Deconned 10k of every raw resource for rate of gold mats

DocFrost72

I decon'd around 10k worth of every raw mat from my surveys over the last few months. Results:

45 rosin (10.7k raw ruby ash)
59 Tempering alloy (11.7k raw ore)
72 Dreugh Wax (6.3k raw silk and 5.9k raw hide)
57 Chromium Grain (9.5 raw dust)

rates:
wood: 4.205%
ore: 5.043%
cloth/hide: 5.902%
dust: 6% flat

The "expected" rate of refining 10 raw materials and finding a gold mat is around 5%, so I'd say that's accurate to my results! The slight variation of around one percent is not surprising given the "low" volume of instances. I did find out you can't refine more than 10,000 of a raw material at a time, however.

EdmondDontes

I recently ran this same exercise and came up with very close to the same results.

Xinihp

The real question however is whether it is better to refine one at a time or in bulk.

I wonder if internally the math.random generator is being seeded between each increment, or if it is seeded once for the entire refinement stack, in which case the potential to get a "bad roll" that affects the entire stack would be a problem.

Doing one at a time the numbers should always even out over time.

tmbrinks

Xinihp wrote: »

The real question however is whether it is better to refine one at a time or in bulk.

I wonder if internally the math.random generator is being seeded between each increment, or if it is seeded once for the entire refinement stack, in which case the potential to get a "bad roll" that affects the entire stack would be a problem.

Doing one at a time the numbers should always even out over time.

There is no difference. I have data between myself and others that have shared in the millions of mats linked in the spreadsheet below in my signature. There is pre-multicraft, post-multicraft, and post-meticulous disassembly data.

Each refine is still an independent event, it just does them x numbers at a time. Save your time, refine all at once.

phaneub17_ESO

Those of us with 18+ characters have the advantage its not refining that gets you the most gold mats its max rank crafting writ boxes. Jewelry writ boxes have a pretty good 1 in 5 chance at 1-2 gold grains and a low chance at a full gold plating.

I_killed_Vivec

tmbrinks wrote: »

Xinihp wrote: »

The real question however is whether it is better to refine one at a time or in bulk.

I wonder if internally the math.random generator is being seeded between each increment, or if it is seeded once for the entire refinement stack, in which case the potential to get a "bad roll" that affects the entire stack would be a problem.

Doing one at a time the numbers should always even out over time.

There is no difference. I have data between myself and others that have shared in the millions of mats linked in the spreadsheet below in my signature. There is pre-multicraft, post-multicraft, and post-meticulous disassembly data.

Each refine is still an independent event, it just does them x numbers at a time. Save your time, refine all at once.

But...

I like to do a "tester" to see how RND is behaving:

Refine 30, get nothing then I don't expect anything special.

Refine 200 more and get nothing... RNG might not be paying out today.

But if those first 30 give 2 dreugh wax then I'm suspicious... refine all I have (about 1K more) and get 15 more. RNG is paying out!!

The problem with statistical analysis is that millions of samples will iron out these discrepancies.

However, I'm convinced that it pays out like a fruit machine (automated gambling machine for non-UK pub goers). These have to pay out at least 75%, but 75% of what? Of ALL takings, not just yours. So if you expect 1 temper for every 150 raw mats and you get none then someone will be getting 2...

In all this excitement I've forgotten how much I'd refined.. so, you have to ask yourself one question: 'Do I feel lucky?' Well do ya, punk?

tmbrinks

I_killed_Vivec wrote: »

tmbrinks wrote: »

Xinihp wrote: »

The real question however is whether it is better to refine one at a time or in bulk.

I wonder if internally the math.random generator is being seeded between each increment, or if it is seeded once for the entire refinement stack, in which case the potential to get a "bad roll" that affects the entire stack would be a problem.

Doing one at a time the numbers should always even out over time.

There is no difference. I have data between myself and others that have shared in the millions of mats linked in the spreadsheet below in my signature. There is pre-multicraft, post-multicraft, and post-meticulous disassembly data.

