EDIT - I'm mainly done with this, and happy with the results of:
1) Top, top left, and bottom left boxen are always bounty reward when there are five reward boxes. Top right comes from the party. Bottom right is random.
2) When there's a legendary coin option for rewards: HC bounty reward boxes are approximately 27% legendary, 29% Epic, and 44% Rare. The legendary rate is approximately halved, to around 13.5%, when on normal.
3) HC data is across Felwood and BRM; normal data is from Coren and Emperor. No anomaly was found in my data for normal Emperor.
I'm looking at doing a few hundred runs of something like Fallen guardians HC and a few others for looking at bounty coin droprate breakdowns and testing my hypothesis regarding rates between the chars (specifically, my hypothesis is that there's 10 equal buckets, 6 for the mercs in your party, 3 for bounty rewards, and 1 random, and each of the five rewards picks randomly. No weighting based on rarity etc., and no difference between HC/normal). First, though, I'd like to check if there's any data on this already that people have gathered? Initially I need data with no overlap between party and bounty (how that interacts, or smaller parties if the game allows, would be a followup) - if someone's already gathered this, would be nice to know ahead of time!
Should be fine to mix and match bounties as long as I avoid overlaps in rewards/team. Overlap between random and others should be easily accounted for, I hope, but no way to prevent that.
Edit:
Interesting preliminary report - it doesn't quite seem to work like that. Low sample size, but looking like 1/5 rewards - specifically the one bottom right - is *always* a random merc - only exception is once when it's Tamsin which seems likely to be the random hitting one of the bounty rewards. Top right appears to be always a party merc. Top, top left, and bottom left are always a bounty reward. Far too early to look at rates between the mercs, though.
It's worth bearing in mind here that the top right and bottom right may actually be due to the order I'm opening rewards in - so it'd go first is bounty reward, second is party, third is random, 4th/5th are bounty rewards (which would scale down to the lower bounties with three rewards). Interesting idea here is that the first bounty would give you more random coins, though it'll take a lot of data to determine coin rarities etc..
Edit 2 - Early data on rarities shows that the rare and epic coins from bounties are at roughly the same rate. Legendaries are notably rarer - might be a 2:2:1 ratio, but early. No pattern yet in the significantly smaller datasets for party/random mercs. Variance on rarity breakdowns is hilaiously high. Looked like 1:2:2 for a good while, but now at 17/22/31, so we'll see what more data does. With only one leg in the group and the rest being rares, no visible signal there, but small dataset. Random shows no skew so far, but again needs a LOT more data. Lots.
Since I know how party overlap will be treated now, I can go for any bounty/group composition to gather more data - pretty much done on Malf, so will switch.
Very interested in the results of your experiment, keep posting. I will try to add to it by keeping track a little. Mostly rewards, keeping track of rarities is a bit much but maybe vaguely.
Still seems optimal farming strategy is to go for the easiest, non-HC bounty with the hero you want, ideally with that hero in party. Which is fairly obvious but did not know that HC gave equal coins, that is dumb.
Very interested in the results of your experiment, keep posting. I will try to add to it by keeping track a little. Mostly rewards, keeping track of rarities is a bit much but maybe vaguely.
Still seems optimal farming strategy is to go for the easiest, non-HC bounty with the hero you want, ideally with that hero in party. Which is fairly obvious but did not know that HC gave equal coins, that is dumb.
HC doesn't give equal coins - but it's a small increase. I'm not tracking the number of coins, but on normal I was seeing sizes of 10-20 coins, whilst heroic seems to be 15-25 20-30.
Not quite double rewards for significantly harder. Though Emperor HC was more fun, honestly - every death was still my fault, though I only did it twice and didn't get bad RNG on the fire sworn.
Pretty preliminary but yeah I found roughly what you did. Three coins bounty, one party and one random. Epics and rares seem to be about equal, legends noticeably rarer and smaller.
I open the boxes in random order, and like you, I always get top right = party; bottom right = random; other three = bounty reward. (Bottom right happens to match reward or party about as often as you'd expect from a random pull, but randos never show up in the other slots.)
Rollback Post to RevisionRollBack
"Why, you never expected justice from a company, did you? They have neither a soul to lose nor a body to kick." -- Lady Saba Holland
Okay, since I've gained a bit of data and some traction, I'll go ahead and share the data, along with some assumptions and comments.
