lillybit wrote: »

virtus753 wrote: »

Yes, I filed a bug report the first week regarding the zero tempers in blacksmithing, clothing, and jewelrycrafting.

I also filed a bug report this week because the updated tooltip still makes no sense for a crafter with maxed extraction passives. It continues to refer to “more powerful” tempers of each category, but there are no more powerful tempers than gold, which can already be unlocked of course with 3/3 in the extraction skill point passives. Once that last level is unlocked in all four crafts, the node is worthless if its tooltip is to be believed.

Is this node supposed to give more of those tempers? Or is it just an alternative to using skill points? They really need to clarify this.

You never know, maybe it's for a future update when they plan on bringing out mats for mythic upgrades!

I wouldn’t be surprised to see that down the line, but any future update like that is well in the future. The node needs to be functional and its tooltip coherent in the next 23 days.

If this node is just going to be redundant on a toon with 3/3 extraction passives, at least for now, I’d be happy to put those 100 CP into something actually worthwhile. But we need to know either way.

commdt wrote: »

So it is released now and what does it actually do? Couldnt find any specific information

In theory... it adds a hidden +1 to all four extraction passives. So, if you're 2/3 in each, it would bump you up to 3/3. If you're already at 3/3 it should bump you to 4/3? Or, it doesn't do anything. It's really unclear at this point.

commdt wrote: »

Well, then it would be good if @ZOS_GinaBruno could clarify it a bit. Not the exact formula, but at least does it do anything if you already have all crafting passives maxed out? The problem is that it is slottable, so you must sacrifice something to use it

Earlier in this thread it was shown that it does add an effective level to 3/3 to make it about 4/3. More specifically, it increases temper yields by around 12.5% compared to 3/3 extraction alone:

https://forums.elderscrollsonline.com/en/discussion/562770/meticulous-disassembly-still-broken-in-6-3-4-for-jc-other-craft-not-tested

Gina already confirmed it’s meant to stack with maxed extraction passives in the patch notes for 6.3.4, the last incremental before Update 29 hit live. She explained that it would increase drop rates for tempers even above the 3/3 extraction rate, but she added that it wasn’t feasible to include a specific number as to how big an increase because that depends on how many extraction passives you have for each craft and differs depending on the color of the temper as well.

commdt wrote: »

Thanks! Now it is clear. Sorry, Im a returning player so I didnt pay attention to PTS cycle

No worries! I understand wanting clarification.

Here is the link to 6.3.4, if you want to see the original note:

https://forums.elderscrollsonline.com/en/discussion/562719/pts-patch-notes-v6-3-4

Fairly sure it's an unreleased Rush track.

tmbrinks wrote: »

Data was collected using 800k items, so 80k refines

Yes, there is still a large margin of error in the exact drop rate increase.

At 80k refines, that puts your margin of error at ~.34% meaning that it is still technically possible that Meticulous disassembly actually reduces the drop rates, based on the data you produced.

With respect, saying, "it increases drops by 12.5%" is a fantasy based on the data collected. Yes, that is within your margin of error, but saying it increases drops by 25% is also within your margin of error.

The mistake being made in this data is the assumption that 12.5% is a whole value. If something had a 12% drop rate, and you sampled 8k times, you could make a reasonable guess that it was around that point.

However, this is a stacked percentage. Meaning, it's a percentage of a percentage. In this case, a 12.5% increase would raise 0.5% to 0.625%. (I erroneously listed the drop rate of Chromium as 3%, that's not correct, that's the drop rate for Dreugh Wax, Tempering Alloy and Rosin.)

The illustration of the danger here is in your own data, when one of your control groups (with 2500 refinements) returned a drop rate for Chromium of 0.486% This does not mean that the drop rate for Chromium was reduced, however, it demonstrates the importance of accounting for margin of error. In this specific case the issue is that a population of 2500 has a nearly 2% margin of error, meaning, you could have seen Chromium with a 2% drop rate in that test and still been well within the margin of error.

So, looping back again, your 80k refines test returned a drop rate of 0.56%. Except, your margin of error was .34%. Meaning, based on your data, with Meticulous assembly slotted, we can only confirm that the drop rate of Chromium grains is somewhere between .22% and .9%.

