You know i have not posted much on this topic as it really seems like a waste but.......

You are complaining about 2 things 

RNG, from my experience in all games i have played, it is streaky. you either accept this or you should move on.

Free stuff, while I understand this to a degree as I did not like that they removed the 5* chicken and Sacred Shard from the daily log in rewards, too me CB is different because it is part of RNG.

I guess for me you either accept RNG and trust when people tell you nothing has changed or if you don't then move on because all this complaining about RNG will get you no place in a RNG based game IMO.

Bimodal distributions are not uncommon and frequently indicate poor sampling. (EDIT: I guess I shouldn't say poor sampling, but that sampling methods need to be refined based on the results.)  In essence, the authors point out the reasons, there are two populations being combined in one sample.

E.g.  If I sample the height of all persons in a given geographic area, a bimodal distribution may not be immediatly appearent in a visual histogram, but analysis will certainly show that the average height of males is different from the average height of females.  Expand this to something with a bit more resolution, say the average weight of eagles.  The females generally weigh 20%-40% more than males.  In a smallish sample, the bimodal distribution will be visually appearant.

in the given situation, we know that the authors intended to sample UNM chests, but also included both NM chests and individuals opening 2 chests.  Any sample that includes both NM and UNM chests should be expected to show a bimodal distribution.

I admit the bimodal distribution doesn't necessarilly conform to this example, but it illustrates the issue of bimodal distributions.  There is certainly something going on.  But the 'conclusions' are inconclusive in my mind.  Correlation is a far cry from causation.

To go back to your previous post, I don't think filtering data from users that did not submit 90% is the issue. The authors stated they removed data from "suspicious" reporters.  They also stated that their criteria for determining who was suspicious would remain "secret."  This is unhead of!

Hi Angwil,

Thank you it seems you have experience with bimodal statistics.   Since the authors made their data available would it not be easier to just look at the data and test our own hypothesis on how the authors could be in error?  For example, the number of data points shows that the authors did not combine UNM and NM chests in their histograms.

While I understood your bimodal examples. I did not see how they illustrated your comment that poor sampling could produce a false bimodal distribution.  Your examples seemed to consider how a bimodal distribution could be hidden.  However in this case the histogram plot is clearly bimodal and the question was why someone required confirmation of this with a significance test?

I do not think the UNM and NM were combined but were reported separately as level 6 (UNM) which showed the bimodal distribution and level 5 (NM) which may or may not show it.  However, the bimodal distribution was not clearly apparent in the level 5 data.    

I was also not clear on this in my original comments.  Before I looked at the data I thought that their main data set of over 90% submission was a sum of 2NM and 2UNM chests to maximise the data pool.  I understood that the sum of bimodal parameters (in your example men's height + women height) would hide the bimodal distribution not produce a false one.

The person who I was responding to did take issue with the authors rejecting respondents who contributed less than 90% of the survey This person seemed to think that implied they discarded 90% of their data. I am at a loss to explain how they came to this conclusion.

Why is the author's refusal to disclose the data rejection criteria of their analysis such a critical issue?  This is often the case where there are security issues, or where the algorithm used is proprietary.  Surely anyone who takes issue with this can just look at their raw data and make their own selection criteria.  That is what I did.  I found I was excluding about 2000 data points out of 96748 UNM chest wins (1 or 2 chests).  Did the authors discard significantly more than this?  Could this have created a bimodal distribution where there was none?  

Now I did not read too much into the author's hypothesis of what caused the bimodal distribution.  I do not think they were trying to show causality.  What I am interested in is if this bimodal distribution could have been created accidentally by the sampling process.  The raw data is there for people who are interested to answer this by example, not by further speculation.  For example, the "hypothesis" that the authors mixed the NM and UNM was a good hypothesis, but looking at the data this clearly isn't so.  In fact I don't think the means of the two histogram peaks even allowed for this.

Best Regards.

My interest and how I am using the data...

I have no connection at all with this study, but I am a real live scientist and an old one at that.  I hope to show you that I have both scientific understanding and an independent interest in these results.