Assumptions: 1) Normal and HC bounties are the same. Not really relevant due to all ones done so far being HC.
2) All bounties have the same rarity rates. This should be roughly visible in the data when I have more collected, but overall figure assumes that they're all the same. No clue how bounties without legendary options are handled, that's outside the scope of this project, since the important thing is grinding those legs. Obviously, my dataset for Emperor is not going to be even vaguely reliable for comparisons.
3) The locations could be tied to order opened in, but comments in this thread imply not. I'm still opening in a fixed clockwise order from the top, though, for safety's sake.
D) Only bounties with 5 boxes as rewards are examined.
5) Assume that the rate of legendaries:epics:rares is the same across all three reward boxen for bounty rewards.
Conclusions/comments:
Far too early to tell actual rates on leg:epic:rare ratios with any sort of confidence, other than that it's not that bad overall, though high variance.
Yes, I should have the analysis on another sheet, really, but tough.
Anacondra got subbed in after a few on Corrupted Ancient. That's why she's in those coins.
I have less than 100 datapoints currently. That's too early to say with any sort of certainty.
Legendary ratio within group feels like it's not really relevant, so not focussed on.
Random pool is likely to be 100% random with no rarity weighting, but would need stupid piles of data to determine.
That boxed set of figures is the overall.
in party/rewards/other table was meaningful, but was changed since the pattern was clear - it shows exactly what you would expect in the first dataset if the top and both left boxes are bounty keyed, top right from your team, and bottom-right entirely random. This is pretty much guaranteed to be true. Impacts in that you get three slots that are always bounty coins from 5.
I don't have anything in particular to add to this thread, I just wanted to say it's great to see a well framed, analytical post that supports its premise with empirical information. Well done.
I'm relatively happy saying that legendary droprate in HCs for the bounty rewards is approximately 28% or so. As I'm struggling with winterspring and BRM, I'm moving onto normal - and already it's looking different enough that I'm separating out the data. I had been hoping to combine these, but it really feels as though it's not going to converge, despite being early days and small data set.
Three bundles will be for bounty rewards, one for mercs in your team, one random, for the five reward bounties.
Appears pretty flat distributionf or random and in-party mercs, though not really the scope of the study.
On HC, each of the three bounty rewards has an approximately 28% chance to be of the legendary (assuming you are doing a bounty with legendary coins). On normal that drops to approximately... 8-9%.
Yes, this is a split between normal and HC, and the 28% HC figure tracks between BRM and Felwood heroics.
Conclusion: Only farm HC if you can. Normal is not worth it whatsoever, particularly when you factor in that bundles are smaller across the board in N as well.
Three bundles will be for bounty rewards, one for mercs in your team, one random, for the five reward bounties.
Appears pretty flat distributionf or random and in-party mercs, though not really the scope of the study.
On HC, each of the three bounty rewards has an approximately 28% chance to be of the legendary (assuming you are doing a bounty with legendary coins). On normal that drops to approximately... 8-9%.
Yes, this is a split between normal and HC, and the 28% HC figure tracks between BRM and Felwood heroics.
Conclusion: Only farm HC if you can. Normal is not worth it whatsoever, particularly when you factor in that bundles are smaller across the board in N as well.
Something is not right here. I've been farming normal Emperor Thaurissan for diablo coins and the split was roughly equal (possibly tiny penalty to legendary like 2-3%, I. e 30% of coins where diablo's).
Three bundles will be for bounty rewards, one for mercs in your team, one random, for the five reward bounties.
Appears pretty flat distributionf or random and in-party mercs, though not really the scope of the study.
On HC, each of the three bounty rewards has an approximately 28% chance to be of the legendary (assuming you are doing a bounty with legendary coins). On normal that drops to approximately... 8-9%.
Yes, this is a split between normal and HC, and the 28% HC figure tracks between BRM and Felwood heroics.
Conclusion: Only farm HC if you can. Normal is not worth it whatsoever, particularly when you factor in that bundles are smaller across the board in N as well.
Something is not right here. I've been farming normal Emperor Thaurissan for diablo coins and the split was roughly equal (possibly tiny penalty to legendary like 2-3%, I. e 30% of coins where diablo's).
