Random rolls? I think not

[krimsonrane]

For the past few days something strange has been happening to my daily dice rolls. Since VIPs get the golds daily now I have been logging into 5 servers and rolling.
I started noticing that a #3 roll popped up a lot on my first roll. But I became convinced something was afoot yesterday when all 5 servers rolled a 3 on my first attempt. ALL 5! Then again today. 4 out of 5 did the same. I usually do silver first. So I click roll, there's a slight lag pause, and boom. I get a 3.

That is impossible on a random roll.

[Suggestion]

For the past few days something strange has been happening to my daily dice rolls. Since VIPs get the golds daily now I have been logging into 5 servers and rolling.
I started noticing that a #3 roll popped up a lot on my first roll. But I became convinced something was afoot yesterday when all 5 servers rolled a 3 on my first attempt. ALL 5! Then again today. 4 out of 5 did the same. I usually do silver first. So I click roll, there's a slight lag pause, and boom. I get a 3.

That is impossible on a random roll.

I hate to disappoint you, but it is NOT impossible on a random dice roll, it is just unlikely, but with thousands of people rolling dice on the daily rolls, some will get unlikely rolls. And since it is easy to see the last rolls it is easy to track and for the one person every few months who gets several really bad rolls to complain about bad luck.

[krimsonrane]

I hate to disappoint you, but it is NOT impossible on a random dice roll, it is just unlikely, but with thousands of people rolling dice on the daily rolls, some will get unlikely rolls. And since it is easy to see the last rolls it is easy to track and for the one person every few months who gets several really bad rolls to complain about bad luck.

It's the lag pause in 9 rolls then a 3 pops up that convinced me there is more than meets the eye going on. So now when I see that lag pause I expect a 3 and it shows up.

[Weemadarthur]

It's the lag pause in 9 rolls then a 3 pops up that convinced me there is more than meets the eye going on. So now when I see that lag pause I expect a 3 and it shows up.

Please make sure all tin foil hats are firmly affixed to your head before reading further.

There is a regular conspiracy theory that if you roll your dice as soon as you log in that the dice won't actually roll and will just give a set low dice roll as the result. This would fit with your experience but if that is the case there is some good news. When you log in don't do your dice rolls immediately and instead type "/roll d100" into chat. Do that a few times and you will reset the counter and the next rolls should be different. Taking this conspiracy theory a little further there is another one that goes along the lines of if you keep rolling until you get 2 rolls of under 20 then do your daily dice roll the next roll will be a high one to compensate (normally 70+).

You may now remove your tin foil hats.

[LeoLionxxx]

It's the lag pause in 9 rolls then a 3 pops up that convinced me there is more than meets the eye going on. So now when I see that lag pause I expect a 3 and it shows up.

Quick sanity check: are you looking at the COST that's filled in after going a silver roll?

That 3 indicates the cost of an additional roll. The actual roll number appears in your chat log.

Example:
https://imgur.com/a/Rwanovo

[CSQ]

I've had that lag and rolled results other than a 3 frequently. Even this morning, I rolled a 21-40 (don't remember the roll, but remember the bracket) after experiencing the lag for my gold roll.

If you're really really worried about it, you can always wait a bit before rolling your daily dice after logging in to avoid the lag. Obviously, if you're logging on five servers, then you might be rushing and wouldn't want to wait, but hitting a vendor, running a quest, or even just waiting about a minute usually takes long enough to clear it.

And, finally, make sure that you're checking the chat log for the results. I don't want to assume you're looking at the astral shards cost, but I've had the messages disappear very quickly after having the lag on rolling, and I have misread or just missed the roll result quite a few times. It also, if I remember correctly, doesn't always properly display the result in the rolling window when it lags- not that it lists an incorrect value, but rather that it disappears very quickly as if the window refreshes.

I don't know how DDO's dice roller's internal mechanisms work, so you may just have *extremely* bad luck, but it's exceedingly rare.

[DYWYPI]

The Daily dice rolls are independent rolls; rolling a certain value on one server does not alter the chance of getting the exact same number again on either the same server or another server. It's all about fractions.

[AlcoArgo]

For the past few days something strange has been happening to my daily dice rolls. Since VIPs get the golds daily now I have been logging into 5 servers and rolling.
I started noticing that a #3 roll popped up a lot on my first roll. But I became convinced something was afoot yesterday when all 5 servers rolled a 3 on my first attempt. ALL 5! Then again today. 4 out of 5 did the same. I usually do silver first. So I click roll, there's a slight lag pause, and boom. I get a 3.

That is impossible on a random roll.

Tell you what, why don't you track every roll every day in a spreadsheet for the next 10 years and at that point we will have a valid statistical universe to analyze.

Until then, relax. Random does not mean patterns never emerge. It just means that any particular pattern is just as likely as any other pattern. 3,3,3,3,3,3,3 is just as likely as 7,6,5,4,3,2,1 and as 8,6,7,5,3,0,9.

[Verlok_the_Red]

Tell you what, why don't you track every roll every day in a spreadsheet for the next 10 years and at that point we will have a valid statistical universe to analyze.

Until then, relax. Random does not mean patterns never emerge. It just means that any particular pattern is just as likely as any other pattern. 3,3,3,3,3,3,3 is just as likely as 7,6,5,4,3,2,1 and as 8,6,7,5,3,0,9.

let me save you the time...

Classical computers can only generate pseudo-random numbers, not true random numbers. Generation of true random numbers is a feature of even the most primitive quantum computers, today, which cannot be matched by any classical computer, today, or ever.

[Krinoshi2]

Sorry to break it to ya all but there was never anything random about the dice rolls.
They are tied to a random number generator which is tied to server time.
Rolling dice at the same time on separate clients results in the same roll result. It's been that way since it came out.

And in case you aren't a complete dimwit and wonder why it is so, it's because when it's done server side you have no way to cheat the system.

Now, OP being able to get a value of 3 consistently after logging on to different servers, that's either a lie, or some lightning fast connection speeds where the whole process took less than a minute to complete.

But you will get the same number if you roll at the same minute. That's a guarantee.

[DYWYPI]

There is insufficient evidence to suggest the "Daily Dice" are rigged regarding the rolls themselves (refer to ).


[...] Applying (Pearson's chi-squared test) for that dice should be plenty enough to tell you if the dice is likely to be biased. :-)

One of the more interesting "goodness-of-fit" applications of the chi-square test is to examine issues of "fairness" and "cheating" in games of chance, such as cards, dice, and roulette. ;-)

So we need to look to see if the 'observed values' are close to the 'expected values', which I crudely mentioned in the prior post.

[...]

Robust Statistical Analysis using (Chi-squared test)

Total amount of Observations (dice rolls): 1274
Expected Frequency per any d100 single face occurrence (1274 × 1/100): 12.74

Null Hypothesis (H0): Dice is fair (unbiased)
Alternative Hypothesis (H1): Dice is unfair (biased)

Level of Significance, i.e. area in the tail: (alpha) 0.01

Degrees of Freedom (DF); Sample size (total rolls, i.e. distinct possible categories, we used 100 rows to record), minus one (N – 1): 100 – 1 = (DF): 99.

For finding the Critical value (tail of graph and rejection region area). We'll just use 99 for DF, as I don't have specialised statistics software and I'm not calculating it longhand (DF 99 is solid enough) for this dice.

Looking on an official Chi-square table using (DF 99) and (alpha 0.01) Critical value, i.e. P-value (probability): 134.642 rounded 134.64.

Therefore if our dice exceeds the (above) P-value [134.642] "Daily Dice" is likely loaded.

https://i.imgur.com/IT4eZE7.png

Chi square: X2 = sum [(Observed - Expected)2/Expected]

For example using [[post=6273711]Post #10] data: Roll 1; Observed value 12, minus Expected value: 12.74 (squared) divisible by Expected 12.74, equals: 0.042983... Then rinse and repeat, i.e. total all the X2 for 1274 rolls.

Our Test Statistic equals: 108.4176

We can easily see: 108.4176 is far less than P-value: [134.642].

Therefore we fail to reject: Null Hypothesis (H0): (we accept) Dice is fair (unbiased)

The roll results and calculations can be found within: [Post #42].