Each refine is still an independent event, it just does them x numbers at a time. Save your time, refine all at once.

But...

I like to do a "tester" to see how RND is behaving:

Refine 30, get nothing then I don't expect anything special.

Refine 200 more and get nothing... RNG might not be paying out today.

But if those first 30 give 2 dreugh wax then I'm suspicious... refine all I have (about 1K more) and get 15 more. RNG is paying out!!

The problem with statistical analysis is that millions of samples will iron out these discrepancies.

However, I'm convinced that it pays out like a fruit machine (automated gambling machine for non-UK pub goers). These have to pay out at least 75%, but 75% of what? Of ALL takings, not just yours. So if you expect 1 temper for every 150 raw mats and you get none then someone will be getting 2...

In all this excitement I've forgotten how much I'd refined.. so, you have to ask yourself one question: 'Do I feel lucky?' Well do ya, punk?

Yes, large sample sizes irons out data gotten from small sample sizes, where randomness has an outsized effect on the outcome. Extrapolation of data from small samples is the most dangerous (wrong) thing done in math. I choose to interpolate the data by having large sample sizes and lots of data before making conclusions.

The math teacher in me is crying right now.

code65536

I_killed_Vivec wrote: »

The problem with statistical analysis is that millions of samples will iron out these discrepancies.

The problem with statistical are people who draw incorrect conclusions despite not having the millions of samples necessary in order to draw meaningful conclusions.

Drink deep, or taste not the Pierian Spring.

DMuehlhausen

Xinihp wrote: »

The real question however is whether it is better to refine one at a time or in bulk.

I wonder if internally the math.random generator is being seeded between each increment, or if it is seeded once for the entire refinement stack, in which case the potential to get a "bad roll" that affects the entire stack would be a problem.

Doing one at a time the numbers should always even out over time.

Because nothing man made is truly random it's best to do them in bulk. Have I seen the code no, but I can practically guarantee you that the more you each one slightly raises your chance at gold mats. It's most like the same on Crown Crates for everything, but specifically the Radiant mounts.

It's just Candy Crush style games. You get to a super hard level can't beat it for like a day. Then just suddenly you start it with zero boosts and boom everything is just exploding and clearing out. If they make it impossible or extremely hard for a person to get past it after so many times that person will quit playing. So you make it slightly easier every attempt and eventually they get by it.

DocFrost72

tmbrinks wrote: »

Xinihp wrote: »

The real question however is whether it is better to refine one at a time or in bulk.

I wonder if internally the math.random generator is being seeded between each increment, or if it is seeded once for the entire refinement stack, in which case the potential to get a "bad roll" that affects the entire stack would be a problem.

Doing one at a time the numbers should always even out over time.

There is no difference. I have data between myself and others that have shared in the millions of mats linked in the spreadsheet below in my signature. There is pre-multicraft, post-multicraft, and post-meticulous disassembly data.

Each refine is still an independent event, it just does them x numbers at a time. Save your time, refine all at once.

Very interesting, thank you for sharing!

I_killed_Vivec wrote: »

tmbrinks wrote: »

Xinihp wrote: »

The real question however is whether it is better to refine one at a time or in bulk.

I wonder if internally the math.random generator is being seeded between each increment, or if it is seeded once for the entire refinement stack, in which case the potential to get a "bad roll" that affects the entire stack would be a problem.

Doing one at a time the numbers should always even out over time.

There is no difference. I have data between myself and others that have shared in the millions of mats linked in the spreadsheet below in my signature. There is pre-multicraft, post-multicraft, and post-meticulous disassembly data.

Each refine is still an independent event, it just does them x numbers at a time. Save your time, refine all at once.

But...

I like to do a "tester" to see how RND is behaving:

Refine 30, get nothing then I don't expect anything special.

Refine 200 more and get nothing... RNG might not be paying out today.

But if those first 30 give 2 dreugh wax then I'm suspicious... refine all I have (about 1K more) and get 15 more. RNG is paying out!!