Interesting. You got some recorded data? As shown in the linked SS, I've got 15 runs, or 45 data points (since you get three reward bundles per run), of which 4 were Uther (HC and N being identical would expect...12.6ish), from normal Coren (Those with N in the first column are normal; H denotes heroic). I see no reason for Emperor and Coren to be different rates. I presume your roughly equal is based on the top and the two left (as you face the screen) bundles etc.. What I can do is do more data collection on normal Coren, though probably not too much for the next few days due to commitments. If that's still showing a significantly lower rate, I'll grind more Diablo coins on normal and see how 45 data points of that looks.
It's been over a decade since I did stats and even then my masters used bespoke HEP software to kick out the analysis, so I'm wary of actually running analysis to glean P-values and similar.
Three bundles will be for bounty rewards, one for mercs in your team, one random, for the five reward bounties.
Appears pretty flat distributionf or random and in-party mercs, though not really the scope of the study.
On HC, each of the three bounty rewards has an approximately 28% chance to be of the legendary (assuming you are doing a bounty with legendary coins). On normal that drops to approximately... 8-9%.
Yes, this is a split between normal and HC, and the 28% HC figure tracks between BRM and Felwood heroics.
Conclusion: Only farm HC if you can. Normal is not worth it whatsoever, particularly when you factor in that bundles are smaller across the board in N as well.
Something is not right here. I've been farming normal Emperor Thaurissan for diablo coins and the split was roughly equal (possibly tiny penalty to legendary like 2-3%, I. e 30% of coins where diablo's).
Interesting. You got some recorded data? As shown in the linked SS, I've got 15 runs, or 45 data points (since you get three reward bundles per run), of which 4 were Uther (HC and N being identical would expect...12.6ish), from normal Coren (Those with N in the first column are normal; H denotes heroic). I see no reason for Emperor and Coren to be different rates. I presume your roughly equal is based on the top and the two left (as you face the screen) bundles etc.. What I can do is do more data collection on normal Coren, though probably not too much for the next few days due to commitments. If that's still showing a significantly lower rate, I'll grind more Diablo coins on normal and see how 45 data points of that looks.
It's been over a decade since I did stats and even then my masters used bespoke HEP software to kick out the analysis, so I'm wary of actually running analysis to glean P-values and similar.
Sadly I haven't recorded anything. I didn't know it would be useful. And yes, I'm taking about the top and two leftside ones, I'm quite sure I didn't get any diablo coin from random "slot" (cause that would make me happy). Also it doesn't seem to be just subjective filling since I was at nearly 300 coins with diablo when I was able to craft Brightwing(epic) and they both started from 0. There could be a bit more Samuro there but definitely not even close to 80% (assuming both diablo and brightwing are 10% each). And finally if diablo was really 10% only, if my math is correct I would need to do nearly 100 runs and there's no way I did that many (I think I did around 30 runs but that might be confirmation bias in my head).
Math: I got last 75 coins from Velen's task, which leaves 425 from emperor, 45 total coins per run [left+top side only], assuming 10% that would be 4.5 diablo coins per run which gives 94.(4) runs) .
I will most likely do few more runs to level up diablo's skills faster so I might write down few records but don't expect huge amount of data.
Just pitching in my (limited) coins data from Thaurissan runs. I did 9 runs to farm Diablo coins. I didn't record all the coins I got, but I did record the Diablo ones and those were: 20+15, 15+15, 20, 15, 15, 10, 0, 0, 0. So a total of 135 coins in 9 runs, or an average of 15 coins per run. Of course the sample is small meaning the standard deviation is quite large (15) and thus the standard error is 5. Assuming the normal distribution of the sampling mean, the 95% confidence intervals are 15+/-10 coins per run.
Conclusion: We need more data, meaning more runs;)
Just pitching in my (limited) coins data from Thaurissan runs. I did 9 runs to farm Diablo coins. I didn't record all the coins I got, but I did record the Diablo ones and those were: 20+15, 15+15, 20, 15, 15, 10, 0, 0, 0. So a total of 135 coins in 9 runs, or an average of 15 coins per run. Of course the sample is small meaning the standard deviation is quite large (15) and thus the standard error is 5. Assuming the normal distribution of the sampling mean, the 95% confidence intervals are 15+/-10 coins per run.
Conclusion: We need more data, meaning more runs;)
That's 9 runs, 8 diablo procs, which is 8/27 or approx 30%. Very interesting, since even with expanded data Uther coins from Coren is approximately 10% and really not moving upwards much. At 20 normal runs, which is 60 boxes, and exactly 6 Uther coin bundles. What I might do is cut this short after another 5 runs (15 more data points) if there's no significant improvement and collect data on Emperor (N).