Conclusion

The "Daily dice" DO NOT appear to be loaded. In fact the statistical model shows there is an excellent chance of ~ 99% percent the dice aren't loaded. (That's what the alpha 0.01 value we tested against means basically).

Furthermore:

[...] Also the Dice roll rewards have different size ranges for example the: 100 reward, is a single figure whereas: 81-90 covers ten figures. Basically there is no need for the roll itself to be skewed or unfairly loaded because the reward ranges already have different sizes.

The statistical term 'normal distribution' doesn't mean "ordinary" it represents real-valued random variables whose distributions are not known. ...

Like I've mentioned in the past the ranges differ so there is no need whatsoever to weight the [d100] dice roll in the first place. The set of possible rewards offered may also use a different range when you roll a specific number. For example you won't always get a Specific Ability Tome on roll of: 100.

[Jerevth]

The RNG is not really random- it's a program that does a fair job given the amount of resources allowed to do that task (It's not a state of the art, top of the line sub routine- the game is 20 years old)

If you want to try and avoid a garbage roll, run a quest to shake up the RNG.

When I have gold rolls, I'll use the silver roll to test the waters; if it's good then I immediately hit the gold roll- my result is usually (Not always but more often than not) very close to the same number. I figure the RNG usually stays in a same number set. If the silver roll is rubbish, I'll do something else for a bit- quest, sort inventory (Sell random loot), etc.

It's anecdotal: I won't spend my play time building a spreadsheet to "validate" something that SSG could turn around and adjust later in their improvement process. I work with spreadsheets all day- I'm not getting paid to play so why work at it? Ignore the posters who tell you to run data sets and build spreadsheets. They haven't bothered to build the proof themselves, either.

[droid327]

There is insufficient evidence to suggest the "Daily Dice" are rigged regarding the rolls themselves.



Furthermore:


Like I've mentioned in the past the ranges differ so there is no need whatsoever to weight the [d100] dice roll in the first place. The set of possible rewards offered may also use a different range when you roll a specific number. For example you won't always get a Specific Ability Tome on roll of: 100.

I dont think anyone's suggesting the dice are rigged in the sense that the devs are deliberately weighting the code to make it less rewarding than it appears to be. People are suggesting the RNG simply isnt very good at actual randomization for some reason.

Chi-squared would not be an adequate test in this case because we're looking for non-random patterns within an otherwise random set. 132312213231 might be a statistically random set of numbers, but 111122223333 is not, even though its the same set of digits and a whole-sample test like chi-squared wouldnt be able to discriminate between them.

There's a lot of anecdotal evidence that short-term variability is not random; ie you'll get the same results several times in a row, even if the specific result you get is sufficiently random between runs

[krimsonrane]

Quick sanity check: are you looking at the COST that's filled in after going a silver roll?

That 3 indicates the cost of an additional roll. The actual roll number appears in your chat log.

Example:
https://imgur.com/a/Rwanovo

It was the actual rolls. Also, this is something that only started occurring in the past 2-3 days.

[SlowDM]

Computer tools aren’t the same as Dice rolls. Games don’t configure most random eleMents . Favourite dice, size of dice, worn edges due to use. A brand new dice isn’t the same a a 1-20 yr old dice. Table angle, force of throw, bouncing against something. Landed on the floor under the table, a multitude of real world, gravity, problems. Computers just randomly stop after milliseconds on a number. Same as the lottery . Lottery . Just swear and cuss and move on.

[krimsonrane]

I've had that lag and rolled results other than a 3 frequently. Even this morning, I rolled a 21-40 (don't remember the roll, but remember the bracket) after experiencing the lag for my gold roll.

If you're really really worried about it, you can always wait a bit before rolling your daily dice after logging in to avoid the lag. Obviously, if you're logging on five servers, then you might be rushing and wouldn't want to wait, but hitting a vendor, running a quest, or even just waiting about a minute usually takes long enough to clear it.

And, finally, make sure that you're checking the chat log for the results. I don't want to assume you're looking at the astral shards cost, but I've had the messages disappear very quickly after having the lag on rolling, and I have misread or just missed the roll result quite a few times. It also, if I remember correctly, doesn't always properly display the result in the rolling window when it lags- not that it lists an incorrect value, but rather that it disappears very quickly as if the window refreshes.

I don't know how DDO's dice roller's internal mechanisms work, so you may just have *extremely* bad luck, but it's exceedingly rare.

I'm not worried about it. Just seeing and relaying the pattern as I experienced it. 9 rolls of 3 in a row across 2 days and 5 servers. But only on the first (silver) roll of two daily. 5 rolls yesterday didn't prompt me to post. 4 more today did. Especially with that tell of a lag pause just before the 3 popped up.

[AlcoArgo]

let me save you the time...

Classical computers can only generate pseudo-random numbers, not true random numbers. Generation of true random numbers is a feature of even the most primitive quantum computers, today, which cannot be matched by any classical computer, today, or ever.

Let me save you some time... I am going to be condescending (that means I am talking down to you).

I am a software engineer and know exactly how random numbers are generated in computers and the limitations are generating a random number. One of the great jokes of computer science is that the generation of random numbers is too important to be left to chance. If you weren't so eager to try and show off the one thing you think you know and point out that computer generated random numbers aren't truly random you would have acknowledged my points that 1) the 10 or so rolls the OP is complaining about is not a statistical universe even with the worst random number generator ever created and 2) with true randomness it is still just as likely any particular pattern emerges as any other - in fact if you go on long enough it is guaranteed. Humans look for patterns, often when they aren't even there.

No matter how primitive the DDO random number generator is it is still random enough that a dozen or so rolls will not make any discernable pattern with regard to any other dozen rolls.

[Eantarus]

For the past few days something strange has been happening to my daily dice rolls. Since VIPs get the golds daily now I have been logging into 5 servers and rolling.
I started noticing that a #3 roll popped up a lot on my first roll. But I became convinced something was afoot yesterday when all 5 servers rolled a 3 on my first attempt. ALL 5! Then again today. 4 out of 5 did the same. I usually do silver first. So I click roll, there's a slight lag pause, and boom. I get a 3.

That is impossible on a random roll.

I have similarly noticed that, without fail, every single time I log in and do a DD role right away, I roll less than 20. I've never tracked the actual amount, but it is still unerringly consistent.

[dogsoldier]

I have similarly noticed that, without fail, every single time I log in and do a DD role right away, I roll less than 20. I've never tracked the actual amount, but it is still unerringly consistent.

I have noticed this as well. Almost without fail, stupidly low daily dice rolls upon 1st login. Go run a quest first, then try daily dice, much better rolls.

[boredGamer]

I have similarly noticed that, without fail, every single time I log in and do a DD role right away, I roll less than 20. I've never tracked the actual amount, but it is still unerringly consistent.

Counterpoint I always roll immediately because I’m too ocd to ignore the icon.

Rolls seems fine.

But probably it’s a login recency bias problem PLUS a user ID problem, right ? ;)

[Yvonne_Blacksword]

Was there some kind of bonus for VIP dice rolls last week or the week before?
My VIP accounts were getting 200xp on silver single digit rolls and FTP/Prem were getting 100xp. Said it in the list. Also, my gold rolls were so Plus+. Not complaining. Just want to understand.

Also, 3 runs through Saltmarsh wilds netted me 4 gloves, a shield and 4 rings(named items) on Kyber... And no quest chest got a QM chit.
4th run so far (same account) netted me zilch, but I got 5 QM chits out of 3 quests.

Also, that char. on Kyber pulled 3 vorpals in one day. Nothing I can use, Great xbow, light mace and khopesh. (Lvls3-5)

So, maybe it is server oriented? Just kidding...about the favoritism

N E 1 know if bravery bonuses affect loot==>>^^^
seems weird.

[krimsonrane]

Sorry to break it to ya all but there was never anything random about the dice rolls.
They are tied to a random number generator which is tied to server time.
Rolling dice at the same time on separate clients results in the same roll result. It's been that way since it came out.

And in case you aren't a complete dimwit and wonder why it is so, it's because when it's done server side you have no way to cheat the system.

Now, OP being able to get a value of 3 consistently after logging on to different servers, that's either a lie, or some lightning fast connection speeds where the whole process took less than a minute to complete.