I am that 1 in 40000 player who has not had a single sacred shard drop for 42+ days now.  If you think I am fabricating that just put me on ignore because I can do nothing to convince you otherwise.  I did the math and I understand that if there are 40000+ people opening UNM chests each day then the most likely explanation is that I am special.  But it has made me curious.

I needed to find the probability of a sacred dropping to see just how unlucky I could be.

A while ago some other authors published clan boss drop tables that were data mined from the game code, which most players accepted as it agreed with their instinct. The authors were entirely credible and I accept their work. However, they did not disclose how to obtain the data so I could not repeat it. 

I found this bot report which came out about the same time I started looking for some aggregated clan boss data.  This bot data seemed credible because it also confirmed the data mined probabilities. It also showed this curious bimodal distribution of data.

I have not seen anything else even approaching the size of this data set for RSL. I expected the discussion to show further insights from people looking at the data. However, instead, I see a bunch of criticisms and what if's about the data that are not even backed up by alternative analysis of the data. There is no other data presented by anyone else that this data disagrees with.  Some of the comments seemed to discredit it solely because it was Russian.  I took offence to this, since two of my previous research topics (General relativity and Sea Ice Growth have had excellent and significant contributions made by Russian and East Asian researchers.

I have not looked at the full distribution of drops or the books. I only consider a simple Binomial distribution for zero sacreds dropping after n chests to test against drop probabilities. Basically just what the authors reported regarding sacreds.  I thought maybe this would show a significant overabundance of people getting zero shards.  Unfortunately for me, this was not backed by the data.  This does make me feel special though.

My next step is to use the implied bimodal distribution of Epic books from UNM drops to see if this implies a lower probability of shards dropping.  Not conclusive I know but it would change my odds from 1 in 40000 special to 1 in 10000 unlucky, and I would then expect to see others in my situation.

Now if you bravely read this far and are still interested, you could make one of two comments.  Firstly if you have a similar long run of zero sacreds that is ongoing then I would like to hear from you.  Secondly, I would like to estimate better how many players obtain one and two of the top chests from UNM each day.  My clan is ranked > 1000 and only 2 of us get a top UNM chest.  So my best limit guess is most maybe 1000 clans of 40 players getting a chest or 40000.  But this is wild speculation. Can someone with a numerical clan rank share how many chest drops your clan generates per day?  Is my 40000 chest drops wildly out compared to 10000 or 100000...

Best Regards

PS please don't tell me to contact customer support.  You know it takes many days for them to respond especially after big changes going on recently.  I have asked them to confirm the last time I got a sacred shard drop and also a list of how many other shard drops.

The person who I was responding to did take issue with the authors rejecting respondents who contributed less than 90% of the survey This person seemed to think that implied they discarded 90% of their data. I am at a loss to explain how they came to this conclusion.

Hey! That's me! The young scientist to your "old scientist". 

I see you have a lot to say here, including that I neither know basic statistics nor read the articles.  And it seems you are quite personally and emotionally invested, as you haven't gotten a sacred shard in 42+ days. And all of your posts are on the probability of this being close to statistically impossible for this to have happened to you.

Unfortunately, I feel like you have yet to read all the articles yet, as my statements about them rejecting all but approximately 10% of respondants comes from their own words.

i

i

There you go. 644 users from an initial pool of 5799. 11.1% of their total population of respondants, I guess, not 10%. I apologize for my error.  I do not feel it was some egregious criticism to point out such a sampling could include some confounding variables.

And honestly, I'm not sure if I need to respond anymore to someone who encourages me to "please remember your high school statistics classes on probability distributions"  or that says " it is understandable if you don't have time to read the two papers and are unfamiliar with statistics or the scientific processes" while they themselves haven't read through the work closely enough that I can just cut and paste from those papers to answer the criticisms they level at me.

Even more so, you don't need to pontificate to me in general. The authors of those posts were nice enough to reach out to me and we spoke at length. Instead of taking the things on faith that you do about their significance values, their sample groupings, the selection methods, the statistical tests they ran, or why they began collecting this data in the first place, I just asked. And we had a productive and friendly discussion.

Because, come to find out they are both very nice people and, in my experience with science, nice people can ask questions of each other and challenge each others work and that is not only okay, it is encouraged. I live with and deal with it all the time, without getting all bothered about it. It also doesn't typically come with attacks about "fake scientific criticism" or "high school statistics".

