SMN     |   559  3.8      22     0.2    |   27   2.8       1     0.1
THF     |   764  5.3      42     0.3    |   45   4.7       2     0.2
WAR     |   783  5.4      38     0.3    |   51   5.3       3     0.3
WHM     |   805  5.5      31     0.2    |   61   6.3       6     0.6

Total   | 13937 96.0     587     4.0    |  923  95.6      42     4.4

How do I read the data table?

What does this data tell us? (Results)

The rate of obtaining an upgraded earring from a case is roughly 4.0%. This applies for obtaining "+1" earrings from NQ cases and "+2" earrings from "+1" cases.
No "+2" earrings were obtained from NQ cases
The job distribution is roughly evenly spread, but some jobs appear to be programmatically more likely. As a first assumption, we expect each job to have a 1/22 (4.55%) chance of showing up, but the observed data deviates from simulations with this assumption (see below). In other words: there's a chance that some job earrings are actually more common than others.
An interesting note from Andromida's analysis of his data: "I believe SE are using a random number between 0 and 255 which each job being given 11 numbers with 6 numbers remaining which appear to be allotted to WHM BLM MNK WAR RDM THF so most jobs have 11 in 255 chance but those jobs have 12 in 255 chance. This is speculation but apears to match the numbers". Which suggests that the original 6 jobs have a higher probability of occurring and is exactly what we're seeing in the data.
The chance of obtaining a specific "Job +2" earring from a "+1" case is roughly (4.0%)*(4.55%) = 1.82%.
Assuming a Binomial distribution and an equal chance for obtaining each job: If you opened 50 "+1" cases, then you have roughly a 90% chance to see at least one specific "Job +1" earring and roughly 9% chance to see at least one specific "Job +2" earring.

I wrote a code to open 14524 cases (the number of NQ cases in the sample above) 5000 times to check the expected per-job numbers, assuming each job is equally likely to occur (but see Andromida's note about the 6 original jobs potentially being more probable). Using the 84th and 16th percentiles as the 1-sigma 68% confidence region, I find that each job should see 633+-25 (with 3-sigma 99.9% confidence range of 567-708) NQ earrings and 26+6-5 (with 3-sigma 99.9% confidence range of 13-43) "+1" earrings, which is not what we're seeing in the sample of 14524 actual case openings. Some jobs are significantly outside of this range, such as WHM with 805 NQ earrings and DRK with only 516. The HQ job distribution agrees well with the 3-sigma expectation, but the sample size is much smaller on a per-job basis.

The simulations also suggest an upgraded earring probability of 3.9% +- 0.7%, which agrees well with the 4.0% we're seeing in the data, but also allows for a 4.5% chance of obtaining an upgraded earring. This would be an interesting situation if we consider how the FFXI code might handle providing upgraded earrings. Perhaps each case rolls two jobs with a 4.5% chance of any one job occuring. The resulting earring is then the first/second roll's job, which is upgraded if both rolls are the same job. If it had three job rolls, then we could argue that the chance of a +2 earring from a NQ case should be (1/22)^3 = 4.5% * 4.5% * 4.5% = 0.09%. This is just interesting speculation and was not observed in the sample of 14524 NQ cases.

What does this data not tell us?

How can you check your own log files?

Great job on this data collect, but :Cringe: at the results.
Fortunately, for me, I have given up on +2 earrings. Hopefully one will come for my mains while making primes, but not counting on it.

On the other hand, I have noticed with my group's augment results on ??? Earring:"Job", they tend to favor low rolls. Considering how difficult it is to attain the +2's, its going to be disappointing to finally attain a +2 for a desired job, yet have the rolls be absolute low(like my MNK+2...) Hopefully some change to at least the augment rate in the near future will occur.

Taint

Offline

Posts: 3543

By Taint 2023-07-08 12:55:33

Link | Quote | Reply

KujahFoxfire said: »

Some stats to add to the pile in here.

Since 1st June my groups had 10x +1 Cases, of which we've produced about 3-4 +2 earrings.

Every one of them has minimum rolled. I personally have 5x +2 earrings since sortie released, all minimum rolled.

