Torch Drops (Gonna Lose My S*** Soon)

I've done ubers 10 times with my Paladin so far. Torches: 6 Necro, 2 Assassin, 1 Druid, 1 Sorc.

Yeah 6 Necromancer torches in 10 attempts. This should be a freakin' joke. Does MF somehow affect these drops? My uber-run gear is focused on skill and resistance boosts instead of MF.

Nope.

Even if it did, MF would only affect the "unique or not" (which it doesn't as torch is always guaranteed) and not the class or stats it rolls.

Pure dumb luck. Or lack thereof I guess in your case.

Look at it this way...if you hang in there long enough, eventually you'll get a string of 6 pally/sorc torches instead of 6 necro. Eventually...

Now find some necromancer players who have gotten 6 paladin torches

For me it was druid torches at first but it evens out over time. I think out of the last 10 ubers I ran, I got 5 sorc torches.

Here's my Hypothesis.

The problem with drop rates of items in D2r is that rng isn't completely random or magic find percentages wouldn't work. Every class of items in the game have a different range of possible numbers that can be generated. A tiered set of numbers determine what class of item is dropped such as a sword or

Helm or whatever, and then another set will determine it's quality. The higher the item quality or desirability the higher the range to make it more rarefied. Magic find for instance can alter the rate of rng by changing the range of possible numbers generated. This means Blizzard can quickly change the drop rate of loot by altering the range of all or some of the items. Magic Find weekend demonstrated just how quickly and easily Blizzard can adjust drop rate. If we took a poll on

Hellfire Torch drop rate, I wouldn't at all be surprised if Paladin and Sorceress Torches drop at a much lower rate than Druid, Barbarian, or Necromancer torches. Since a torch drops 100% of the time at the finish of an Uber run, all that's left is class, quality, and range which is adjustable.

I've done Ubers 19 times. I've got 5 Assassin, 3 Druid, 3 Necromancer, 3 Paladin, 3 Sorceress, and 2 Amazon torches.
First I needed for my Sorc but I got one only after about 5-6 tries.
Now I saw that I don't have one for Barbarian. I don't need one because I don't have any Barb character, but the stats are weird. 19 times, never dropped one.

You need to do thousands of Uber runs before the drops between classes would become even(ish). That's just rng and statistics.

Alvin did a thousand on single player and uploaded the results to a spreadsheet, link in the video. Hilariously Druid got the lowest amount ;0



Sometimes it's better to just buy the one you are after (assuming you have the runes).

But I'm with mhlg regarding drops. I got an obscene amount of essences, keys, unique items first few weeks of ladder. Now nada.

Florian_G wrote: 7 months ago

I've done Ubers 19 times. I've got 5 Assassin, 3 Druid, 3 Necromancer, 3 Paladin, 3 Sorceress, and 2 Amazon torches.
First I needed for my Sorc but I got one only after about 5-6 tries.
Now I saw that I don't have one for Barbarian. I don't need one because I don't have any Barb character, but the stats are weird. 19 times, never dropped one.

Odd of this happening are ~3.1%. When you look at the odds of a

Jah rune dropping in P1 non-TZ Cows is ~0.000000654%. Doesn't make 3.1% look so crazy anymore.

M1ck wrote: 7 months ago

You need to do thousands of Uber runs before the drops between classes would become even(ish). That's just rng and statistics.

Alvin did a thousand and uploaded the results to a spreadsheet, link in the video. Hilariously Druid got the lowest amount ;0



Sometimes it's better to just buy the one you are after (assuming you have the runes).

But I'm with mhlg regarding drops. I got an obscene amount of essences, keys, unique items first few weeks of ladder. Now nada.

The larger a sample size the more statistically accurate a result, but that doesn't mean that smaller sample sizes are completely inaccurate, because even with smaller samples a trend will emerge. For instance, if we were to collect the Uber drop data of several players and then continue to do so over time, the result would become increasingly more accurate. An issue I see, is that there are a lot of D2/r expert sites that post data on item drop rates without revealing how they arrived to their conclusion. Without methodology how do we know that expert site information is accurate? That leaves us with at best empirical data, which is usually a good start in forming a hypothesis but not in itself a result. One way to gather data is tolook at forum trade data, but even that is somewhat anecdotal. because which

Hellfire Torch's are more likely to be traded and which kept?

Knappogue wrote: 7 months ago

Florian_G wrote: 7 months ago

I've done Ubers 19 times. I've got 5 Assassin, 3 Druid, 3 Necromancer, 3 Paladin, 3 Sorceress, and 2 Amazon torches.
First I needed for my Sorc but I got one only after about 5-6 tries.
Now I saw that I don't have one for Barbarian. I don't need one because I don't have any Barb character, but the stats are weird. 19 times, never dropped one.

Odd of this happening are ~3.1%. When you look at the odds of a

Jah rune dropping in P1 non-TZ Cows is ~0.000000654%. Doesn't make 3.1% look so crazy anymore.