This is tending towards the conclusion that every bounty sampled so far has a similar rate EXCEPT normal Coren. Which, honestly, sounds ridiculous and I want to dismiss it out of hand (though I can't, obviously). Uther/Manno/Murk is L/E/R, much like most other bounties including Emperor. Maybe some rarities got tweaked late as an explanation if this is true?
So, on to-do list is 20ish runs normal Emp, 20ish runs of a couple of other normals, and try to understand what the outlier is.
ETA: 25 runs on Coren N now. 75 data points, 10 Uthers. 13ish%. I'd expect double that if the rate's consistent with the HC rate. Still higher than the initial, but nowhere near the HC data. Should grind out another bunch before switching, really, to nail it down a bit more.
Just pitching in my (limited) coins data from Thaurissan runs. I did 9 runs to farm Diablo coins. I didn't record all the coins I got, but I did record the Diablo ones and those were: 20+15, 15+15, 20, 15, 15, 10, 0, 0, 0. So a total of 135 coins in 9 runs, or an average of 15 coins per run. Of course the sample is small meaning the standard deviation is quite large (15) and thus the standard error is 5. Assuming the normal distribution of the sampling mean, the 95% confidence intervals are 15+/-10 coins per run.
Conclusion: We need more data, meaning more runs;)
@fusilli: Is this even meaningful, though, if we don't know which slots the Diablo coins appeared in? A Diablo coin in the party slot is not the same as one in a bounty slot.
It was always one of the 3 slots on the left. That was me farming coins to craft Diablo therefore he was not in my party.
I am now collecting the same data for Malfurion in Fallen Guardians, I'll post again results when I am done.
What I can claim with a high degree of confidence is that:
Legendary coins for the merc you are trying to farm (the legendary merc that is listed when you click on the bounty box before starting the run) seem to only show up in one of the 3 slots on the left
The outcome of each of these three slots seems to be independent from the other two, meaning that each of them offers coins for one of the three mercs that are part of that run's bounty with some fixed, unknown probability for each of the three
The number of coins you get on each slot is also random. I've received 10, 15, 20 or 30 so far. I haven't seen 25 yet. I have no idea how likely each number is and whether it is conditional on the mercenary (meaning, different probabilities for legendary coins than for epic/rare ones)
Finally one thing that I have not considered at all is whether number of battles and difficulty of battles during the run affect final rewards and in what way
Just pitching in my (limited) coins data from Thaurissan runs. I did 9 runs to farm Diablo coins. I didn't record all the coins I got, but I did record the Diablo ones and those were: 20+15, 15+15, 20, 15, 15, 10, 0, 0, 0. So a total of 135 coins in 9 runs, or an average of 15 coins per run. Of course the sample is small meaning the standard deviation is quite large (15) and thus the standard error is 5. Assuming the normal distribution of the sampling mean, the 95% confidence intervals are 15+/-10 coins per run.
Conclusion: We need more data, meaning more runs;)
@fusilli: Is this even meaningful, though, if we don't know which slots the Diablo coins appeared in? A Diablo coin in the party slot is not the same as one in a bounty slot.
I'm assuming that they're all appearing in bounty slots - I'm not building that into my data, though, instead using it to say "this looks odd, I'd better investigate" and then gathering data. I'll plough through a bunch of Emps soon enough!
It was always one of the 3 slots on the left. That was me farming coins to craft Diablo therefore he was not in my party.
I am now collecting the same data for Malfurion in Fallen Guardians, I'll post again results when I am done.
What I can claim with a high degree of confidence is that:
Legendary coins for the merc you are trying to farm (the legendary merc that is listed when you click on the bounty box before starting the run) seem to only show up in one of the 3 slots on the left
The outcome of each of these three slots seems to be independent from the other two, meaning that each of them offers coins for one of the three mercs that are part of that run's bounty with some fixed, unknown probability for each of the three
The number of coins you get on each slot is also random. I've received 10, 15, 20 or 30 so far. I haven't seen 25 yet. I have no idea how likely each number is and whether it is conditional on the mercenary (meaning, different probabilities for legendary coins than for epic/rare ones)
Finally one thing that I have not considered at all is whether number of battles and difficulty of battles during the run affect final rewards and in what way
Normal or HC Fallen guardians? Got some data for HC, which came out as 17:23:32 getween the three, but that's 72 datapoints, so 24 runs before I got Malf and stopped. Curious to see if normal is the same, considering the oddities with Emp/Coren.