The problem with statistical analysis is that millions of samples will iron out these discrepancies.

However, I'm convinced that it pays out like a fruit machine (automated gambling machine for non-UK pub goers). These have to pay out at least 75%, but 75% of what? Of ALL takings, not just yours. So if you expect 1 temper for every 150 raw mats and you get none then someone will be getting 2...

In all this excitement I've forgotten how much I'd refined.. so, you have to ask yourself one question: 'Do I feel lucky?' Well do ya, punk?

I entirely get where you're coming from, but the issue is one of control. For sake of argument assume you are entirely correct, and that there would be a hypothetical "single" slot machine distributing gold materials.

Even then, you're not going to be able to pick your spot in line. Best you can do is try to find a pattern to access the machine with, but it would be akin to being in a pub with millions of people all chaotically plating the single slot machine at once. Even if your read of loot distribution were correct (and I'm not convinced it is one slot machine, but countless slot machines you never get to "choose"), you have no way of controlling for your input in any meaningful manner. In essence, even under how I read your interpretation, your pulls are still random due to interference by other factors (players).

We go right back to aggregated data, which is why that is typically sought after in these sorts of equations.

Fun thought experiment though.

Paulytnz

phaneub17_ESO wrote: »

Those of us with 18+ characters have the advantage its not refining that gets you the most gold mats its max rank crafting writ boxes. Jewelry writ boxes have a pretty good 1 in 5 chance at 1-2 gold grains and a low chance at a full gold plating.

To add to this I also find if all your crafters have the daily mail rewards coming in too, that is also a good reliable source for the rare mats vs refining.

I_killed_Vivec

DocFrost72 wrote: »

I entirely get where you're coming from, but the issue is one of control. For sake of argument assume you are entirely correct, and that there would be a hypothetical "single" slot machine distributing gold materials.

Even then, you're not going to be able to pick your spot in line. Best you can do is try to find a pattern to access the machine with, but it would be akin to being in a pub with millions of people all chaotically plating the single slot machine at once. Even if your read of loot distribution were correct (and I'm not convinced it is one slot machine, but countless slot machines you never get to "choose"), you have no way of controlling for your input in any meaningful manner. In essence, even under how I read your interpretation, your pulls are still random due to interference by other factors (players).

We go right back to aggregated data, which is why that is typically sought after in these sorts of equations.

Fun thought experiment though.

Oh, I'm not saying that the fruit machine is a good analogy, just that sometimes it seems like it is.

I have had bad experiences refining mats, when a couple of thousand just disappear at the click of a mouse and you don't even get a single gold temper.

I have also had good experiences, where refining 30 mats do give a gold temper and then refining 1K gives 15 more.

So it looks like at times it's ready to pay out (maybe a lot of people have been refining mats) and at others it isn't...

Now there's obvious confirmation bias, you do tend to remember the extremes - "nothing for 1K mats?", "15 for 1K mats!". Also if you don't think it's going to pay out and walk away you will never know if it was about to pay out...

Given the way RNG works in this game I'm sure they will just roll dice for every decision (refining each batch of 10 mats), making each decision independent. By far the simplest mechanism, and probably the fairest, even though when you're at the wrong end of the RNG stick it doesn't seem like it...

Amottica

I_killed_Vivec wrote: »

tmbrinks wrote: »

Xinihp wrote: »

The real question however is whether it is better to refine one at a time or in bulk.

I wonder if internally the math.random generator is being seeded between each increment, or if it is seeded once for the entire refinement stack, in which case the potential to get a "bad roll" that affects the entire stack would be a problem.

Doing one at a time the numbers should always even out over time.

There is no difference. I have data between myself and others that have shared in the millions of mats linked in the spreadsheet below in my signature. There is pre-multicraft, post-multicraft, and post-meticulous disassembly data.

Each refine is still an independent event, it just does them x numbers at a time. Save your time, refine all at once.

But...

I like to do a "tester" to see how RND is behaving:

Refine 30, get nothing then I don't expect anything special.