I found only one

Jah on TZ P8, after 8 months of playing

Florian_G wrote: 7 months ago

Knappogue wrote: 7 months ago

Florian_G wrote: 7 months ago

I've done Ubers 19 times. I've got 5 Assassin, 3 Druid, 3 Necromancer, 3 Paladin, 3 Sorceress, and 2 Amazon torches.
First I needed for my Sorc but I got one only after about 5-6 tries.
Now I saw that I don't have one for Barbarian. I don't need one because I don't have any Barb character, but the stats are weird. 19 times, never dropped one.

Odd of this happening are ~3.1%. When you look at the odds of a

Jah rune dropping in P1 non-TZ Cows is ~0.000000654%. Doesn't make 3.1% look so crazy anymore.

I found only one

Jah on TZ P8, after 8 months of playing

I understand your struggle. I have found only one

Jah and it was P1 Non-TZ cows within my first 2 weeks of D2R. Have since found lots of

Ber and below, but no more

Jah

mhlg wrote: 7 months ago

The larger a sample size the more statistically accurate a result, but that doesn't mean that smaller sample sizes are completely inaccurate, because even with smaller samples a trend will emerge. For instance, if we were to collect the Uber drop data of several players and then continue to do so over time, the result would become increasingly more accurate. An issue I see, is that there are a lot of D2/r expert sites that post data on item drop rates without revealing how they arrived to their conclusion. Without methodology how do we know that expert site information is accurate? That leaves us with at best empirical data, which is usually a good start in forming a hypothesis but not in itself a result. One way to gather data is tolook at forum trade data, but even that is somewhat anecdotal. because which

Hellfire Torch's are more likely to be traded and which kept?

The drop calculators you find online are based on the actual game data. The entire drop system is bundled with the game, all parameters. Anyone, at any time, can delve into it and parse the internal txt files. The only difference between the online and offline version of the game is that in online mode, the random numbers are generated server side whereas in offline mode the random numbers are generated by the local game client. You can alter these values at will for offline play to experiment with how it works. The drop calculators are not always 100% accurate, because the drop calculations are inherently extremely complex and depends on a lot of factors. Theoretically, if you have a 100% correct interpretation of the drop algorithm, internal data files and know all the factors affecting a possible drop, you should end up with an accurate prediction. It will always be a prediction though. Also, the human mind is terrible at grasping probability and great at seeing patterns and trends, even where there are none. You can't ID 20 torches and extrapolate that you'll get the same distribution of torches over 1000s of ID:s.

Also, if you're curious, you can ask them. The people making the drop calculators and the experts as you call them are not some mysterious nameless figureheads, they're people just like you, you can find them on various prominent D2R discord servers, some may even be members here (just not very active).

Also, if you're curious, you can ask them. The people making the drop calculators and the experts as you call them are not some mysterious nameless figureheads, they're people just like you, you can find them on various prominent D2R discord servers, some may even be members here (just not very active).

My post wasn't intended to insult any D2r sites, and especially not well ordered sites such as Maxroll which I frequent regularly. I don't see any of them as "mysterious nameless figureheads," but I'm a skeptic by nature and data driven when it comes to facts and source citation. That being said without naming any particular site, I have encountered a site or two with less than accurate information that conflicted with information found on another site. Putting that aside, my point was more about how rng isn't all that random because it can be manipulated.

My previous post

The problem with drop rates of items in D2r is that rng isn't completely random or magic find percentages wouldn't work. Every class of items in the game have a different range of possible numbers that can be generated. A tiered set of numbers determine what class of item is dropped such as a sword or

Helm or whatever, and then another set will determine it's quality. The higher the item quality or desirability the higher the range to make it more rarefied. Magic find for instance can alter the rate of rng by changing the range of possible numbers generated. This means Blizzard can quickly change the drop rate of loot by altering the range of all or some of the items. Magic Find weekend demonstrated just how quickly and easily Blizzard can adjust drop rate. If we took a poll on

Hellfire Torch drop rate, I wouldn't at all be surprised if Paladin and Sorceress Torches drop at a much lower rate than Druid, Barbarian, or Necromancer torches. Since a torch drops 100% of the time at the finish of an Uber run, all that's left is class, quality, and range which is adjustable.

mhlg wrote: 7 months ago

I'm a skeptic by nature

Always a good practice to question

mhlg wrote: 7 months ago

without naming any particular site, I have encountered a site or two with less than accurate information that conflicted with information found on another site. Putting that aside, my point was more about how rng and how isn't all that random because it can be manipulated.

The million dollar problem. While there are staple resources like AB/PK forums; many of which are highly accurate reverse engineered information, there is indeed a ton of misinformation floating around.