The top and left boxes are for the bounty rewards; top right is party, bottom right is completely random (so even if Malf comes in there, discount it). Always that order.
Okay, absolutely no support of N Emperor being significantly higher rate after 15 runs (45 datapoints). Very swingy data, but 7:8:30 is on the high end but within reason for a 13% or so rate for normals, whilst not fitting a 27% rate whatsoever. Much like with N Coren, I would expect approx double the amount of legendaries.
I'll burn through a few more runs, but not expecting any stellar changes. Not sure if there's reporting bias or just great luck from the anecdata, honestly.
Rollback Post to RevisionRollBack
To post a comment, please login or register a new account.
EDIT - I'm mainly done with this, and happy with the results of:
1) Top, top left, and bottom left boxen are always bounty reward when there are five reward boxes. Top right comes from the party. Bottom right is random.
2) When there's a legendary coin option for rewards: HC bounty reward boxes are approximately 27% legendary, 29% Epic, and 44% Rare. The legendary rate is approximately halved, to around 13.5%, when on normal.
3) HC data is across Felwood and BRM; normal data is from Coren and Emperor. No anomaly was found in my data for normal Emperor.
I'm looking at doing a few hundred runs of something like Fallen guardians HC and a few others for looking at bounty coin droprate breakdowns and testing my hypothesis regarding rates between the chars (specifically, my hypothesis is that there's 10 equal buckets, 6 for the mercs in your party, 3 for bounty rewards, and 1 random, and each of the five rewards picks randomly. No weighting based on rarity etc., and no difference between HC/normal). First, though, I'd like to check if there's any data on this already that people have gathered? Initially I need data with no overlap between party and bounty (how that interacts, or smaller parties if the game allows, would be a followup) - if someone's already gathered this, would be nice to know ahead of time!
Should be fine to mix and match bounties as long as I avoid overlaps in rewards/team. Overlap between random and others should be easily accounted for, I hope, but no way to prevent that.
Edit:
Interesting preliminary report - it doesn't quite seem to work like that. Low sample size, but looking like 1/5 rewards - specifically the one bottom right - is *always* a random merc - only exception is once when it's Tamsin which seems likely to be the random hitting one of the bounty rewards. Top right appears to be always a party merc. Top, top left, and bottom left are always a bounty reward. Far too early to look at rates between the mercs, though.
It's worth bearing in mind here that the top right and bottom right may actually be due to the order I'm opening rewards in - so it'd go first is bounty reward, second is party, third is random, 4th/5th are bounty rewards (which would scale down to the lower bounties with three rewards). Interesting idea here is that the first bounty would give you more random coins, though it'll take a lot of data to determine coin rarities etc..
Edit 2 -
Early data on rarities shows that the rare and epic coins from bounties are at roughly the same rate. Legendaries are notably rarer - might be a 2:2:1 ratio, but early. No pattern yet in the significantly smaller datasets for party/random mercs.Variance on rarity breakdowns is hilaiously high. Looked like 1:2:2 for a good while, but now at 17/22/31, so we'll see what more data does. With only one leg in the group and the rest being rares, no visible signal there, but small dataset. Random shows no skew so far, but again needs a LOT more data. Lots.Since I know how party overlap will be treated now, I can go for any bounty/group composition to gather more data - pretty much done on Malf, so will switch.
I had seen the same pattern that you mentioned: left zone always bounty, right zone random and party
Luckily in the malfurion bounty i receive two times diablo coins in the random spot, getting close to craft him and avoid the f****ing thaurissan
Very interested in the results of your experiment, keep posting. I will try to add to it by keeping track a little. Mostly rewards, keeping track of rarities is a bit much but maybe vaguely.
Still seems optimal farming strategy is to go for the easiest, non-HC bounty with the hero you want, ideally with that hero in party. Which is fairly obvious but did not know that HC gave equal coins, that is dumb.
HC doesn't give equal coins - but it's a small increase. I'm not tracking the number of coins, but on normal I was seeing sizes of 10-20 coins, whilst heroic seems to be
15-2520-30.Not quite double rewards for significantly harder. Though Emperor HC was more fun, honestly - every death was still my fault, though I only did it twice and didn't get bad RNG on the fire sworn.