Refine 200 more and get nothing... RNG might not be paying out today.

But if those first 30 give 2 dreugh wax then I'm suspicious... refine all I have (about 1K more) and get 15 more. RNG is paying out!!

The problem with statistical analysis is that millions of samples will iron out these discrepancies.

However, I'm convinced that it pays out like a fruit machine (automated gambling machine for non-UK pub goers). These have to pay out at least 75%, but 75% of what? Of ALL takings, not just yours. So if you expect 1 temper for every 150 raw mats and you get none then someone will be getting 2...

In all this excitement I've forgotten how much I'd refined.. so, you have to ask yourself one question: 'Do I feel lucky?' Well do ya, punk?

all good. However, @tmbrinks has significant data to demonstrate actual drop rates vs what can happen with refining just a few. The big picture tells us what the drop rate is and if anything changes when a new patch drops.

HumbleThaumaturge

DocFrost72 wrote: »

I decon'd around 10k worth of every raw mat from my surveys over the last few months. Results:

45 rosin (10.7k raw ruby ash)
59 Tempering alloy (11.7k raw ore)
72 Dreugh Wax (6.3k raw silk and 5.9k raw hide)
57 Chromium Grain (9.5 raw dust)

rates:
wood: 4.205%
ore: 5.043%
cloth/hide: 5.902%
dust: 6% flat

The "expected" rate of refining 10 raw materials and finding a gold mat is around 5%, so I'd say that's accurate to my results! The slight variation of around one percent is not surprising given the "low" volume of instances. I did find out you can't refine more than 10,000 of a raw material at a time, however.

Your "percent" numbers are off (high) by a factor of 10. For example, 45 divided by 10700 is 0.004205, which is 0.4205%, not 4.205%. So your rates should be: (1) Wood: 0.4205%; (2) Ore: 0.5042%; (3) Cloth/Hide: 0.5902%; and (4) Dust: 0.6000%.

But your raw numbers give us a good glimpse of the gold-mat drop rate. As you say, you would need a much higher raw-mats sample size to get closer to the actual number(s).

Lots of folks in these Forums have conducted similar "experiments." Check them out. A guy with a forum name of tmbrinks once posted a spreadsheet.

My own personal experience without CP buff is a drop rate of 1 gold-mat per 200 raw mats (0.5 % drop rate), consistent with your raw numbers. With the CP passive "Meticulous Disassembly" slotted, the drop rate may be around 0.55% or one gold-mat per 182 raw mats. You did not mention if you had "Meticulous Disassembly" slotted or not.

tmbrinks

HumbleThaumaturge wrote: »

DocFrost72 wrote: »

I decon'd around 10k worth of every raw mat from my surveys over the last few months. Results:

45 rosin (10.7k raw ruby ash)
59 Tempering alloy (11.7k raw ore)
72 Dreugh Wax (6.3k raw silk and 5.9k raw hide)
57 Chromium Grain (9.5 raw dust)

rates:
wood: 4.205%
ore: 5.043%
cloth/hide: 5.902%
dust: 6% flat

The "expected" rate of refining 10 raw materials and finding a gold mat is around 5%, so I'd say that's accurate to my results! The slight variation of around one percent is not surprising given the "low" volume of instances. I did find out you can't refine more than 10,000 of a raw material at a time, however.

Your "percent" numbers are off (high) by a factor of 10. For example, 45 divided by 10700 is 0.004205, which is 0.4205%, not 4.205%. So your rates should be: (1) Wood: 0.4205%; (2) Ore: 0.5042%; (3) Cloth/Hide: 0.5902%; and (4) Dust: 0.6000%.

But your raw numbers give us a good glimpse of the gold-mat drop rate. As you say, you would need a much higher raw-mats sample size to get closer to the actual number(s).

Lots of folks in these Forums have conducted similar "experiments." Check them out. A guy with a forum name of tmbrinks once posted a spreadsheet.