But you will get the same number if you roll at the same minute. That's a guarantee.

In one comment you managed to call me a dimwit and a liar. .... people feel brave as hell when behind a keyboard.

[thwart]

I read this post and scoffed. Then I went to take my rolls ... my first roll was a 3!

[LightBear]

No tin foil hat needed.

I did some testing on this subject and can confirm that what everybody's results are on this subject are correct.
And I do mean everything from everybody.


My bias is that whenever I spend shards to roll I get better results.

[Akivasha]

Me thinks VIP are due for one free 100 roll this week! What say you all?

[krimsonrane]

I read this post and scoffed. Then I went to take my rolls ... my first roll was a 3!

lol welcome to the 3 zone!

[ahpook]

...

But you will get the same number if you roll at the same minute. That's a guarantee.

I doubt several things in this guarantee. For one that the RNG seed is based on the minute of the day. But mostly that you know enough about their RNG usage to make this statement.


lol welcome to the 3 zone!

I logged in and got a 3 on my first roll using /roll d100 ! I then quit the app and logged in again and got a 44 on my first /roll d100

[krimsonrane]

I doubt several things in this guarantee. For one that the RNG seed is based on the minute of the day. But mostly that you know enough about their RNG usage to make this statement.



I logged in and got a 3 on my first roll using /roll d100 ! I then quit the app and logged in again and got a 44 on my first /roll d100

Well, you got a 3 albeit on test mode lol

[Brandwynn]

I read this post and scoffed. Then I went to take my rolls ... my first roll was a 3!

You shouldn't have read this thread.. Neither should I have read a post on this thread.. **** poor rolls for the last couple of days. That is it.. And as I posted here my luck on the rolls will be worse too.. This thread is jinxed, I tell ya.. Jinxed!
LOL

[Raithe]

Like I've mentioned in the past the ranges differ so there is no need whatsoever to weight the [d100] dice roll in the first place. The set of possible rewards offered may also use a different range when you roll a specific number. For example you won't always get a Specific Ability Tome on roll of: 100.

Of course there wouldn't be any need to use weighted dice when you are already specifying a probability breakdown for the different rewards, that doesn't mean that incompetence might introduce a non-random or non-psuedo-random influence. It seems to me that chi-squared tests of particular data sets taken from DDO have failed to confirm randomness in the past. As a personal anecdote, it seemed at times that I would quite frequently get the same dice roll that I had just gotten, as if the RNG wasn't the problem, it was the code that was dishing out dice rolls to different instances. It has been a long time since I noticed that particular strange behavior, however, so it may have been resolved when the statistician brought it up.

In any case, it would be incompetence in coding that would be the culprit here, no one would expect intentional foul play.

[Strimtom]

omg I did my rolls today on stream and I got a 5 for both silver and gold roll! Momma mia!

[cdbd3rd]

I started to type actual words, but the thread is already hanging over both sides of the rope with it's tails tied together.

Just gonna say...

https://media4.giphy.com/media/VDmzGCWmAa50Y/200w.gif?cid=82a1493b3y9502u348bufexyz25jsuue773ig k0kwig1wclo&rid=200w.gif&ct=g

[LightBear]

In any case, it would be incompetence in coding that would be the culprit here, no one would expect intentional foul play.

What is the one thing taking up most resources in a DnD RPG... Ah dice rolls.
And then someone writes some code to make it less server intensive by just caching it and make it and have everybody pull from that same cache.

[KoobTheProud]

The dice rolls are all random.

I went VIP on one of my alt accounts for some random reason. Logged in every day and rolled for about a week and got results all over the chart, including a 98 and a 91. Then life intervened and I wasn't able to roll for a week or so. First roll after that was 100.

So then I wondered if the gold VIP rolls were being boosted somehow, since 98, 91 and 100 over a period of maybe 9 days seemed unusually good. I let it go for a few days and then rolled 20, 16 and 40-something.

It's just random.

The 100 and 98 were good rolls also. I got a +5 heart, a nice cosmetic outfit, a few K of sentient XP and a mount.

[CaptainPurge]

I rolled a 3.

and other numbers

Party Time!

[DYWYPI]

It is unlikely a specific number outcome (as originally suggested) was rolled more than twice consecutively. For example, the probability (before the event) of three consecutive rolls showing the same number: (1/100) × (1/100) × (1/100) that could be a one in 1-million chance.

One of the dice astral shard costs is 3 AS and that figure "3" appears within the graphic (where the actual "Roll!" result goes) after a roll result has completed. The chat channel however would show the correct Daily Dice roll result. The graphic display result would be brief and possibly not observed - unreliable with sufficient lag.

The Chi-square I used was a 'Goodness-of-Fit' test for a population. A population is a whole; it's every member of a group. A cryptologist investigating the random generator for errors would also use a Chi-square test; to help see if the generator was functioning correctly.

No correlation between a seed and any value generated from that seed should be evident; each element of the sequence should appear to be the outcome of an independent random event whose probability is 1/2.

Part of the reason I was also mentioned gambling was because there was evidence of Gambler's fallacy in some of the posts.

The gambler's fallacy is a cognitive bias, meaning that it's a systematic pattern of deviation from rationality, which occurs due to the way people's cognitive system works. It is primarily attributed to the expectation that even short sequences of outcomes will be highly representative of the process that generated them, and to the view of chance as a fair and self-correcting process.

Why do people act this way time and time again? We can discover intriguing insights; it seems, by recruiting monkeys and getting them to gamble too. If these animals make dumb choices like us, perhaps it could tell us more about ourselves. :-)

[KoobTheProud]

Just wanted to point out that your odds on rolling any specific trifecta are a million to one.

So 3,3,3 or 3,3,46 or 3,46,3...

[Kelledren]

I have similarly noticed that, without fail, every single time I log in and do a DD role right away, I roll less than 20. I've never tracked the actual amount, but it is still unerringly consistent.

I roll silver and gold on on every server each day as soon as I log into each server- takes 15 minutes to do all the rolls and I’m a busy guy- and my rolls are all seemingly random per the usual. I do get some laginess in some roles- sometimes it pops up in 1-20, but also pops up in 90s and 100- thought maybe had to with rolling multiple prize tables.

[krimsonrane]

I roll silver and gold on on every server each day as soon as I log into each server- takes 15 minutes to do all the rolls and I’m a busy guy- and my rolls are all seemingly random per the usual. I do get some laginess in some roles- sometimes it pops up in 1-20, but also pops up in 90s and 100- thought maybe had to with rolling multiple prize tables.

If only my 3roll 9xs run were all 20's and on gold. :)

[DYWYPI]

True biases do sometimes occur (e.g., faulty equipment), but more often an apparent bias will just be a random fluke that will not allow one to predict future events (fundamental uncertainty; independence of events).

It is important to note, that the joint probability of two events occurring refers only to events that have not happened yet. With a 6-Sided dice chances of any number coming up twice in a row are: 1/6, not 1/36. This is because there are six possible ways (opportunities) of getting the same number twice in a row: (1/6) × (1/6) × 6 = 6/36 = 1/6. "The dice have no memory".

<Off-topic>

Historic: 2000 Rolls (d20)

Incidentally I recently received a PM relating to this thread; it referenced an "old thread" regarding perceived bias with Success rates.

UMD and Scroll rolls, were also linked to within that "old thread" (UMD RNG analysis) by Dirac. Dirac, didn't actually give their calculations for: 2000 Rolls, (Degrees of Freedom: 19). Using my calculations, that UMD test statistic comes out at around: 18.2, less than P-value: [30.144] for [0.95] 95% chance of being true. Therefore we fail to reject: Null Hypothesis, (we accept) Dice is fair for those [UMD rolls].

NB: Dirac essentially divided: 18.2/19 to get that: 0.95[7] 'Reduced Chi-square' value (equal to the ratio of the observed experimental variance divided by the theoretical variance). It's just easier to visualize 0.95[7] (being near to 1). Than the actual result for; 19 Degrees of Freedom in this case 18.2.