Pretty preliminary but yeah I found roughly what you did. Three coins bounty, one party and one random. Epics and rares seem to be about equal, legends noticeably rarer and smaller.
Thanks also for the info on HC rewards.
I open the boxes in random order, and like you, I always get top right = party; bottom right = random; other three = bounty reward. (Bottom right happens to match reward or party about as often as you'd expect from a random pull, but randos never show up in the other slots.)
"Why, you never expected justice from a company, did you? They have neither a soul to lose nor a body to kick." -- Lady Saba Holland
Okay, since I've gained a bit of data and some traction, I'll go ahead and share the data, along with some assumptions and comments.
Assumptions:
1) Normal and HC bounties are the same. Not really relevant due to all ones done so far being HC.
2) All bounties have the same rarity rates. This should be roughly visible in the data when I have more collected, but overall figure assumes that they're all the same. No clue how bounties without legendary options are handled, that's outside the scope of this project, since the important thing is grinding those legs. Obviously, my dataset for Emperor is not going to be even vaguely reliable for comparisons.
3) The locations could be tied to order opened in, but comments in this thread imply not. I'm still opening in a fixed clockwise order from the top, though, for safety's sake.
D) Only bounties with 5 boxes as rewards are examined.
5) Assume that the rate of legendaries:epics:rares is the same across all three reward boxen for bounty rewards.
Conclusions/comments:
Far too early to tell actual rates on leg:epic:rare ratios with any sort of confidence, other than that it's not that bad overall, though high variance.
Yes, I should have the analysis on another sheet, really, but tough.
Anacondra got subbed in after a few on Corrupted Ancient. That's why she's in those coins.
I have less than 100 datapoints currently. That's too early to say with any sort of certainty.
Legendary ratio within group feels like it's not really relevant, so not focussed on.
Random pool is likely to be 100% random with no rarity weighting, but would need stupid piles of data to determine.
That boxed set of figures is the overall.
in party/rewards/other table was meaningful, but was changed since the pattern was clear - it shows exactly what you would expect in the first dataset if the top and both left boxes are bounty keyed, top right from your team, and bottom-right entirely random. This is pretty much guaranteed to be true. Impacts in that you get three slots that are always bounty coins from 5.
Google sheet is at https://docs.google.com/spreadsheets/d/18yRuv8an8lYMzA8b1TdY8UanG7b-jyWjCB75xBhm1hQ/edit?usp=sharing - no, you may not add your own data. I don't trust random people not to troll.
I don't have anything in particular to add to this thread, I just wanted to say it's great to see a well framed, analytical post that supports its premise with empirical information. Well done.
I'm relatively happy saying that legendary droprate in HCs for the bounty rewards is approximately 28% or so. As I'm struggling with winterspring and BRM, I'm moving onto normal - and already it's looking different enough that I'm separating out the data. I had been hoping to combine these, but it really feels as though it's not going to converge, despite being early days and small data set.
Okay, results!
Three bundles will be for bounty rewards, one for mercs in your team, one random, for the five reward bounties.
Appears pretty flat distributionf or random and in-party mercs, though not really the scope of the study.
On HC, each of the three bounty rewards has an approximately 28% chance to be of the legendary (assuming you are doing a bounty with legendary coins). On normal that drops to approximately... 8-9%.
Yes, this is a split between normal and HC, and the 28% HC figure tracks between BRM and Felwood heroics.
Conclusion: Only farm HC if you can. Normal is not worth it whatsoever, particularly when you factor in that bundles are smaller across the board in N as well.
Something is not right here. I've been farming normal Emperor Thaurissan for diablo coins and the split was roughly equal (possibly tiny penalty to legendary like 2-3%, I. e 30% of coins where diablo's).
Interesting. You got some recorded data? As shown in the linked SS, I've got 15 runs, or 45 data points (since you get three reward bundles per run), of which 4 were Uther (HC and N being identical would expect...12.6ish), from normal Coren (Those with N in the first column are normal; H denotes heroic). I see no reason for Emperor and Coren to be different rates. I presume your roughly equal is based on the top and the two left (as you face the screen) bundles etc.. What I can do is do more data collection on normal Coren, though probably not too much for the next few days due to commitments. If that's still showing a significantly lower rate, I'll grind more Diablo coins on normal and see how 45 data points of that looks.