Out latest 10 came from:

E Boss, G Boss, G Boss, A Boss, Blue Chest in B, H Boss, H Boss, G Boss, B Boss, G Boss.

Would guess there's a higher drop rate from basement bosses than anything else, but also the weighting/distribution on the augs is totally skewed unless there's some kind of trick to rolling them decently like opening them on the right job or something.

Basement chest/bosses are the best source. Out static is on a break this week and doing low man chest runs we've gotten 10 +1 cases. Only boss we killed was E once.

[+]

Bahamut.Kaius

Offline

Server: Bahamut

Game: FFXI

user: endoplatypus

Posts: 26

By Bahamut.Kaius 2023-07-08 14:42:28

Link | Quote | Reply

I have acquired ~4,650,000 muffins since Sortie was released. As you can imagine that is a great many chests, cases, and +1 cases, and a whole lot of basement boss kills.

I have 0 +2 earrings so I definitely endorse this thread.

The best was when some dude in LS returns to game after an extended break, does like 5 sorties, gets a +2 and posts on discord "I FINALLY got +2 guys!!"

[+]

Carbuncle.Nynja

Online

Server: Carbuncle

Game: FFXI

user: NynJa

Posts: 3860

By Carbuncle.Nynja 2023-07-08 15:03:41

Link | Quote | Reply

say it with me:

BA
RA
SU

Taint

Offline

Posts: 3543

By Taint 2023-07-08 15:14:04

Link | Quote | Reply

Bahamut.Kaius said: »

I have acquired ~4,650,000 muffins since Sortie was released. As you can imagine that is a great many chests, cases, and +1 cases, and a whole lot of basement boss kills.

I have 0 +2 earrings so I definitely endorse this thread.

The best was when some dude in LS returns to game after an extended break, does like 5 sorties, gets a +2 and posts on discord "I FINALLY got +2 guys!!"

We took a newly returned BRD with us last night, he got his first +1 case ever and got DRG+2 earring. Pretty awesome luck.

Asura.Illuminate

Offline

Server: Asura

Game: FFXI

user: ThordainSylph

Posts: 145

By Asura.Illuminate 2023-07-09 03:05:32

Link | Quote | Reply

Izanami said: »

Since each case opening is an independent event, we can collect unbiased data of case openings from all players by simply being near the player when they open their case. If we use Windower's logger, then we can analyze data collected since the release of Sortie.

I've convinced Andromida (Asura) to let me post his data here, which he conveniently provided to me in a fancy table format already (thanks). About half of this data was obtained with a character AFK near the Sortie entrance/exit around September 2022, which provided a quick way to obtain results since many players wait until immediately after their run to open cases.

Code

Old Box 15489   NQ 14524   HQ 965

        | NQ Old Box                    | HQ Old Box
Job     | NQ#   NQ%     NQ+#    NQ+%    | HQ#   HQ%     HQ1+#   HQ1+%
BLM     |   772  5.3      38     0.3    |   58   6.0       3     0.3
BLU     |   597  4.1      22     0.2    |   34   3.5       1     0.1
BRD     |   594  4.1      28     0.2    |   45   4.7       2     0.2
BST     |   585  4.0      25     0.2    |   31   3.2       1     0.1
COR     |   556  3.8      28     0.2    |   45   4.7       4     0.4
DNC     |   579  4.0      21     0.1    |   35   3.6       4     0.4
DRG     |   576  4.0      22     0.2    |   37   3.8       1     0.1
DRK     |   516  3.6      21     0.1    |   36   3.7       2     0.2
GEO     |   560  3.9      23     0.2    |   28   2.9       0     0.0
MNK     |   770  5.3      24     0.2    |   53   5.5       2     0.2
NIN     |   615  4.2      19     0.1    |   42   4.4       1     0.1
PLD     |   610  4.2      27     0.2    |   42   4.4       1     0.1
PUP     |   587  4.0      21     0.1    |   40   4.1       2     0.2
RDM     |   787  5.4      29     0.2    |   54   5.6       2     0.2
RNG     |   581  4.0      23     0.2    |   37   3.8       1     0.1
RUN     |   556  3.8      24     0.2    |   37   3.8       1     0.1
SAM     |   611  4.2      27     0.2    |   42   4.4       1     0.1
SCH     |   574  4.0      32     0.2    |   43   4.5       1     0.1
SMN     |   559  3.8      22     0.2    |   27   2.8       1     0.1
THF     |   764  5.3      42     0.3    |   45   4.7       2     0.2
WAR     |   783  5.4      38     0.3    |   51   5.3       3     0.3
WHM     |   805  5.5      31     0.2    |   61   6.3       6     0.6

Total   | 13937 96.0     587     4.0    |  923  95.6      42     4.4

How do I read the data table?