For the Regression (Reduced Chi-squared) to be a valid size; it would require an 'Expected Frequency' of 100 and greater. For (d20) in 20 rows and 1 column (100 × 20) meaning: 2000 Rolls would be required. Whereas with Chi-squared it requires the 'Expected Frequency' to be 5 or greater thus (d20) requiring: 100 Rolls to be valid. People tend to confuse that magic "5" value, thus wrongly assuming; it requires an 'Observed frequency' greater than 5 (number of observations per cell).


Modern-day 500 Rolls (d20)

Last Thursday, I rolled myself a Tabaxi Rogue, and it said; she had a 25% chance of UMD 'Success' with: Heal Scrolls. Thus any roll of 16 or higher on a d20 would succeed. I hid in the Lamannia Test Dojo physically clicking 500 times. I wasn't testing success rates.

Results: 500 Rolls (d20) Frequency

The DDO forum isn't really suitable for displaying the formula but basically the "i" is the incident (roll); "E" is expected, and "O" is observed frequency values.


i Oi Ei Oi-Ei (Oi-Ei)2 (Oi-Ei)2/Ei
01 28 25 3 9 0.36
02 25 25 0 0 0
03 25 25 0 0 0
04 34 25 9 81 3.24
05 18 25 -7 49 1.96
06 27 25 2 4 0.16
07 29 25 4 16 0.64
08 19 25 -6 36 1.44
09 19 25 -6 36 1.44
10 26 25 1 1 0.04
11 27 25 2 4 0.16
12 29 25 4 16 0.64
13 17 25 -8 64 2.56
14 33 25 8 64 2.56
15 27 25 2 4 0.16
16 24 25 -1 1 0.04
17 29 25 4 16 0.64
18 20 25 -5 25 1
19 22 25 -3 9 0.36
20 22 25 -3 9 0.36
500 Sum 17.76

For the results of [500 rolls] the 'mean' of the "sample size" was 25 and the 'medium' 25.5 and the 'bimodal' 27 reoccurrences of a single number. Variance: 22.2 and Standard Deviation: 4.711.

There was a 23% Success result for the Heal scrolls which is fine. The test subject was only rated at 25% success chance. As it so happens there was an even split [250/250] for rolling: 1-10 and 11-20.

Results: 500 Rolls (d20) Frequency for UMD Heal Scroll

Robust Statistical Analysis using (Chi-squared test)

Total amount of Observations (dice rolls): 500
Expected Frequency per any d20 single face occurrence (500 × 1/20): 25

Null Hypothesis (H0): Dice is fair (unbiased)
Alternative Hypothesis (H1): Dice is unfair (biased)

Level of Significance, i.e. area in the tail: (alpha) 0.01

Degrees of Freedom (DF); Sample size (total rolls, i.e. distinct possible categories, we used 20 rows to record), minus one (N – 1): 20 – 1 = (DF): 19.

For finding the Critical value (tail of graph and rejection region area). We'll just use 19 for DF, as I don't have specialised statistics software and I'm not calculating it longhand (DF 19 is solid enough) for this dice.

Looking on an official Chi-square table using (DF 19) and (alpha 0.01) Critical value, i.e. P-value (probability): 36.191 rounded 36.19.

Therefore if our dice exceeds the (above) P-value [36.191] "d20 Dice" is likely loaded.

https://i.imgur.com/IT4eZE7.png

Chi square: X2 = sum [(Observed - Expected)2/Expected]

For example using data: Roll 1; Observed value 28, minus Expected value: 25 (squared) divisible by Expected 25, equals: 0.36... Then rinse and repeat, i.e. total all the X2 for 500 rolls.

Our Test Statistic equals: 17.76

We can easily see: 17.76 is far less than P-value: [36.191].

Therefore we fail to reject: Null Hypothesis (H0): (we accept) Dice is fair (unbiased)

Conclusion

The "dice" do not appear to be loaded. In fact the statistical model shows there is an excellent chance of ~ 99% percent the dice aren't loaded. (That's what the alpha 0.01 value we tested against means basically).

There is insufficient evidence to suggest the "Dice" are weighted regarding the rolls themselves. We've already rolled enough. If someone wants to supply 1000s of rolls feel free.


Discussion

Things do not have to even out, but sometimes seem to, as more observations are added (law of large numbers).

Law of Large Numbers: As the sample size increases the average of the actual outcomes will more closely approximate the mathematical probability.

</Off-topic>

I do believe it's possible the 'Daily Dice' Random Number Generator could theoretically have artefacts but still you'd need a cryptographer. Or do severe stress tests to overload (or compromise) the DDO program, server and so forth. (If I hadn't received that PM, I wouldn't have bothered posting the "Off-topic" section).

SSG are not going to freely give you the Seed value or a special 'reserved number' that never ever gets output by the generator, e.g. 000000. ;-)

See: https://en.wikipedia.org/wiki//dev/random

[CaptainPurge]

Holy cow folks, I did it! I proved the conspiracy correct!

https://i.gyazo.com/b0ad47187ed482d3e0d5d283d6c6cabf.png

[Eantarus]

I roll silver and gold on on every server each day as soon as I log into each server- takes 15 minutes to do all the rolls and I’m a busy guy- and my rolls are all seemingly random per the usual. I do get some laginess in some roles- sometimes it pops up in 1-20, but also pops up in 90s and 100- thought maybe had to with rolling multiple prize tables.

Has it occurred to you that an issue might affect some players, and not others?

This has been a big part of what makes it so difficult to convince anyone of the bug. I can reliably reproduce statistically unlikely events every time I sit down at the game(such as rolling the exact same number 3 times in a row on daily dice) but others cannot. Ergo, whatever bug is affecting me, but possibly not you.

This is very similar to the wi flag problem from Asheron's Call(from which DDO derives the lion's share of it's code). AC players went through the exact same thing: those impacted by the bug(and it was not everyone) spent literal years just trying to be taken seriously, and years longer waiting for a fix. Eventually the problem was solved.

But we can't even start solving it until forumites stop shouting down everyone who complains.

[boredGamer]

Has it occurred to you that an issue might affect some players, and not others?

This has been a big part of what makes it so difficult to convince anyone of the bug. I can reliably reproduce statistically unlikely events every time I sit down at the game(such as rolling the exact same number 3 times in a row on daily dice) but others cannot. Ergo, whatever bug is affecting me, but possibly not you.

This is very similar to the wi flag problem from Asheron's Call(from which DDO derives the lion's share of it's code). AC players went through the exact same thing: those impacted by the bug(and it was not everyone) spent literal years just trying to be taken seriously, and years longer waiting for a fix. Eventually the problem was solved.

But we can't even start solving it until forumites stop shouting down everyone who complains.

So just record your actual data.

[Weemadarthur]

So just record your actual data.

This will have no effect at all. In another thread on the subject I actually posted my findings on several thousand rolls along with positing a plausible theory to why clumping takes place (not saying my theory was correct just saying it was plausible and if correct those results made perfect sense). Immediately someone posted several thousand rolls is too small a sample size and several million would be needed. Now I would be willing to bet my last dollar that if I posted a million rolls someone would immediately say that the sample size was still too small and billions would be needed. In short (and tbh I don't think those saying this are completely wrong) no sample size will ever be big enough as random is random and any person doing the same test could in theory get completely different results.

The bigger problem though is most players still don't understand how random works. If you tell them there is a 10% chance they will drop an item they will complain if they don't get it in 10 pulls. The fact is though on every roll the chance of them dropping that item is still 10%. None of the rolls previous to that single roll hold any value. Every roll is and always will be the same 10% chance. The player is just as likely to pull the same item 6 times as pull 6 different items. Combine that with selection bias and human natures fondness for trying to make things follow a narrative pattern and it becomes incredibly difficult to prove anything.

[DYWYPI]

True biases do sometimes occur (e.g., faulty equipment), but more often an apparent bias will just be a random fluke that will not allow one to predict future events (fundamental uncertainty; independence of events).

It is important to note, that the joint probability of two events occurring refers only to events that have not happened yet. With a 6-Sided dice chances of any number coming up twice in a row are: 1/6, not 1/36. This is because there are six possible ways (opportunities) of getting the same number twice in a row: (1/6) × (1/6) × 6 = 6/36 = 1/6. "The dice have no memory".

