Dupes

shteev

Now, normally I'm not one to talk about duplicates. Randomness is clumpy, I know that. You're unlikely to get an even distribution of random cards. Perceiving a skew in the RNG is generally just confirmation bias, or the Birthday Problem rearing it's merry head.

But.

I've just got my 10th Consulate Crackdown as my RatC reward, and I very much feel like something fishy is going on. Out of the other AER rares, I have 7 Release the Gremlins, and no more than 3 of anything else.

I'm an impatient person, and I always collect my reward right when the event ends... and it always seems to be a Consulate Crackdown.

I shall be revisiting this thread at the end of NoP and RatC, and let's see if I can get another one or two, eh?

Houdin

And I just got my 7th walking Batista.
Hmmm. Doesn't seem like there is anything going on at all. Why would anyone think the laws of probability would apply to mtgpq

shteev

In the meantime, a though occured to me. Since the drop rates for this game aren't published at all, are we just assuming that the drops are random?

Houdin

Ahh @shteev I think you may just have been the first to "speak out loud" what we all deep down no to be true but have tried to convince ourselves isn't. Let's be honest. If drop rates are not random. In fact if drop rates have been programmed to consistently give a player duplicates of cards they have already received almost 100% of the time, then the reality is that D3 has been purposely lying to us for quite some time in an effort to take our money. I don't think many of us really want to have to contemplate the ramifications of that.

Ohboy

I have to assume you at least know that random numbers fall under a normal distribution shteev. 

Your description perfectly describes a bell curve.

https://en.m.wikipedia.org/wiki/Normal_distribution

This is what you EXPECT to find in randomness. A uniform drop rate is the most basic tell of a non random process. 

shteev

Ohboy said:

I have to assume you at least know that random numbers fall under a normal distribution shteev. 

Your description perfectly describes a bell curve.

https://en.m.wikipedia.org/wiki/Normal_distribution

This is what you EXPECT to find in randomness. A uniform drop rate is the most basic tell of a non random process. 

You are right, of course. So this is what I'm going to do... I'm going to come in here just before I open my prizes for NoP and RatC, and I'm going to make a big show of myself declaring 'Ooh, I wonder if THIS card will be a Consulate Crackdown!'.

And if neither of them are, then I will sneak off with my tail between my legs, and we'll hear no more about it.

madwren

What's always odd to me is the streaks. I had a run wehre I got 4 Glint-Sleeve Siphoner in a row from events, followed by 3 Solemn Recruits, followed by 3 Greenbelt Rampagers and a pair of Quicksmith Spies. Today I got a Crackdown, I wonder if I'll get 2 more?

shteev

Houdin said:

Ahh @shteev I think you may just have been the first to "speak out loud" what we all deep down no to be true but have tried to convince ourselves isn't. Let's be honest. If drop rates are not random. In fact if drop rates have been programmed to consistently give a player duplicates of cards they have already received almost 100% of the time, then the reality is that D3 has been purposely lying to us for quite some time in an effort to take our money. I don't think many of us really want to have to contemplate the ramifications of that.

I think it's certainly worth contemplating that the drops aren't random... and hey, @Brigby, now might be a good time to confirm that the drops are, at the very least, 'random' in whatever legal sense you deem necessary to cover yourselves.

But, actually, I don't really suspect that the drops are calculated specifically to annoy us. What I think is more likely is that Hibernum keep re-seeding the RNG with the system clock in an ill fated attempt to make it 'more random'. I worked at a place where development did that once. You would not BELIEVE how difficult it was to persuade them that their random number generator was demonstrably not random.

Ohboy

I think everyone learns from that mistake the first time they try to be clever. 

shteev

Ok wait. Should random numbers fall on a normal distribution? My low sample size of dupe rares is unlikely to have a flat distribution, but surely, as I collect more and more, lets say, a billion more or so, the distribution should become flatter, shouldn't it? And not retain the same shape with some cards being far more common than others?

Normal distribution is about averages, isn't it, and not absolute values of random numbers?

I confess it's been a few years since I studied statistics at school.

Steeme

Why does everyone assume that their algorithm is purely random?

They are a business, with a business model, and there is typically business logic behind everything in their application.

shteev

Steeme said:

Why does everyone assume that their algorithm is purely random?

They are a business, with a business model, and there is typically business logic behind everything in their application.

Its a disturbing thought, isn't it, that 'common' or 'rare' could just be names for types of cards which have nothing to do with rarity at all.

I'm reminded of Subway and their 'Footlong sub' which turned out not to be a foot long, and Apple's famously nonsense 'Twice as fast, half the price' ad.

murtagon

My base assumption has been you have equal odds of receiving cards of a specific rarity.  I've always thought the method that determines which cards you receive would be a two phase process.  First rarity then the card, versus each card having a unique chance of getting the card.  Something along the lines of 60% common, 38% uncommon, 1.8% rare,.2% mythic.  Then the cards within the rarity pool were even.   This would generate a bell curve but within the rarity the curve would be flat.  But what people seem to be experiencing is a bell curve within the rarity, and that curve seems to be based on the players collection.