What does this data tell us? (Results)

The rate of obtaining an upgraded earring from a case is roughly 4.0%. This applies for obtaining "+1" earrings from NQ cases and "+2" earrings from "+1" cases.
No "+2" earrings were obtained from NQ cases
The job distribution is roughly evenly spread, but some jobs appear to be programmatically more likely. As a first assumption, we expect each job to have a 1/22 (4.55%) chance of showing up, but the observed data deviates from simulations with this assumption (see below). In other words: there's a chance that some job earrings are actually more common than others.
An interesting note from Andromida's analysis of his data: "I believe SE are using a random number between 0 and 255 which each job being given 11 numbers with 6 numbers remaining which appear to be allotted to WHM BLM MNK WAR RDM THF so most jobs have 11 in 255 chance but those jobs have 12 in 255 chance. This is speculation but apears to match the numbers". Which suggests that the original 6 jobs have a higher probability of occurring and is exactly what we're seeing in the data.
The chance of obtaining a specific "Job +2" earring from a "+1" case is roughly (4.0%)*(4.55%) = 1.82%.
Assuming a Binomial distribution and an equal chance for obtaining each job: If you opened 50 "+1" cases, then you have roughly a 90% chance to see at least one specific "Job +1" earring and roughly 9% chance to see at least one specific "Job +2" earring.

I wrote a code to open 14524 cases (the number of NQ cases in the sample above) 5000 times to check the expected per-job numbers, assuming each job is equally likely to occur (but see Andromida's note about the 6 original jobs potentially being more probable). Using the 84th and 16th percentiles as the 1-sigma 68% confidence region, I find that each job should see 633+-25 (with 3-sigma 99.9% confidence range of 567-708) NQ earrings and 26+6-5 (with 3-sigma 99.9% confidence range of 13-43) "+1" earrings, which is not what we're seeing in the sample of 14524 actual case openings. Some jobs are significantly outside of this range, such as WHM with 805 NQ earrings and DRK with only 516. The HQ job distribution agrees well with the 3-sigma expectation, but the sample size is much smaller on a per-job basis.

The simulations also suggest an upgraded earring probability of 3.9% +- 0.7%, which agrees well with the 4.0% we're seeing in the data, but also allows for a 4.5% chance of obtaining an upgraded earring. This would be an interesting situation if we consider how the FFXI code might handle providing upgraded earrings. Perhaps each case rolls two jobs with a 4.5% chance of any one job occuring. The resulting earring is then the first/second roll's job, which is upgraded if both rolls are the same job. If it had three job rolls, then we could argue that the chance of a +2 earring from a NQ case should be (1/22)^3 = 4.5% * 4.5% * 4.5% = 0.09%. This is just interesting speculation and was not observed in the sample of 14524 NQ cases.

What does this data not tell us?

How can you check your own log files?

This data is pretty interesting. I think assuming an equal probability of getting an earring for each specific job is incorrect. With a sample of over 13,000 and a range of 516 on DRK to 800 on WHM, this doesn't seem to have happened by chance alone. I can plug in the numbers on SPSS and do an ANOVA to get an exact probability that the variation in (an assumed) flat distribution could happen by chance.

[+]

Asura.Lunafreya

Offline

Server: Asura

Game: FFXI

user: Lunafreya

Posts: 732

guildwork

By Asura.Lunafreya 2023-07-09 08:24:20

Link | Quote | Reply

The sortie earring system is utterly broken and stupid even for SE/FFXI standards barring an exchange system eventually being implemented or galli can be used to buy +1 or , optimistically +2, cases.