People will often judge the coin-tossing sequence of; H, H, H, H, H, H as being less random than; H, T, H, H, T, H, even though the probability of obtaining each of these given sequences is identical: 1/2 x 1/2 × 1/2 × 1/2 × 1/2 × 1/2 = 0.015625.

[LightBear]

I have similarly noticed that, without fail, every single time I log in and do a DD role right away, I roll less than 20. I've never tracked the actual amount, but it is still unerringly consistent.

Same here, so yeah you are not alone.

On note of arbitrary dice rolls, I get a teleport (spell) failed waaaay more often than the noted 3%.

[Steeme]

Why not ask the OP if he's still rolling a 3 on every first roll on every server? Enough time has gone by, he can declare whether it was an anomaly or double-down on the original claim.

[boredGamer]

This will have no effect at all. In another thread on the subject I actually posted my findings on several thousand rolls along with positing a plausible theory to why clumping takes place (not saying my theory was correct just saying it was plausible and if correct those results made perfect sense). Immediately someone posted several thousand rolls is too small a sample size and several million would be needed. Now I would be willing to bet my last dollar that if I posted a million rolls someone would immediately say that the sample size was still too small and billions would be needed. In short (and tbh I don't think those saying this are completely wrong) no sample size will ever be big enough as random is random and any person doing the same test could in theory get completely different results.

The bigger problem though is most players still don't understand how random works. If you tell them there is a 10% chance they will drop an item they will complain if they don't get it in 10 pulls. The fact is though on every roll the chance of them dropping that item is still 10%. None of the rolls previous to that single roll hold any value. Every roll is and always will be the same 10% chance. The player is just as likely to pull the same item 6 times as pull 6 different items. Combine that with selection bias and human natures fondness for trying to make things follow a narrative pattern and it becomes incredibly difficult to prove anything.

Right and everyone in this thread is saying they have bias but it’s exhibiting in different ways ?

And then you go on to explain that bias but don’t think it applies to you ? Did you do a statistical analysis on your data ? What did it show exactly ?

[Weemadarthur]

Right and everyone in this thread is saying they have bias but it’s exhibiting in different ways ?

And then you go on to explain that bias but don’t think it applies to you ? Did you do a statistical analysis on your data ? What did it show exactly ?

1st up what I was trying to say is nothing about anyone being right or wrong but rather that telling people to record data when no-one will ever be able to record enough data to give a result that anyone will accept is pointless.

As for my previous experiment if you want to have a read through it's in this thread :- https://forums.ddo.com/forums/showthread.php/531025-Raven-at-the-Door-Loot-Strangeness.

Now as I said in my last post this is just a plausible explanation of why loot clumping takes place due to the psuedo random nature of computers rather than actual random. I am not claiming it's correct, I am not claiming to be right and others to be wrong. I am just saying if the theory is correct that the findings then do make sense. I do however also agree that the sample sizes I used are still too small to get anything definitive from them, that no matter how hard anyone (including myself) tries observational bias will always be in effect and the fact that we don't have the slightest clue as to how rolls are made in DDO (are they server based and we all share a pool or are they individual rolls per account for example) any evidence we do present (including mine) is at best only a small portion of the data needed to draw a conclusion from. That ofc is without going down the rabbit hole of did I make it all up on the fly to push an agenda as lets be honest here you don't know me and have no reason to take my word on face value for any of this lol.

Now all I can say for myself is I try (I may not always succeed but I do try) to be as impartial and objective as possible with all my posts. I may be a disagreeable old fart but I never presume to know better. What I offered in the above thread is an explanation that may be correct not definitive evidence that I am right and anyone who disagrees is wrong.

Now with all that taken into account how much data do you think Eantarus should record before any conclusion can be drawn from it? If thousands of rolls aren't enough data (and I don't disagree there) what amount would be? Or are you willing to concede that maybe asking others to supply data which is based on unknown parameters would really have no effect at all and all in all be an exercise in pointlessness.

[droid327]

I'm at the stage in my EK/PM life right now where I suck up 10% ASF...and, as I always am at this point, I'm struck by how often I'm rolling <10 on my ASF checks. Particularly for Death Aura. I dont have to cast it that often during a quest, but even then I'm often getting 2-4 fizzles per quest when it really should be like 1 fizzle every other quest...

[DANTEIL]

One of my favorite bits of trivia is the fact that Spotify initially had their 'shuffle' feature using something approximating a "true" randomness algorithm, but they received so many complaints from users who didn't like that Spotify was shuffling too many songs by the same artist in a row etc., that they ended up changing the algorithm to be only quasi-random, eliminating runs of similar artists/songs. (link (https://thetab.com/uk/2021/11/17/spotify-shuffle-explained-228639)). Apple iTunes went through the same thing.


source: I teach about psychological beliefs about randomness and probability, among other things

[Kelledren]

Has it occurred to you that an issue might affect some players, and not others?

This has been a big part of what makes it so difficult to convince anyone of the bug. I can reliably reproduce statistically unlikely events every time I sit down at the game(such as rolling the exact same number 3 times in a row on daily dice) but others cannot. Ergo, whatever bug is affecting me, but possibly not you.

This is very similar to the wi flag problem from Asheron's Call(from which DDO derives the lion's share of it's code). AC players went through the exact same thing: those impacted by the bug(and it was not everyone) spent literal years just trying to be taken seriously, and years longer waiting for a fix. Eventually the problem was solved.

But we can't even start solving it until forumites stop shouting down everyone who complains.

Hmmm… I don’t believe I was shouting down complaints but rather giving my experience on the subject. As a VIP and a past season pass holder (with either OCD and/or just cheapness) I’ve rolled a lot of dice on all the servers. There might be something to the lag in the first minute or two you log in, but it’s hard to tell if it’s a real thing. I’ve had high rolls in that early lag period too, but a lot of times it seems low. It could be perception because those low dice tiers are larger than the others.

I actually checked into this thread to see if anyone had a higher rate of 100 rolls this weekend when the double dice was going. I know that it is only xp, but I rolled 100 four times this weekend and in the 90s a bit more than that. Todays rolls were all <20 for the most part. It’s weird but could still be just randomness. As for roll=time on server, it could work that way, but could rotate through those numbers faster than you can swap servers. We would have to get 20-50 people or so to all roll at the exact same time (error for connection speeds) and check the distribution. Plus how do the additional sub-rolls work in that scenario?

[Theolin]

The mere observance of randomness alters it
and randomness is not really random
and humans do not understand randomness
and nature does not do random
and computers have no idea what random is nor do people know how to tell them
....
that being said there does seem to be something different between my different characters in how DDO's random works
though that could just be my confirmation bias and not really reality ... but ...

[LightBear]

Random doesn't exist, you just need to figure out where in the sequence of things you are and how you can influence that scheme to work in your advantage.

[DYWYPI]

Simon Pampena, did a humorous '20 coin flip test' on a friend asking his friend try to; imagine and create 20 'random flips' in his mind, and Simon bet he could read his friend's mind... The video is about 12 minutes long (the final few minutes are sponsor adverts or something).

Randomness is Random (video): https://www.yout-ube.com/watch?v=tP-Ipsat90c


You wouldn't need anywhere near "1 million rolls" for checking bias for a 100-Sided dice; you would likely need a bare minimum of 500 rolls for it to be valid. Typically a few thousand rolls would be used for a 'Test of Independence'. For a 'Goodness of Fit' for the 100-Sided dice, you'd likely need 10,000 rolls. I've already given actual recorded results and worked examples of d100 rolls within this thread.

We are talking about the 100-Sided dice (face) distribution itself, and NOT any reward system or the pseudo random pattern or subroutine.

The d100 dice rolls themselves are not loaded. The rewards table albeit involves a different roll.

[Eantarus]

Same here, so yeah you are not alone.

On note of arbitrary dice rolls, I get a teleport (spell) failed waaaay more often than the noted 3%.

Saaaaaaaame. I once had it fail on me 3 times in a row. My rate seems to be about 1 in 10.



This will have no effect at all. In another thread on the subject I actually posted my findings on several thousand rolls along with positing a plausible theory to why clumping takes place (not saying my theory was correct just saying it was plausible and if correct those results made perfect sense). Immediately someone posted several thousand rolls is too small a sample size and several million would be needed.