You have a good day sir! Thanks for all your feedback on my posts.

"Can someone with a numerical clan rank share how many chest drops your clan generates per day?  Is my 40000 chest drops wildly out compared to 10000 or 100000..."

For what it's worth, my clan is currently ranked 562.  We down brutal, nm, and unm each day.  Some members hit hard and sometimes normal as well.  25 to 30 players hit unm; at this time (13 hours until reset) 17 have earned 2 transcendent chests, 4 have earned 2 celestial chests, 3 have earned 2 divine chests, and 1 member has earned 2 mythical chests. 

(EDIT) Just a few hours before reset now.  All 30 members have hit UNM and earned double transcendent chests.  29 have hit nm and earned double ultimate chests.  20 hae hit brutal and earned double grandmaster chests.  12 have hit hard and eared double guardian chests.  5 have hit normal and earned double knight chests.  1 member hit easy and earned douible warrior chests.  All CB levels (easy to unm) have been downed.

This thread is way to confusing.

Can I just get free stuff please.

No. And that's final.

Go to your room and think about what you did.

Absolutly you can get free stuff.

I would try the neighborhood liquor store or the bank around the corner.  Good luck!

I will preface this reply with I do not have the scientific acumen of some of the people participating in this thread now, namely Quinn and the two people behind the RSL CB bot. My background is in gambling and understanding human nature/psychology. That said:

"Hi there. Have you actually read the studies?" Is typically not a great way to lead off a conversation or debate. Especially when later on it becomes apparent you were the one who misinterpreted, not her. Bit of a piggyback on what Quinn already said to you, but perhaps in the future you can be less condescending, arrogant and presumptious about other people and what they may or may not know.  

Since you do have a scientific background, perhaps ask questions of people and test your hypothesis that way rather than making sweeping generalizations about their background and knowledge. Furthermore, perhaps don't accuse someone of being racist and demeaning them over a joke in passing. Making jokes does not preclude someone from being in a given role or job, much like being condescending likely does not with you.

Given this and the above, I have a hard time believing you don't have any sort of emotional investment in this when if her stances were so flawed, you merely could've picked them apart with points backed up by data. Quinn on the other hand was willing to look into the situation for the better part of a day despite the fact it ran counter to prior held beliefs based on what we knew about clan boss drop rates, an endeavor that provided zero value to her outside of finding out the truth about the situation. 

The professional gambler can read the thread and determine that a statistician is not needed to refute the OP theory :)

I am neither a gambler or a statistician, 

Oh right, I didn't even address the claims that I am racist for noting the data was only collected from within the Russian-speaking community. 

When doing research of any kind, noting such a variable is important.  It is a fact that this data is from Raid's Russian community, and to quote the developers of the bot from my conversation with them "All our users are from Russia (we have russian-only interface and told about bot only inside russian community). Also 99% of our users have ru-lang  as default at telegram." 

Noting this is important if you are trying to find out the whys/hows the data set has developed in such a way, is it due to a specific server in a specific region? Or could it only apply to a specific subset of the global population? And if so, why? Noting the sampling population plays a vital role in analyzing why something might be, typically in the discussion section of your work.

What's so funny about this accusation is that I absolutely adore Russian culture, literature and language. So much so that I dual-majored in Russian Literature and Biochemistry during my undergraduate degree. I am particularly fond of  The Master and Margarita, by Buglakov. Worth a read for anyone looking for a way to spend some of the holiday break.

I apologise wholeheartedly for my language and for missing the posts after the one I referred to.  I thought I had posted an apology as soon as I noticed the thread continued on another page but I may have miss posted that.  My language was indeed condescending right from the start.  In particular, I apologise to @harleQuinn .( hope that is a tag of some sort).  Even if I was taking the comments I quoted too literally and had been correct in my assumptions, it was still a totally inappropriate way to respond.  Now reading more it seems my assumptions were wrong making it an even deeper hole.  Instead of jumping at you for being a moderator, I should have stopped to think of the additional credibility that should be attached to someone in authority.  