Lakshmi.Cesil

Offline

Server: Lakshmi

Game: FFXI

user: Valkyrie

Posts: 560

By Lakshmi.Cesil 2023-07-09 09:22:25

Link | Quote | Reply

Asura.Lunafreya said: »

The sortie earring system is utterly broken and stupid even for SE/FFXI standards barring an exchange system eventually being implemented or galli can be used to buy +1 or , optimistically +2, cases.

Agreed! It's crazy :( made me take a really long break, hope they do something about it.

Antisense

Offline

By Antisense 2023-07-09 09:26:45

Link | Quote | Reply

Having the probability distribution null be like 1/18 for the original 6 jobs and 1/24 for the other 16 jobs doesn't result in a "badness" of fit of the above sample.

11x16 + 12x6 != 256 nor 255

The data aren't consistent with a uniform 22-nomial distribution in any case since the non-conforming bins would be for the original 6 + DRK.

[+]

Fenrir.Jinxs

Offline

Server: Fenrir

Game: FFXI

user: Jinxs

Posts: 540

By Fenrir.Jinxs 2023-07-09 10:01:09

Link | Quote | Reply

This is one of those things that we should brigade an official forum post. At the very least we need something to do with excess sapphires/starstones and nq/+1 earrings.
A trade in system would be great.

Asura.Eiryl

Offline

Server: Asura

Game: FFXI

user: Eiryl

By Asura.Eiryl 2023-07-09 10:11:21

Link | Quote | Reply

Shoulda done that ***8 months ago.

Asura.Geriond

Offline

Server: Asura

Game: FFXI

user: Gerion

Posts: 3184

By Asura.Geriond 2023-07-09 10:34:06

Link | Quote | Reply

Just a couple small math corrections.

Izanami said: »

Since each case opening is an independent event, we can collect unbiased data of case openings from all players by simply being near the player when they open their case. If we use Windower's logger, then we can analyze data collected since the release of Sortie.

I've convinced Andromida (Asura) to let me post his data here, which he conveniently provided to me in a fancy table format already (thanks). About half of this data was obtained with a character AFK near the Sortie entrance/exit around September 2022, which provided a quick way to obtain results since many players wait until immediately after their run to open cases.

[snip]

4% * 4.55% is 0.182%, not 1.82%
4.5% * 4.5% * 4.5% is 0.009%, not 0.09%.

Asura.Illuminate said: »

TLDR: What this means is the probability of just getting a max augmented +1 earring from a NQ case, (irrespective of its job) is = (0.05)(0.071)
so equals 3.5%! Dear lord....

(0.05)*(0.071) is actually 0.355%, even worse.

[+]

Izanami

Offline

Posts: 308

By Izanami 2023-07-09 10:45:52

Link | Quote | Reply

Asura.Illuminate said: »

This data is pretty interesting. I think assuming an equal probability of getting an earring for each specific job is incorrect. With a sample of over 13,000 and a range of 516 on DRK to 800 on WHM, this doesn't seem to have happened by chance alone. I can plug in the numbers on SPSS and do an ANOVA to get an exact probability that the variation in (an assumed) flat distribution could happen by chance.

You have the correct interpretation of the results. I simulated equal probability for each job exactly to demonstrate that the data does not follow such a distribution. It might seem obvious by glancing at the numbers, but it helps and is important to provide a baseline confidence interval to compare the data to. A uniform probability has been the assumed distribution, so it's best to compare against that. The average of the 6 base jobs is ~780, while the average of the 16 advanced jobs is ~580. This is a 34% increased drop rate for the 6 base jobs.

The data clearly shows that the 6 base jobs have significantly higher drop rates and that DRK is unusually low. However, we could argue that DRK is not a special case, and is instead simply a 3-sigma outlier in the distribution of the 16 advanced jobs. I can't imagine why SE would pick DRK, out of all of the advanced jobs, to reduce the rate of.

Hopefully somebody with a real statistics background can tease out the true probabilities from this data.

Asura.Geriond said: »

Just a couple small math corrections.
4% * 4.55% is 0.182%, not 1.82%
4.5% * 4.5% * 4.5% is 0.009%, not 0.09%.

Thanks. I've updated the post. I didn't even consider checking those numbers since I thought "just multiply by 100, you can't possibly mess it up."