This is patently false. You can derive useful information from small sample sizes, just with less certainty than you get from larger ones. Its all a question of how much certainty you want. If you're happy with 95% certainty(which is pretty darn reasonable for a video game) then you can that with a few thousand dice roles easy.

What is your theory? I would love to share.



Now I would be willing to bet my last dollar that if I posted a million rolls someone would immediately say that the sample size was still too small and billions would be needed. In short (and tbh I don't think those saying this are completely wrong) no sample size will ever be big enough as random is random and any person doing the same test could in theory get completely different results.

Yeah that's the big problem. The amateur statisticians running rampant on these boards think you need to get 100% certainty. You just don't. 90% should be more than plenty to prove your point.



The bigger problem though is most players still don't understand how random works. If you tell them there is a 10% chance they will drop an item they will complain if they don't get it in 10 pulls. The fact is though on every roll the chance of them dropping that item is still 10%. None of the rolls previous to that single roll hold any value. Every roll is and always will be the same 10% chance. The player is just as likely to pull the same item 6 times as pull 6 different items. Combine that with selection bias and human natures fondness for trying to make things follow a narrative pattern and it becomes incredibly difficult to prove anything.

Yes, each 10% is a different chance. However, if you go 200 pulls without seeing a single item, that's pretty strong evidence that the odds are not 10%.

I did some highly interesting work on feytwisted chests using the one in Make Believe that you can see is feytwisted without fighting any mobs. Supposedly the drop rate on feytwisted chests is somewhere around 2%, or 1 in 50. I did runs on R10 to maximize the odds. It took 170 trips to see the first feytwisted chest, 150 to see the second, and 96 to see the third. After that, they started dropping like clockwork every 25-30 runs, with a few happening as frequently as 4 or 5 runs apart.

My point is and always has been: yes, clumps in RNG are normal, but these kinds of clumps make for a terrible gameplay experience.

[LightBear]

Yes, each 10% is a different chance. However, if you go 200 pulls without seeing a single item, that's pretty strong evidence that the odds are not 10%.

I did some highly interesting work on feytwisted chests using the one in Make Believe that you can see is feytwisted without fighting any mobs. Supposedly the drop rate on feytwisted chests is somewhere around 2%, or 1 in 50. I did runs on R10 to maximize the odds. It took 170 trips to see the first feytwisted chest, 150 to see the second, and 96 to see the third. After that, they started dropping like clockwork every 25-30 runs, with a few happening as frequently as 4 or 5 runs apart.

My point is and always has been: yes, clumps in RNG are normal, but these kinds of clumps make for a terrible gameplay experience.

Could be that the game is keeping score of that chest being fey server wide, so if multiple people are doing the same party trick as you it could be that it is you that winds up with the short end of the stick and someone else getting lucky multiple times in a row.

And yeah, clumps of rolling a one could occur, I'm just tired to always be the one that gets them.

The more reason to have fail safe mechanics in play, like named items on the reward list of the arc.
Which Fey has btw.

Would still like to see named items appear on the reward lists.

[Weemadarthur]

What is your theory? I would love to share.

OK here's a brief outline of my theory of what causes clumping in loot rolls. There is a link in my previous post to a thread where I go into a little more depth on my findings and explain why and what I was actually trying to accomplish.

My theory is that although loot rolls on paper are equally distributed due to the psuedo random effects in play of the random generator used this isn't what happens in practice. I think that over a period of time rolls tend to favor numbers nearer the center of the spread so if you roll a d/100 100 times you will see a pattern that more numbers appear in the 30-70 range than those on either side of that bracket. If you then apply that same result to the chances of 5 pcs of loot dropping you would then end up with 1 piece that drops more often (the 40-60 roll) that then lowers as you get to the outside of the range. If correct that would reinforce the fact that players frequently see 1 item drop more than others, 2 items drop fairly frequently and 2 items that become hard to pull.

Now to be very clear here I think there are several flaws in my own theory so am not putting this forward as fact. The biggest issue Bjond raised in that other thread and that is we have no idea at all how rolls are determined. Do we all share one pool of numbers, are rolls shared across each server or are all rolls individual? Depending on what the answer to that question is will completely change how results can be perceived. Then take into account that some games have algorithms in place that reward spending and as much as we would like to think that isn't so here the fact is we just don't know. Lastly my results are just those of 1 player. A second player could have vastly different results that would in effect disprove my hypothesis.

All in all although I think the results from my testing do seem to back up my own theory I would not in any way claim that it is correct. I would genuinely go as far as to say without the knowledge of exactly how rolls are made that the best we could draw from this is that it's definitely possible that maybe I might be right. This is why I stated above that I don't think recording data will help to any large degree. Without a large sample group running a large sample size there are just too many unknown variables at play.

Still I do think it is a good theory that if correct does make sense of what a lot of players have seen. As there is no way to prove or disprove it we can just appreciate it for what it is and agree or disagree with it as we all see fit.

[Sylvado]

let me save you the time...

Classical computers can only generate pseudo-random numbers, not true random numbers. Generation of true random numbers is a feature of even the most primitive quantum computers, today, which cannot be matched by any classical computer, today, or ever.

Actually there is no such thing as random because everything has a cause. However, it is very simple to create a value that is unpredictable for all practical purposes. My Timex computer some 40 years ago just had a counter connected to the horizontal frequency oscillator.

[Eantarus]

Could be that the game is keeping score of that chest being fey server wide, so if multiple people are doing the same party trick as you it could be that it is you that winds up with the short end of the stick and someone else getting lucky multiple times in a row.

Plausible. But if that's true it creates this problem:



And yeah, clumps of rolling a one could occur, I'm just tired to always be the one that gets them.

Which regardless of anything else is at the core of the issue. Some players are being very negatively affected by the clumping and further negatively affected by the gaslighting which goes on all over the forum. It's not all in your head.



The more reason to have fail safe mechanics in play, like named items on the reward list of the arc.
Which Fey has btw.

Would still like to see named items appear on the reward lists.

The game desperately needs some kind of failsafe. I'm presently trying to get a Legendary Tail of the Scorpion out of Toxic Treatment. I've done about 20 chest pulls so far, and am expecting it to take about upwards of 200.



OK here's a brief outline of my theory of what causes clumping in loot rolls. There is a link in my previous post to a thread where I go into a little more depth on my findings and explain why and what I was actually trying to accomplish.

My theory is that although loot rolls on paper are equally distributed due to the psuedo random effects in play of the random generator used this isn't what happens in practice. I think that over a period of time rolls tend to favor numbers nearer the center of the spread so if you roll a d/100 100 times you will see a pattern that more numbers appear in the 30-70 range than those on either side of that bracket. If you then apply that same result to the chances of 5 pcs of loot dropping you would then end up with 1 piece that drops more often (the 40-60 roll) that then lowers as you get to the outside of the range. If correct that would reinforce the fact that players frequently see 1 item drop more than others, 2 items drop fairly frequently and 2 items that become hard to pull.

This is a good theory. Putting it together with what I know and what I've observed, I think it works out like so:

Let's say a quest has 4 named items that drop according to the 10/15/33% mechanic. Each item is weighted equally(per dev post). The loot table looks something like this:
Item 1: 1-25
Item 2: 26-50
Item 3: 51-75
Item 4: 76-100

Now, if there is a bias in the system as you suspect, then that would make items 2 and 3 appear much more frequently. So your theory jives with what the devs have told us about how loot mechanics work.

Personally, based on observing DD roles I think the bias is more towards low numbers. E.G. I roll a 1-20 far more often than the laws of probability would dictate in any balanced system. Applying this logic to quest loot explains why one item will appear in huge numbers and others will appear more reasonably. You can really see this effect in quests with more than 4 items. Even with the bug on daily dice I still occasionally see a 75 or higher but very very scarcely anything into the 80s or 90s. Apply this to a quest like Reach for the Sky with 6 named items in the end chest: each named item only gets about 16 spaces on loot table, if the Enigma Core occupies spaces 84-100, that would explain why it took me 2,000 chest-pulls to get one(not a lie, no a joke, 100% serious).