Many of the drop calculators are inaccurate in many ways which is why I've been a huge advocate of Nishi who has been motivated to solve; but not even his is perfect. Him and I have continually QA'd and caught errors. Good thing is, he always acts swiftly to fix. He hangs in my Discord which is nice which is a conduit to direct engagement with him and feedback.

But when in doubt, its always good to cross-check a number of resources in an attempt to find the most consistent common ground. For me, I have about 3-4 trusted "advisors" that I find highly credible so whenever I encounter something unusual/contrarian, I will personally try to consult.

Th3ory wrote: 7 months ago

mhlg wrote: 7 months ago

I'm a skeptic by nature

Always a good practice to question

mhlg wrote: 7 months ago

without naming any particular site, I have encountered a site or two with less than accurate information that conflicted with information found on another site. Putting that aside, my point was more about how rng and how isn't all that random because it can be manipulated.

The million dollar problem. While there are staple resources like AB/PK forums; many of which are highly accurate reverse engineered information, there is indeed a ton of misinformation floating around.

Many of the drop calculators are inaccurate in many ways which is why I've been a huge advocate of Nishi who has been motivated to solve; but not even his is perfect. Him and I have continually QA'd and caught errors. Good thing is, he always acts swiftly to fix. He hangs in my Discord which is nice which is a conduit to direct engagement with him and feedback.

But when in doubt, its always good to cross-check a number of resources in an attempt to find the most consistent common ground. For me, I have about 3-4 trusted "advisors" that I find highly credible so whenever I encounter something unusual/contrarian, I will personally try to consult.

I've already bookmarked Nishi because of an earlier post by you recommending the site.

https://nishicode.com/#/

mhlg wrote: 7 months ago

My post wasn't intended to insult any D2r sites, and especially not well ordered sites such as Maxroll which I frequent regularly. I don't see any of them as "mysterious nameless figureheads," but I'm a skeptic by nature and data driven when it comes to facts and source citation. That being said without naming any particular site, I have encountered a site or two with less than accurate information that conflicted with information found on another site. Putting that aside, my point was more about how rng and how isn't all that random because it can be manipulated.

Apologies, I suppose I read a bit more into your post than what you meant. Sorry about that.

Regarding your original post, it's a common fallacy to extrapolate way more than is reasonable based on a small set of data. Random numbers don't tend to even out until a very large amount of numbers have been drawn, and they never truly even out because if they did they wouldn't be random. The standard deviation (the spread) will get smaller the more numbers are drawn, but there's no law of nature that guarantees that it evens out perfectly at any point. A theoretical pool of 100 torches is very likely to produce close to or even over 20 of any one class and around 10 of any one class, but it's going to vary which is the ~20 and which is the ~10. Even a pool of 10000 torches won't even out perfectly (but the spread will most likely be much lower than that of 100).

Yea, I tend to just buy them lol, unless I want to do Ubers just because, but that's not very often

ShadowHeart wrote: 7 months ago

The only difference between the online and offline version of the game is that in online mode, the random numbers are generated server side whereas in offline mode the random numbers are generated by the local game client.

Yes, exactly... But unlike some people may think everything in this (and many other games) IS global/server-side, well at least for the Online part of the game. What does this mean statistically? Well that means if you do 7 runs and it drops 7 times the same class torch for you in a row, and 6 more people out there does the same thing but for them the drop 7 Torches of same but different class than yours and each others it means, statistically, on average that everyone is dropping a different Torch every single time over the 7 runs even tho it's not the case actually.

It doesn't even need to be that "rare" for server to balance it out -- a simple math => if you get 5 Necro Torches from 5 runs, player No. 2 gets 2x Paladin, 2x Sorc and 1x Necro also from 5 runs, player No. 3 gets 2x Barb, 2x Sin and 1x Necro and player No. 4 gets 2x Druid, 2x Zon and 1x Necro, on average (server wise) every single different Torch dropped exactly 2 times no matter how many exact ones each player actually got.
"Statistics are like a good swimsuit -- they cover the main parts and reveal irrelevant stuff" or whatever is the quote. But you get what I wanted to say.

ShadowHeart wrote: 7 months ago

Also, the human mind is terrible at grasping probability and great at seeing patterns and trends, even where there are none. You can't ID 20 torches and extrapolate that you'll get the same distribution of torches over 1000s of ID:s.

That is true, psychologically human mind tends to save/pinpoint negative data/experience/patterns and trends (and often neglect) much more than positive ones. However, true RNG is exactly what I described above. It can happen that you, as a singular player, can have a massively bad luck. On the other hand, you'll get lucky over something bizzare or unneeded that you won't even notice since it doesn't affect you as much as something else does that is dear to you.

From a random distribution, it's very unlikely that you'll get a perfect even distribution from a small sample set. It's far more likely that a small sample set will have clumps of one kind or another.

It's unlikely to get any particular result, of course, but there are far more permutations that are clumped than there are that are evenly distributed. People expect the perfect distribution, but they should expect the opposite.