Ohboy

I think we can at least agree that none of us have drawn anywhere close to a billion cards yet. I believe it completely flattens out at infinity. I'm not sure how many rares I've opened, but I think I can safely assume it's less than a million. I did a million random rare open simulation to see how well it flattens out.

The point is that we will all have rares that are more "common" than others. That's expected. Even after a million run simulation, it looks pretty flat with about a 250 card variance between the max and min number of copies of a certain rare.

If everyone's most common rare were the same, it would lend credibility to the theory of biased tiers within rarities, but as far as I can tell, that's just not happening. I have I think 12 inexorable blobs from SOI. How many do you have? It's just normal variance.

As for streaks...it's interesting to note that another common way to detect non random numbers is the lack of streaks, because humans tend to think streaks reveal non randomness and seek to remove them.

shteev

You believe that Normal Distributions flatten out at infinity? That's not true, is it?

Ohboy

shteev said:

You believe that Normal Distributions flatten out at infinity? That's not true, is it?

Of course that's not true. Normal distributions by definition do not flatten.

Should random numbers fall on a normal distribution?

I was referring to this part when you mentioned large numbers

murtagon

Ohboy said:

I think we can at least agree that none of us have drawn anywhere close to a billion cards yet. I believe it completely flattens out at infinity. I'm not sure how many rares I've opened, but I think I can safely assume it's less than a million. I did a million random rare open simulation to see how well it flattens out.

The point is that we will all have rares that are more "common" than others. That's expected. Even after a million run simulation, it looks pretty flat with about a 250 card variance between the max and min number of copies of a certain rare.

If everyone's most common rare were the same, it would lend credibility to the theory of biased tiers within rarities, but as far as I can tell, that's just not happening. I have I think 12 inexorable blobs from SOI. How many do you have? It's just normal variance.


As for streaks...it's interesting to note that another common way to detect non random numbers is the lack of streaks, because humans tend to think streaks reveal non randomness and seek to remove them.

I'm not sure why you think you need 1 billion cards to see odd variance within a rarity pool.  The current set has 58 non exclusive rares.  I opened 70 premium packs.  Each had a guaranteed rare.  Unless the formula is set to ensure dups occur for certain cards I should not see such a large distribution on certain cards.  I do agree 70 cards would not be enough to guarentee one of each card but the odds of getting 8 of a single card are very remote unless there is an additional variable to increase those odds.   Even my Alt that only has 12 unique mythics has 3 Soul Swallowers and 4 Nahiri's Wrath, 1 of each of the other 10 mythics.  So while i agree human perception has a significant impact when trying to see patterns within a formula, the streaks and high dup rates people experience do show something is up and that a truly random distribution is not in place here.

murtagon

But for those that really care here are industry tests used for randomness

https://en.m.wikipedia.org/wiki/Diehard_tests

shteev

Ohboy said:

The point is that we will all have rares that are more "common" than others. That's expected. Even after a million run simulation, it looks pretty flat with about a 250 card variance between the max and min number of copies of a certain rare.

If everyone's most common rare were the same, it would lend credibility to the theory of biased tiers within rarities, but as far as I can tell, that's just not happening. I have I think 12 inexorable blobs from SOI. How many do you have? It's just normal variance.

As for streaks...it's interesting to note that another common way to detect non random numbers is the lack of streaks, because humans tend to think streaks reveal non randomness and seek to remove them.

These things all appear true to me.

Ohboy said:

I think we can at least agree that none of us have drawn anywhere close to a billion cards yet. I believe it completely flattens out at infinity. I'm not sure how many rares I've opened, but I think I can safely assume it's less than a million. I did a million random rare open simulation to see how well it flattens out.

But this... this confuses the hell out of me. Please elucidate. If a distribution of random numbers should fall on a curve of any description, then changing the sample size will not change the shape of the curve. A small sample size may deviate from the curve to a greater degree than a large sample size, with outliers such as my 10 Consulate Crackdowns, but the shape of the underlying curve itself won't change.

Is this a troll? Am I falling for a troll? Or... do you need to think about what you're saying just a little bit more and admit that you shouldn't, in fact, expect to find a bell curve in the distribution of rares in my collection, no matter what the size of it? Or am I failing to see something because it's nearly 5am here now???

Ohboy

Nah it's me. I'm doing this while pretending to work, and it got away from me. I agreed with your assumption too hastily.

murtagon

shteev said:

But this... this confuses the tinykitty out of me. Please elucidate. If a distribution of random numbers should fall on a curve of any description, then changing the sample size will not change the shape of the curve. A small sample size may deviate from the curve to a greater degree than a large sample size, with outliers such as my 10 Consulate Crackdowns, but the shape of the underlying curve itself won't change.

That is my point as well Shteev.  The number of cards available and the number of attempts at those cards just do not explain the consistent streaks people are experiencing.  There do appear to be other factors impacting the odds to increase dup chances.