My own personal experience without CP buff is a drop rate of 1 gold-mat per 200 raw mats (0.5 % drop rate), consistent with your raw numbers. With the CP passive "Meticulous Disassembly" slotted, the drop rate may be around 0.55% or one gold-mat per 182 raw mats. You did not mention if you had "Meticulous Disassembly" slotted or not.

It just depends on if you look at the percentages with respect to "each item" or "each refine". Since you can only refine in multiples of 10, it is reasonable to look at it that way as well. (I went back and forth on my spreadsheet on which way to report it)

Spreadsheet is still up to date and linked in my sig!

YandereGirlfriend

phaneub17_ESO wrote: »

Those of us with 18+ characters have the advantage its not refining that gets you the most gold mats its max rank crafting writ boxes. Jewelry writ boxes have a pretty good 1 in 5 chance at 1-2 gold grains and a low chance at a full gold plating.

This is the way. Plus all the surveys that you get after after a few months.

It's tedious but having 10-15 maps for every zone in the game every 3 months yields a TON of refining materials.

DocFrost72

HumbleThaumaturge wrote: »

DocFrost72 wrote: »

I decon'd around 10k worth of every raw mat from my surveys over the last few months. Results:

45 rosin (10.7k raw ruby ash)
59 Tempering alloy (11.7k raw ore)
72 Dreugh Wax (6.3k raw silk and 5.9k raw hide)
57 Chromium Grain (9.5 raw dust)

rates:
wood: 4.205%
ore: 5.043%
cloth/hide: 5.902%
dust: 6% flat

The "expected" rate of refining 10 raw materials and finding a gold mat is around 5%, so I'd say that's accurate to my results! The slight variation of around one percent is not surprising given the "low" volume of instances. I did find out you can't refine more than 10,000 of a raw material at a time, however.

Your "percent" numbers are off (high) by a factor of 10. For example, 45 divided by 10700 is 0.004205, which is 0.4205%, not 4.205%. So your rates should be: (1) Wood: 0.4205%; (2) Ore: 0.5042%; (3) Cloth/Hide: 0.5902%; and (4) Dust: 0.6000%.

But your raw numbers give us a good glimpse of the gold-mat drop rate. As you say, you would need a much higher raw-mats sample size to get closer to the actual number(s).

Lots of folks in these Forums have conducted similar "experiments." Check them out. A guy with a forum name of tmbrinks once posted a spreadsheet.

My own personal experience without CP buff is a drop rate of 1 gold-mat per 200 raw mats (0.5 % drop rate), consistent with your raw numbers. With the CP passive "Meticulous Disassembly" slotted, the drop rate may be around 0.55% or one gold-mat per 182 raw mats. You did not mention if you had "Meticulous Disassembly" slotted or not.

The percentages are based on the baseline of 1 refine, not one mat. It was an intentional and conscious decision based on the impossibility of refining just one raw item.

And yes, tmbrinks is here in the comments! He was able to confirm there's no difference between lump refining and refining individually through his work, which is very informative.

The character in question does have meticulous disassembly. Some research into tmbrink's other works shows that does increase the chance of getting higher level mats too!

Jayman1000

DocFrost72 wrote: »

I decon'd around 10k worth of every raw mat from my surveys over the last few months. Results:

45 rosin (10.7k raw ruby ash)
59 Tempering alloy (11.7k raw ore)
72 Dreugh Wax (6.3k raw silk and 5.9k raw hide)
57 Chromium Grain (9.5 raw dust)

rates:
wood: 4.205%
ore: 5.043%
cloth/hide: 5.902%
dust: 6% flat

The "expected" rate of refining 10 raw materials and finding a gold mat is around 5%, so I'd say that's accurate to my results! The slight variation of around one percent is not surprising given the "low" volume of instances. I did find out you can't refine more than 10,000 of a raw material at a time, however.

And interestingly 5% is exactly one golden mat per stack of 200 It all fits: 1 stack = 1 golden.

Gratz on the rolls also, you were lucky to get the higher ratio for the more expensive mats