Now to be very clear here I think there are several flaws in my own theory so am not putting this forward as fact. The biggest issue Bjond raised in that other thread and that is we have no idea at all how rolls are determined. Do we all share one pool of numbers, are rolls shared across each server or are all rolls individual? Depending on what the answer to that question is will completely change how results can be perceived. Then take into account that some games have algorithms in place that reward spending and as much as we would like to think that isn't so here the fact is we just don't know. Lastly my results are just those of 1 player. A second player could have vastly different results that would in effect disprove my hypothesis.

This all comes back to the point I hope I've hammered home by now: it doesn't actually matter how the mechanics are working on the back-end if its resulting in a very negative gameplay experience for some of the players.

Let's just for fun compare this to a hypothetical experience. Let's say you go to a restaurant and the waiter screws up your order. The manager tells you "Oh, I'm sorry. He had 30 other tables and didn't mess up any of them, you just got unlucky". Ok, no big deal, it happens. Then you go back the next week, and this time its the chef who screws up your order. The manager says "Oh, sorry; the chef fixes 300 meals a night and he got the rest of them ok, you just got unlucky." Then the next week you go back yet again and the valet scratches your car.

In my hypothetical example, no meant you any harm, and everyone did their job right for the rest of the night. You just got hit with a string of bad luck. The big difference between my example and DDO: if that happened to you at a restaurant, the manager would do something to make it right before you took your business elsewhere.

I used to be in a guild with 14 other friends. I now play the game alone. All of them quit one by one because of the issue with item drop rates.

[Eantarus]

Actually there is no such thing as random because everything has a cause. However, it is very simple to create a value that is unpredictable for all practical purposes. My Timex computer some 40 years ago just had a counter connected to the horizontal frequency oscillator.

Pseudorandom is plenty random enough for the casual applications we're asking for here. I've never understood by the amateur statisticians all over these boards like to trot out that "computers can't make random numbers" "fact"(sarcastic air quotes) as if its at all relevent.

In any case its my belief that the bug does not lie in the random number generator itself. Any first-week programmer can write a random number generator from scratch that's random enough for practical purposes.

[Sylvado]

Pseudorandom is plenty random enough for the casual applications we're asking for here. I've never understood by the amateur statisticians all over these boards like to trot out that "computers can't make random numbers" "fact"(sarcastic air quotes) as if its at all relevent.

In any case its my belief that the bug does not lie in the random number generator itself. Any first-week programmer can write a random number generator from scratch that's random enough for practical purposes.

Exactly, it only gets complicated if you didn't want it to be pseudorandom.

[amessi1]

As a pseudo math geek , computer nerd, that gets to play with large data sets, this has been a fun thread to read. I've always kind of playfully imagined there was some type of AI or algorithmic logic governing the rolls. Of course we're talking about an old game that doesn't need fancy algorithms, so like the professor was telling us, it might be a bit like the Spotify Playlist, wherein certain situations there is some history kept in calculating the dice, ideally a temp cache. So for certain stages of the game's development, there may have been times where this history became stale or can become stale or was provided by devs in order to solve something else, meaning this could be attributable to bugs or halo, but is to intermittent to make a persuasive argument.

We're talking about a game that wants to emulate a dungeon master, so the idea of some type of algorithmic adjustments to the dice rolls, seems to fit the this game.

Although what would really Tickle My Fancy, is the ability to filter out the junk Rewards from the daily dice that I delete anyway.

Also, while I gather the context of RNG, can someone verbosly tell me what it means?

And despite my bias toward science, I may start trying some warm up /roll d100...tin foil helmet of playfulness equipped!

[Eantarus]

Also, while I gather the context of RNG, can someone verbosly tell me what it means?


RNG = Random Number Generation - you mean that?

[amessi1]

RNG = Random Number Generation - you mean that?

Doh, that makes sense. I hate acronyms...I thought it was the name of app/method that randomly generated numbers...lolz

[Weemadarthur]

And despite my bias toward science, I may start trying some warm up /roll d100...tin foil helmet of playfulness equipped!

If you do try this would you mind coming back and letting us know if it works. I have had pretty positive results from the exercise but would appreciate having a 2nd set of results to compare with even if it means I have to remove my foil hat (which I think looks quite dashing lol).

[GODDEATH]

for awhile i even ran 1 quest before rolling dice, then decided it was silly to do so.
logged in today and pow 10000Xp a Nice new shiny armor kit and a fate tome. upon login no waiting.
last week a +6 cha tome on another toon on login no waiting
i do alternate toons who roll for the day though. just because i dont know why.

please add a free dice in treasure (loot a dice get a free roll!) al chests anywhere anytime. FUN.

[Eantarus]

Doh, that makes sense. I hate acronyms...I thought it was the name of app/method that randomly generated numbers...lolz

Yeah, I hate acronyms to. Especially a problem as an MMO gamer.


If you do try this would you mind coming back and letting us know if it works. I have had pretty positive results from the exercise but would appreciate having a 2nd set of results to compare with even if it means I have to remove my foil hat (which I think looks quite dashing lol).

Its actually really easy to reproduce the kind of loot problems some of us are seeing. Just pick any modern, short quest and ransack it on a couple characters. I guarantee in 20 or 30 chest pulls you'll see it.

[Eantarus]

for awhile i even ran 1 quest before rolling dice, then decided it was silly to do so.


I've mentioned this a couple of times but these bugs don't seem to be affecting all players universally. I consistently get a 19 or lower on every single DD role if I do it upon logging in.

[Weemadarthur]

Its actually really easy to reproduce the kind of loot problems some of us are seeing. Just pick any modern, short quest and ransack it on a couple characters. I guarantee in 20 or 30 chest pulls you'll see it.

I know 1st hand about the loot clumping issue and have encountered it on a regular basis (I have posted my findings in a few different posts over the last few years). The post you quoted though was regarding a different issue involving daily dice rolls being low when you 1st log in and a potential way to avoid that. I think it was my 1st post in this thread where I put forth an often quoted conspiracy theory based on doing random rolls before your daily dice roll to break that cycle. I was just curious if Amessi did give it a try whether they would mind letting me know if their results improved or not.

At the end of the day it's just a few extra seconds to copy and paste /roll d100 in chat a few times before doing the daily dice roll so if anyone else wants to give it a go and let me know if their results improve or not feel free to test yourselves and PM me the results. If I get enough data I will update on whether it works for the majority or not.

[Eantarus]

I know 1st hand about the loot clumping issue and have encountered it on a regular basis (I have posted my findings in a few different posts over the last few years). The post you quoted though was regarding a different issue involving daily dice rolls being low when you 1st log in and a potential way to avoid that. I think it was my 1st post in this thread where I put forth an often quoted conspiracy theory based on doing random rolls before your daily dice roll to break that cycle. I was just curious if Amessi did give it a try whether they would mind letting me know if their results improved or not.

Yeah we've been discussing a lot of issues with the rangen appearing decidedly non-random at times.



At the end of the day it's just a few extra seconds to copy and paste /roll d100 in chat a few times before doing the daily dice roll so if anyone else wants to give it a go and let me know if their results improve or not feel free to test yourselves and PM me the results. If I get enough data I will update on whether it works for the majority or not.

I actually tried this and could observe no correlation. Was getting the same low roles no matter what. For me the only thing that ensures reasonably random dice roles is playing/being online for a while. I get the best results immediately after a quest completion(and by "best" I mean "most random").

A friend of mine before quitting did about 50 roles to use up his remaining astral shards. Just logged in and rolled 'em. Of the 50 roles, 29 of them were 19 or less. Really sucked because he wanted to level his character back to 20 one last time.

[Weemadarthur]

I actually tried this and could observe no correlation. Was getting the same low roles no matter what. For me the only thing that ensures reasonably random dice roles is playing/being online for a while. I get the best results immediately after a quest completion(and by "best" I mean "most random").

A friend of mine before quitting did about 50 roles to use up his remaining astral shards. Just logged in and rolled 'em. Of the 50 roles, 29 of them were 19 or less. Really sucked because he wanted to level his character back to 20 one last time.

I don't mean to sound dismissive here but when you start a line with an estimated amount of rolls and then finish with a specific number on that estimated amount it doesn't really help your credibility. I am starting to find it hard to believe that your not letting your bias take over the narrative.