It's been over a decade since I did stats and even then my masters used bespoke HEP software to kick out the analysis, so I'm wary of actually running analysis to glean P-values and similar.
Sadly I haven't recorded anything. I didn't know it would be useful. And yes, I'm taking about the top and two leftside ones, I'm quite sure I didn't get any diablo coin from random "slot" (cause that would make me happy). Also it doesn't seem to be just subjective filling since I was at nearly 300 coins with diablo when I was able to craft Brightwing(epic) and they both started from 0. There could be a bit more Samuro there but definitely not even close to 80% (assuming both diablo and brightwing are 10% each). And finally if diablo was really 10% only, if my math is correct I would need to do nearly 100 runs and there's no way I did that many (I think I did around 30 runs but that might be confirmation bias in my head).
Math: I got last 75 coins from Velen's task, which leaves 425 from emperor, 45 total coins per run [left+top side only], assuming 10% that would be 4.5 diablo coins per run which gives 94.(4) runs) .
I will most likely do few more runs to level up diablo's skills faster so I might write down few records but don't expect huge amount of data.
Just pitching in my (limited) coins data from Thaurissan runs. I did 9 runs to farm Diablo coins. I didn't record all the coins I got, but I did record the Diablo ones and those were: 20+15, 15+15, 20, 15, 15, 10, 0, 0, 0. So a total of 135 coins in 9 runs, or an average of 15 coins per run. Of course the sample is small meaning the standard deviation is quite large (15) and thus the standard error is 5. Assuming the normal distribution of the sampling mean, the 95% confidence intervals are 15+/-10 coins per run.
Conclusion: We need more data, meaning more runs;)
That's 9 runs, 8 diablo procs, which is 8/27 or approx 30%. Very interesting, since even with expanded data Uther coins from Coren is approximately 10% and really not moving upwards much. At 20 normal runs, which is 60 boxes, and exactly 6 Uther coin bundles. What I might do is cut this short after another 5 runs (15 more data points) if there's no significant improvement and collect data on Emperor (N).
This is tending towards the conclusion that every bounty sampled so far has a similar rate EXCEPT normal Coren. Which, honestly, sounds ridiculous and I want to dismiss it out of hand (though I can't, obviously). Uther/Manno/Murk is L/E/R, much like most other bounties including Emperor. Maybe some rarities got tweaked late as an explanation if this is true?
So, on to-do list is 20ish runs normal Emp, 20ish runs of a couple of other normals, and try to understand what the outlier is.
ETA: 25 runs on Coren N now. 75 data points, 10 Uthers. 13ish%. I'd expect double that if the rate's consistent with the HC rate. Still higher than the initial, but nowhere near the HC data. Should grind out another bunch before switching, really, to nail it down a bit more.
@fusilli: Is this even meaningful, though, if we don't know which slots the Diablo coins appeared in? A Diablo coin in the party slot is not the same as one in a bounty slot.
"Why, you never expected justice from a company, did you? They have neither a soul to lose nor a body to kick." -- Lady Saba Holland
It was always one of the 3 slots on the left. That was me farming coins to craft Diablo therefore he was not in my party.
I am now collecting the same data for Malfurion in Fallen Guardians, I'll post again results when I am done.
What I can claim with a high degree of confidence is that:
I'm assuming that they're all appearing in bounty slots - I'm not building that into my data, though, instead using it to say "this looks odd, I'd better investigate" and then gathering data. I'll plough through a bunch of Emps soon enough!
Normal or HC Fallen guardians? Got some data for HC, which came out as 17:23:32 getween the three, but that's 72 datapoints, so 24 runs before I got Malf and stopped. Curious to see if normal is the same, considering the oddities with Emp/Coren.
The top and left boxes are for the bounty rewards; top right is party, bottom right is completely random (so even if Malf comes in there, discount it). Always that order.
Okay, absolutely no support of N Emperor being significantly higher rate after 15 runs (45 datapoints). Very swingy data, but 7:8:30 is on the high end but within reason for a 13% or so rate for normals, whilst not fitting a 27% rate whatsoever. Much like with N Coren, I would expect approx double the amount of legendaries.
I'll burn through a few more runs, but not expecting any stellar changes. Not sure if there's reporting bias or just great luck from the anecdata, honestly.