[+]

Asura.Illuminate

Offline

Server: Asura

Game: FFXI

user: ThordainSylph

Posts: 145

By Asura.Illuminate 2023-07-09 10:55:47

Link | Quote | Reply

So, I made a bar chart (thank you Izanami and Andromida!) which started with the highest frequency of NQ earrings (WHM) from a NQ chest, all the way to the lowest frequency (DRK). This is absolutely NOT a flat distribution. I tried to find a free statistics program that would do an ANOVA of 22 different categories for the job variable and their corresponding frequencies/occurrences, but it is just too many categories for the job variable lol!

This confirms that there is a considerably higher distribution for the original FFXI jobs (WHM RDM WAR BLM MNK THF), as previously noted by Izanami, which shows a slow, steady decrease in frequency.

Then a significant drop by over 150 followed by a second distribution that is relatively flat but shows a slow, steady decrease in frequency for the rest of the jobs, with DRK being the lowest. DRK might be an outlier. But with a sample size of over 13000, it might just be what SE decided.

Basically what I'm getting at here is that if SE claimed this is a flat distribution they need to learn how to math. This agrees with Izanami's conclusion that there is a significantly greater chance of getting earrings on original FFXI jobs than the rest that were added later. Ofc this applies to only NQ earrings, but I would assume the same distribution might apply to the +1 earrings and the +2s even?

[+]

Asura.Illuminate

Offline

Server: Asura

Game: FFXI

user: ThordainSylph

Posts: 145

By Asura.Illuminate 2023-07-09 11:11:21

Link | Quote | Reply

Here is the +1 distribution using Izanami's/ Andromida's work. It also shows five of the original 6 jobs as the most common, followed by thief tied with two others for the 6th highest frequency.

This is a small sample size though (for this kind of analysis).

Asura.Illuminate

Offline

Server: Asura

Game: FFXI

user: ThordainSylph

Posts: 145

By Asura.Illuminate 2023-07-09 11:32:20

Link | Quote | Reply

Bismarck.Zattano said: »

Izanami said: »

Since each case opening is an independent event, we can collect unbiased data of case openings from all players by simply being near the player when they open their case. If we use Windower's logger, then we can analyze data collected since the release of Sortie.

I've convinced Andromida (Asura) to let me post his data here, which he conveniently provided to me in a fancy table format already (thanks). About half of this data was obtained with a character AFK near the Sortie entrance/exit around September 2022, which provided a quick way to obtain results since many players wait until immediately after their run to open cases.

Code

Old Box 15489   NQ 14524   HQ 965

        | NQ Old Box                    | HQ Old Box
Job     | NQ#   NQ%     NQ+#    NQ+%    | HQ#   HQ%     HQ1+#   HQ1+%
BLM     |   772  5.3      38     0.3    |   58   6.0       3     0.3
BLU     |   597  4.1      22     0.2    |   34   3.5       1     0.1
BRD     |   594  4.1      28     0.2    |   45   4.7       2     0.2
BST     |   585  4.0      25     0.2    |   31   3.2       1     0.1
COR     |   556  3.8      28     0.2    |   45   4.7       4     0.4
DNC     |   579  4.0      21     0.1    |   35   3.6       4     0.4
DRG     |   576  4.0      22     0.2    |   37   3.8       1     0.1
DRK     |   516  3.6      21     0.1    |   36   3.7       2     0.2
GEO     |   560  3.9      23     0.2    |   28   2.9       0     0.0
MNK     |   770  5.3      24     0.2    |   53   5.5       2     0.2
NIN     |   615  4.2      19     0.1    |   42   4.4       1     0.1
PLD     |   610  4.2      27     0.2    |   42   4.4       1     0.1
PUP     |   587  4.0      21     0.1    |   40   4.1       2     0.2
RDM     |   787  5.4      29     0.2    |   54   5.6       2     0.2
RNG     |   581  4.0      23     0.2    |   37   3.8       1     0.1
RUN     |   556  3.8      24     0.2    |   37   3.8       1     0.1
SAM     |   611  4.2      27     0.2    |   42   4.4       1     0.1
SCH     |   574  4.0      32     0.2    |   43   4.5       1     0.1
SMN     |   559  3.8      22     0.2    |   27   2.8       1     0.1
THF     |   764  5.3      42     0.3    |   45   4.7       2     0.2
WAR     |   783  5.4      38     0.3    |   51   5.3       3     0.3
WHM     |   805  5.5      31     0.2    |   61   6.3       6     0.6

Total   | 13937 96.0     587     4.0    |  923  95.6      42     4.4

How do I read the data table?