[Artos_Fabril]

Here's a quick mental exercise:
If you do 100 random d100 rolls, what do you think is your chance to roll 100 at least once?
If you don't roll 100 in those attempts, what do you think you chance of rolling 100 on the next roll is?

scroll for answers





























~63%
1%
In order to have a 99% chance to roll 100 (or any other specific number) at least one time, you'd have to roll 458 times. (actually a 98.98% chance, but it's the closest possible number of rolls to 99%)

[Sylvado]

True for the chance in 101 rolls but each roll, even the 101st, the chance is 1:100 regardless of any previous results.

[DYWYPI]

Since the "Daily dice" has 100 sides; using a simple rule of thumb, you'd need around five times that before it even was worth doing any useful statistical analysis on that dice, e.g. 500 rolls. Similar goes for the: Roll /d100. Around 2000 rolls would be a more desirable value, and 10,000 would allow regression.

Events are just plain erratic (fundamental uncertainty). Random events are often described as 'clumpy' because clumps of 'wins' or 'losses' sometimes occur.

Nothing is certain; nothing is ever due to happen (independence of events).

Rolling the Simulated [/roll d100] 100-Sided dice only 100 times won't really give significant results.

For pure amusement; today I decided to: [roll /d100], 500 times, since 100 rolls would be wasting my time... A lot of the numbers within the first 100 were higher numbers. It wasn't until the mid 400s all faces were observed at least once.


Modern-day 500 Rolls [Simulated /Roll d100] (d100)

Results: 500 Rolls (d100) Frequency

The DDO forum isn't really suitable for displaying the formula but basically the "i" is the incident (roll); "E" is expected, and "O" is observed frequency values.


i Oi Ei Oi-Ei (Oi-Ei)2 (Oi-Ei)2/Ei
01 5 5 0 0 0
02 4 5 -1 1 0.2
03 6 5 1 1 0.2
04 1 5 -4 16 3.2
05 8 5 3 9 1.8
06 10 5 5 25 5
07 4 5 -1 1 0.2
08 4 5 -1 1 0.2
09 7 5 2 4 0.8
10 7 5 2 4 0.8
11 7 5 2 4 0.8
12 6 5 1 1 0.2
13 3 5 -2 4 0.8
14 9 5 4 16 3.2
15 4 5 -1 1 0.2
16 13 5 8 64 12.8
17 5 5 0 0 0
18 2 5 -3 9 1.8
19 7 5 2 4 0.8
20 11 5 6 36 7.2
21 3 5 -2 4 0.8
22 4 5 -1 1 0.2
23 7 5 2 4 0.8
24 5 5 0 0 0
25 3 5 -2 4 0.8
26 9 5 4 16 3.2
27 6 5 1 1 0.2
28 2 5 -3 9 1.8
29 1 5 -4 16 3.2
30 3 5 -2 4 0.8
31 4 5 -1 1 0.2
32 5 5 0 0 0
33 2 5 -3 9 1.8
34 9 5 4 16 3.2
35 5 5 0 0 0
36 6 5 1 1 0.2
37 6 5 1 1 0.2
38 3 5 -2 4 0.8
39 3 5 -2 4 0.8
40 2 5 -3 9 1.8
41 4 5 -1 1 0.2
42 4 5 -1 1 0.2
43 5 5 0 0 0
44 6 5 1 1 0.2
45 3 5 -2 4 0.8
46 4 5 -1 1 0.2
47 4 5 -1 1 0.2
48 5 5 0 0 0
49 2 5 -3 9 1.8
50 5 5 0 0 0
51 6 5 1 1 0.2
52 4 5 -1 1 0.2
53 4 5 -1 1 0.2
54 4 5 -1 1 0.2
55 4 5 -1 1 0.2
56 5 5 0 0 0
57 3 5 -2 4 0.8
58 6 5 1 1 0.2
59 7 5 2 4 0.8
60 4 5 -1 1 0.2
61 4 5 -1 1 0.2
62 3 5 -2 4 0.8
63 3 5 -2 4 0.8
64 6 5 1 1 0.2
65 4 5 -1 1 0.2
66 5 5 0 0 0
67 7 5 2 4 0.8
68 6 5 1 1 0.2
69 3 5 -2 4 0.8
70 5 5 0 0 0
71 4 5 -1 1 0.2
72 7 5 2 4 0.8
73 4 5 -1 1 0.2
74 2 5 -3 9 1.8
75 7 5 2 4 0.8
76 1 5 -4 16 3.2
77 4 5 -1 1 0.2
78 3 5 -2 4 0.8
79 4 5 -1 1 0.2
80 6 5 1 1 0.2
81 6 5 1 1 0.2
82 2 5 -3 9 1.8
83 5 5 0 0 0
84 5 5 0 0 0
85 6 5 1 1 0.2
86 3 5 -2 4 0.8
87 6 5 1 1 0.2
88 5 5 0 0 0
89 6 5 1 1 0.2
90 8 5 3 9 1.8
91 2 5 -3 9 1.8
92 8 5 3 9 1.8
93 2 5 -3 9 1.8
94 9 5 4 16 3.2
95 8 5 3 9 1.8
96 7 5 2 4 0.8
97 3 5 -2 4 0.8
98 7 5 2 4 0.8
99 9 5 4 16 3.2
100 5 5 0 0 0
500 Sum 101.2

For the results of [500 rolls] the 'mean' of the "sample size" was 5 and the 'medium' 5 and the 'modal' 13 reoccurrences of a single number. Variance: 5.06 and Standard Deviation: 2.249...

Results: 500 Rolls (d100) Frequency for Simulated: Roll /d100

Robust Statistical Analysis using (Chi-squared test)

Total amount of Observations (dice rolls): 500
Expected Frequency per any d100 single face occurrence (500 × 1/100): 5

Null Hypothesis (H0): Dice is fair (unbiased)
Alternative Hypothesis (H1): Dice is unfair (biased)

Level of Significance, i.e. area in the tail: (alpha) 0.05

Degrees of Freedom (DF); Sample size (total rolls, i.e. distinct possible categories, we used 100 rows to record), minus one (N – 1): 100 – 1 = (DF): 99.

For finding the Critical value (tail of graph and rejection region area). We'll just use 99 for DF, as I don't have specialised statistics software and I'm not calculating it longhand (DF 99 is solid enough) for this dice.

Looking on an official Chi-square table using (DF 99) and (alpha 0.05) Critical value, i.e. P-value (probability): 123.225 rounded 123.

Therefore if our dice exceeds the (above) P-value [123.225] "d100 Dice" is likely loaded.

https://i.imgur.com/IT4eZE7.png

Chi square: X2 = sum [(Observed - Expected)2/Expected]

For example using data: Roll 1; Observed value 5, minus Expected value: 5 (squared) divisible by Expected 5, equals: 0... Then rinse and repeat, i.e. total all the X2 for 500 rolls.

Our Test Statistic equals: 101.2

We can easily see: 101.2 is far less than P-value: [123.225].

Therefore we fail to reject: Null Hypothesis (H0): (we accept) Dice is fair (unbiased)


Conclusion

The "dice" do not appear to be loaded. In fact the statistical model shows there is an significant chance of ~ 95% percent the dice aren't loaded. (That's what the alpha 0.05 value we tested against means basically).

Discussion

Things do not have to even out, but sometimes seem to, as more observations are added (law of large numbers).

Law of Large Numbers: As the sample size increases the average of the actual outcomes will more closely approximate the mathematical probability.




A sampling error is a statistical error that occurs when an analyst does not select a sample that represents the entire population of data. As a result, the results found in the sample do not represent the results that would be obtained from the entire population. The population was 500 rolls, 100 rolls would not really be sufficient.

For the: group ranges of 20, the higher occurrences' were in the outer ranges: 1–20 and 80–100, the inner 21-40, 41–60 and 61–80 were nearly of equal size, which is to be expected, count (123, 88, 89, 88 and 112). Five equal ranges is a sensible number.

However, if you were to split the count ranges: 1–25, 26–50, 51–74 and 75–100 (145, 108, 117 and 130).

The 50/50 Split was distributed: 253/247.

The faces: 16 (frequency 13) and 20 (frequency 11) could possibly be considered outliers; if you were to just do a very basic box-and-whisker plot diagram on the quartiles.