Now, look, in most cases, having a margin of error of .34% would be amazing. However, when we're trying to assertion on exactly how much drop rates were adjusted, it starts to fall apart.

Even we step back and look at the most forgiving example, Terne, we're left still with an item that, by your own testing, runs around a 1.5% drop rate. This still suffers from the .34% margin of error. If the 12.5% number was accurate, then you'd still have a situation where your margin of error was high enough to conceal an actual drop rate decrease.

The problem here is that statistical margins of error are almost impossible to completely eliminate. Given the values are as small as they are, (a difference in drop rate between .2% and .05% depending on the upgrade material.) Even to simply lower the margin of error on Terne dust enough to conclusively demonstrate an increase, you'd need to refine something like 300k times (three million raw materials), to do the same (and get your margin of error under .6%, you'd need to refine three million times (thirty million raw materials.)

It's nice to be able to sit down and say, "yeah, I can prove it's this specific number... but you can't. we're still rolling dice, and when the base chances get this low, it gets really brutal to test. You need to throw a lot of dice.

So, no, it might be 12.5%, it could be 25%, or it could be no change whatsoever. There simply isn't enough here to state the change definitively.

We're talking about 2 different percentages and mixing them up.

My percentages are per material refined. So can multiply by 10 to get the per refinement rate, which is what you're looking at.

So the 0.34% margin of error , would be 5.6% +_ 0.34% (or ~5.3% to ~5.9%) which shows definitively that there is an increase over the old 5% drop rate.

Here's the analysis on a per refine basis



I'm very, very confident that the drop rate has changed

starkerealm wrote: »

tmbrinks wrote: »

I'm simply running a binomial distribution test on the data...

Translation: You took the part, divided it by the whole, multiplied the result by 100 (or, rather, let the cell formatting tool handle that last part), and... oh, right that's what we call a percent.

tmbrinks wrote: »

I've very clearly stated at some point that the 12.5% is my "best guess", and not that it's definitively that.

And I was responding to stating that value as fact, without any acknowledgement of it being speculative.

tmbrinks wrote: »

We're talking about 2 different percentages and mixing them up.

Not exactly. There are two relevant percentages. There's the percentage that the Meticulous Disassembly modifies (or is supposed to modify), the drop rate. So, when you add 12.5% to your drop chance, that means that there is now a 112.5% chance of gold items dropping relative to normal. However, the total percentage chance of that drop is much lower. The thing is, when trying to determine the former, the margin of error only cares about the latter.

tmbrinks wrote: »

My percentages are per material refined. So can multiply by 10 to get the per refinement rate, which is what you're looking at.

No, because if you did that, all of this information is utterly worthless.

There is a 0% chance of a single unrefined material ever producing an upgrade item in ESO. Not, it's low. Not, it's .5%. It is zero. This may come as a shock, but a single unit of raw materials cannot be refined at all. So, if we take your data at face value, we know it is false.

This is specifically taking your survey population, and then changing the number to modify your margin of error in order to lend your data more credibility. If you want that larger sample size, you need to collect the larger sample size.

So, instead basing your calculations on the number actual events you just added a zero to make it sound more legitimate... ugh.

tmbrinks wrote: »

Here's the analysis on a per refine basis



I'm very, very confident that the drop rate has changed

Okay, so now I have an additional methodological issue, and I'm not going to hammer on this.

I've been assuming, up to this point, that you've been basing this data off of Jewelry extraction on the PTS. This is because, the template characters have an unlimited amount of dust. But, of course, this data was accrued over time, from a live character.

Now, that data could be extremely useful... if you preserved the dates when those refinements took place, and didn't just file it all into a slurry. Because it could provide useful trending information (and yeah, at that point we just kinda live with a significant margin of error.)

Unfortunately, when we're looking at this right now... not so much.

It is. This is literally just data from 3/8 onward. There are separate tabs in the spreadsheet that show the date of the refines.

https://docs.google.com/spreadsheets/d/1yQDUjYNQVsIFl0ktkbkSlfYPkzP6pgCOTMQ-qCuzfaI/edit#gid=1892397659

here is the page for the blacksmithing data.

Please look at the entire document.