I would like to be able to excuse my language as due to emotional investment, but there is no emotion involved other than excitement perhaps at getting such outlying results in my own account.   It was only a game to unwind with until I saw something interesting enough to start reading here.  I was maybe more frustrated at reading other people negative comments which I judged as very poor making no contribution to looking at the data.  Instead I ended up making one of the poorest of posts.

I look forward to seeing other peoples conclusions on their raw data.  Despite my rude language in previous posts, I still think it is the best data set I have been able to find on clan boss drops. It agrees with datamined drop probabilities both in the mean drop and the distribution in November and later shows a bimodal distribution.

I have liked all the responses that pointed out the above faults so they maybe remind me in the future.

I agree the bimodel distributio here is hard to explain, and the given hypothesis is reasonable.

However, for me, it's this simple.  There is a difference between;

1) Observing a phenomena, developing a hypothesis to explain such, designing an experiment to test said (inluding what data will be collected, how the data will be collected, and what criteria will be used to exclude outliers), and then collecting said data;

and 2) examining a data set, noticing a pattern, developing a hypothesis to explain such, then pointing at the pattern to support the hypothesis.

For me it's that simple.

Our clan has a rating of 1000+! We have 30 people and absolutely everyone gets the 2 top chests with 4, 5 and 6 clan bosses! and out of 30 people, 5 people received 1-2 sacrals from the bosses of the 5th and 6th clans starting from November 17, the rest receive only epic books every day, sometimes ancient fragments are found, but very rarely

Ok, let's assume they really changed the CB drop rate and give more epic books and less other stuff. I don't believe it's like that, and the collcted data is in no way a proof. Who will report to this? People that got bad drops and read in the internet that CB drops are changed. Who will not report their drops to this? People who get the same drops as allways.

But let's just assume you are right, the CB gives more epic books now and less other stuff, especially less shards. That would be asolutely fantastic for the majority of players, especially f2p players. As a nearly f2p player I could collect (only) 11 legendaries in more than a year of playing. Most champs I use are epic. Most epics I use are still unbooked.

A sacred shard would give a small chance of getting a legendary champ. Most likely it would be another epic, and unless it's Skullcrusher (don't have him) or a dupe for the guardian ring, I don't need more epic champs. But I definetly need more epic books.

If the CB drop rates would be really changed the way you describe, this would be the best thing to f2p players ever happened in Raid. Of course whales, that need legendary dupes to enhance their champs after they allready filled the guardian ring, prefer sacred shards over epic books. 

The epic books are more useful than the other drops, but you can also get them from events and tournaments.  UNM CB was the only regular supply of sacreds for F2P.  Sacred shards fare essential for completing fusion/fragment events  though.  Even keeping all fusions for the next event it has often required 4-6 sacred shards also. 

I am also FTP completely so most legos go unbooked and can be quite mediocre.  But a few lego's are almost required for higher level hard doom tower.  Also completing faction wars requires some key epics I still need, and now with guardians, almost any epic is useful for me.  So without a one regular source of 1-3 Sacreds a week for f2p is quite restrictive.

I think this can all be blamed on removal of sacred shard and chicken from daily rewards.  The conspiracy is much larger than an attack on a few f2p russians with a bad rng run.

Don't believe ? Read social networks, reddit, facebook, reviews in the google store Russian-language forum, there is also this topic! there are a lot of people writing about the fact that only epic books fall! how can you explain this? Did they all agree and just drive you to Plarium? My drop until November 17 is 5-6 sacred shards 15-20 purple and 50-60 ancient per month! So why do I only get epic books for a month? and 3-4 Ancient Shards a week and nothing else? Drought? Why, then, in a year and a half of playing that I beat 4 5 6 clan bosses every day and take from them 2 of the best chests, I have never had such a terrible drop?

Good evening Harle! I created this topic a week ago! And several times you told more than one of me that you handed everything over to the developers! Have they still not commented on this in any way? Although what I'm asking, they never admit it! Although everything suggests that the drop has changed for the worse, or rather was replaced by epic books! Do you know why? Because this is Plarium and they don't give a damn about their community! They have proven this more than once that they have only one thing in their heads, how to suck more money out of the players!