What does this data tell us? (Results)

The rate of obtaining an upgraded earring from a case is roughly 4.0%. This applies for obtaining "+1" earrings from NQ cases and "+2" earrings from "+1" cases.
No "+2" earrings were obtained from NQ cases
The job distribution is roughly evenly spread, but some jobs appear to be programmatically more likely. As a first assumption, we expect each job to have a 1/22 (4.55%) chance of showing up, but the observed data deviates from simulations with this assumption (see below). In other words: there's a chance that some job earrings are actually more common than others.
An interesting note from Andromida's analysis of his data: "I believe SE are using a random number between 0 and 255 which each job being given 11 numbers with 6 numbers remaining which appear to be allotted to WHM BLM MNK WAR RDM THF so most jobs have 11 in 255 chance but those jobs have 12 in 255 chance. This is speculation but apears to match the numbers". Which suggests that the original 6 jobs have a higher probability of occurring and is exactly what we're seeing in the data.
The chance of obtaining a specific "Job +2" earring from a "+1" case is roughly (4.0%)*(4.55%) = 1.82%.
Assuming a Binomial distribution and an equal chance for obtaining each job: If you opened 50 "+1" cases, then you have roughly a 90% chance to see at least one specific "Job +1" earring and roughly 9% chance to see at least one specific "Job +2" earring.

I wrote a code to open 14524 cases (the number of NQ cases in the sample above) 5000 times to check the expected per-job numbers, assuming each job is equally likely to occur (but see Andromida's note about the 6 original jobs potentially being more probable). Using the 84th and 16th percentiles as the 1-sigma 68% confidence region, I find that each job should see 633+-25 (with 3-sigma 99.9% confidence range of 567-708) NQ earrings and 26+6-5 (with 3-sigma 99.9% confidence range of 13-43) "+1" earrings, which is not what we're seeing in the sample of 14524 actual case openings. Some jobs are significantly outside of this range, such as WHM with 805 NQ earrings and DRK with only 516. The HQ job distribution agrees well with the 3-sigma expectation, but the sample size is much smaller on a per-job basis.

The simulations also suggest an upgraded earring probability of 3.9% +- 0.7%, which agrees well with the 4.0% we're seeing in the data, but also allows for a 4.5% chance of obtaining an upgraded earring. This would be an interesting situation if we consider how the FFXI code might handle providing upgraded earrings. Perhaps each case rolls two jobs with a 4.5% chance of any one job occuring. The resulting earring is then the first/second roll's job, which is upgraded if both rolls are the same job. If it had three job rolls, then we could argue that the chance of a +2 earring from a NQ case should be (1/22)^3 = 4.5% * 4.5% * 4.5% = 0.09%. This is just interesting speculation and was not observed in the sample of 14524 NQ cases.

What does this data not tell us?

How can you check your own log files?

Great job on this data collect, but :Cringe: at the results.
Fortunately, for me, I have given up on +2 earrings. Hopefully one will come for my mains while making primes, but not counting on it.

On the other hand, I have noticed with my group's augment results on ??? Earring:"Job", they tend to favor low rolls. Considering how difficult it is to attain the +2's, its going to be disappointing to finally attain a +2 for a desired job, yet have the rolls be absolute low(like my MNK+2...) Hopefully some change to at least the augment rate in the near future will occur.

Using a sample size of 141 NQ earrings, the relative %'s of augments from lowest tier to highest tier was: 55.3%, 19.9%, 10.6%, 7.1% and 7.1%. So a 55.3% chance you'll get the lowest tiered augment and a 7.1% chance you would get the max tiered augment. I don't have enough +1 earrings (yet) to see if those augments are ridiculously skewed towards the sh*tty end of the stick, as well.

Perhaps the augment tier is a ratio, 8:3:1.5:1:1

From:
To:
Subject:
Body: