Spiral Knights - Estimate of the number of prize boxes sold and mediocre evidence SEGA is lying about prizebox odds; Please report keys/halos

Estimate of the number of prize boxes sold and mediocre evidence SEGA is lying about prizebox odds; Please report keys/halos

23 replies [Last post]

Wed, 10/15/2014 - 05:40

Fehzor

Original Thread: Did we find all of the 1% items from the last promotion?

All of these were under the same 1% umbrella... Which is significant because there are 21 varieties in total, making it likely that one or more just wouldn't be found or ever unboxed during the promotion...

So can we verify that all were found, now that it is done? If you've seen one on the market or equipped, please post and I'll add it to the OP later.

Things that have been spotted

Prism pylons: 5 of them recorded

The fact that I was able to locate 5 pylons, yet nowhere near that many of any one windup key/halo indicates that the odds are very very likely to be different from what was advertised to us. All other items here have only been seen once.

Divine wind up Key (auctions)
Regal wind up Key (auctions)
Heavy Wind-up Key (Gigaslash)
Cool Wind-up Key (Anon)

Toasty gear halo (Wiki)
Regal gear halo (Wiki)
Volcanic Gear Halo (Fangel)

Things that no one has spotted yet

Heavy gear halo
Cool Gear Halo
Divine Gear Halo
Dusky Gear Halo
Fancy Gear Halo
Military Gear Halo
Prismatic Gear Halo

Dusky Wind-up Key
Fancy Wind-up Key
Military Wind-up Key
Prismatic Wind-up Key
Toasty Wind-up Key
Volcanic Wind-up Key

____________________________________________

Here is a chart that I made of approximately how many boxes were opened based on how many unique 1% items were found. It DOES account for duplicates.

Number of unique accessories acquired : Expected number of boxes opened up to that point
1 : 100
2 : 205
3 : 315
4 : 432
5 : 555
6 : 686
7 : 826
8 : 976
9 : 1138
10 : 1313
11 : 1504
12 : 1714
13 : 1947
14 : 2210
15 : 2510
16 : 2860
17 : 3280
18 : 3805
19 : 4505

Note that this isn't going to be anywhere near exact- just vaguely ballpark, but more and more vaguely ballpark as we find more and more of these.

Method:
To find the first unique 1% item, it takes 100, as ANY 1% item counts, and 1 / 1 / 100 = 100.
To find the second unique 1%, it takes any of the other 20 1% items, as 1 / 1 / (20/21 * 100) = 105; 105 + 100 from the first = 205;
...and so on.

This discounts already found items from the pool, giving more weight to newer ones. It almost certainly costs us in accuracy, but accounting for duplicates makes finding out which ones exist much, much, harder. This also only accounts for known prize boxes opened- if I open a prize box, who do I tell or show? If the answer is no one that reports it here or on the wiki, we've missed that box.

In other words, this really really shouldn't be used as "PEOPLE OPENED AROUND X PRIZE BOXES AND PROVIDED OOO WITH ~X*5 DOLLARS" because of these huge accuracy issues.

EDIT

The above values are incorrect if SEGA lied about the prize box statistics. The following values predict that each box has a .66 chance of finding a halo or wind up key, and completely disregards the number of pylons found.

Number of unique accessories acquired : Expected number of boxes opened up to that point
1 : 150
2 : 307
3 : 474
4 : 651
5 : 838
6 : 1038
7 : 1252
8 : 1483
9 : 1733
10 : 2006
11 : 2306
12 : 2639
13 : 3014
14 : 3443
15 : 3943
16 : 4543
17 : 5293
18 : 6293

Wed, 10/15/2014 - 08:18

Vohtarak-Forum

im vohtarak, call me by that name

I saw njthug with the pylons
and a regal wind up key in the auctions

Wed, 10/15/2014 - 12:30

Fehzor

Updated the OP with lists. I'll keep this thread active for a while longer. If you see anyone trading the items, on the auctions... wherever, please report them. I wanna know if I was even vaguely right when I said that there was a good chance of a few of these never making it in game.

Wed, 10/15/2014 - 15:01

Fangel

Nah it wasn't heavy

I got the volcanic gear halo.
Try checking the wiki. Whatever items in there have pictures have been unboxed.

Wed, 10/15/2014 - 16:48

Gigaslash

I own a heavy wind up gear

Thu, 10/16/2014 - 15:36

Fehzor

Checked the wiki and fully updated everything. Keep submitting these, and be sure to ask around!!!

EDIT/UPDATE:

So far, we have 7/21 of the 1% items found.

Updated + improved the OP to reflect upon why this information is awesome.

NEW UPDATE! Improved the odds predicting that SEGA has lied to us all about the odds of finding a pylon compared to anything else.

Thu, 10/16/2014 - 15:49

Sweet-Hope

but what we can do "in case" SEGA or well OOO is lying to us with the odd chance on prizebox? people will still buying it, and nothing can be done about that (i mean about prizebox odd, we cant do something about it anyway)

will you do the same thing on every promo by any chance? because there is right now another promo that might worth to look into the odd chances.

Thu, 10/16/2014 - 17:14

Klipik-Forum

I don't think there's anything we can actually do about it other than be salty. As for getting the actual information, it would good for people deciding whether or not to buy boxes.

Thu, 10/16/2014 - 17:20

Lordyasir

You are right the odds are

You are right
the odds are higher than OOO has announced

Thu, 10/16/2014 - 17:59

Avihr

^ lawl

Thu, 10/16/2014 - 19:56

#10

Fehzor

What we can do is take the odds given with a grain of salt.

If SEGA tells us that there is a "15 percent" chance of finding A, B and C, we know that A+B+C = 15, but that A, B and C are NOT all equal to 5 as we would expect they are, provided that they aren't just flat out lying and simply making up a set of stats. Most of what they've told us is at least somewhat true, but this is a ways off.

@Klipik

It's only really possible with this one, because of the 20 items for an assumed .0066 factor. I have no better way to poll information than to ask if it exists at all, which is a pretty good way to do it because it takes from all "known accessories", and measures very close to exactly the boxes that are on the market.

@Lordyasir

The odds are not higher, they're just different. If I wanted say, the divine windup key, the odds are drastically lower than if I wanted a pylon. This is rather upsetting.

Thu, 10/16/2014 - 20:14

#11

Bopp

two problems

This thread will not collect statistically valid evidence, let alone "proof", that Three Rings is lying. It will collect anecdotal evidence with tremendous observation biases.

You assumed that all items under the "1% umbrella" have equal odds? Now you are trying to collect evidence that they don't have equal odds? Even if your evidence pans out, and they don't have equal odds, I still don't see the lie. Did Three Rings promise that they would have equal odds?

Fri, 10/17/2014 - 09:04

#12

Fehzor

It is largely implied that there are equal numbers of each, and not having equal percentages within brackets makes the odds highly ambiguous.

Fri, 10/17/2014 - 09:25

#13

Avihr

Hmm, I've fought worse

What kind of conclusion could you possibly make if you don't have the number of boxes that have been opened in order to know how many rare items per box there are?

Fri, 10/17/2014 - 09:57

#14

Turn-Me-On

Lol, dense OP is dense. Guess

Lol, dense OP is dense.

Guess you're down to the negative 3 sigma range on the IQ scale now fehzor.

The prize box odds are odds, not guarantees. You lack a fundamental understanding of statistics.

Just because something has a 1/6 chance of happening doesnt mean it will happen after 6 occurrences.

Try rolling a dice 6 times. You might notice that you never land on 1,2,etc.

Fri, 10/17/2014 - 10:23

#15

Lordyasir

If it was real luck,even 99%

If it was real luck,even 99% chance items may not see the light despite the tones of boxes opened by the players

Fri, 10/17/2014 - 11:42

#16

Trats-Romra

@Octaslash
It's not the same with computer.
Computers don't have the capacity to generate number that are truly random. There's code that generate numbers more random, but computer can't generate numbers totally random.

This link indicates that to generate random numbers, you need to depend on random nature events and I don't think that OOO would take their time to make unboxing more random. I think that, probably, they use a simple pseudorandom number generator, because you don't need to afford it to gamble(right word?), differently of guarantee of the security of credit card, that you needs to be well made.

I'll give a example. I generated this 100 random numbers from the site www.random.org with range of 1-100:
59 89 41 32 92 61 75 65 23 84 63 49 41 98 2 62 35 5 46 60 81 47 30 92 7 89 57 76 24 44 72 94 53 50 33 23 27 66 43 16 58 24 41 73 11 80 66 53 63 47 12 75 32 34 39 65 92 99 51 96 78 31 15 17 89 29 1 92 56 78 9 49 31 41 31 88 89 94 55 35 22 89 69 79 51 26 96 1 56 39 58 57 36 100 8 31 12 84 44 8
And sorted these numbers with Java:
1 1 2 5 7 8 8 9 11 12 12 15 16 17 22 23 23 24 24 26 27 29 30 31 31 31 31 32 32 33 34 35 35 36 39 39 41 41 41 41 43 44 44 46 47 47 49 49 50 51 51 53 53 55 56 56 57 57 58 58 59 60 61 62 63 63 65 65 66 66 69 72 73 75 75 76 78 78 79 80 81 84 84 88 89 89 89 89 89 92 92 92 92 94 94 96 96 98 99 100

See? There are some numbers that doesn't appear on middle of them, but have numbers on all ranges: 1-10, 11-20, 21-30, 31-40 until 91-100. So it's not totally equal percentages, but they are near when you generate numbers on computers. So, don't compare dices with random number generated by computer.

OOO can even have a good number generator, but don't compare it with dices. They are totally differents. Dices don't folllow programs like computers do. They are totally affected by ambient. The computer relies more on code and because of that, it can't generate totally random numbers.

@Fehzor
I think you should hear Aviri and Bopp a bit. How I showed to Octalash, you can get something next, but we have no idea of how they sort odds. But, I think you can try to get something next. I think you can, because you appear to be very interested on it. You can continue to try with method of exhaustion, but certainly it isn't the most trivial method. May you could search a bit about statistics and generation of random numbers on computers. It could be a step to discover the answer.
But, remember: Statistics begin searching for approximations, not exact numbers, because on most cases you can't determine it with total precision. On most cases, you have a result with margin of error

Fri, 10/17/2014 - 12:21

#17

Fehzor

In order from legitimate concerns to misunderstandings of statistics and people bashing me for no reason-

@Bopp continued cuz last post was from phone and I need to type now

OK. You're 100% correct about it. I didn't mean proof as in actual proof. I wrote that in the OP. But it does pan out to be around 99.9% certain (after the bias, of course), that there is an inaccuracy.

The amount measured is the amount of all people that I know, and that visit the forums, that know people that have seen the items. This number grows very quickly, until it presumably fills up the majority of the playerbase. There are always ones that will be hidden, and we can never measure those, so this only applies to the percentage of the items that are in the market OR equipped OR publicly available in any way at all. The actual results will be very shaky, but the more people we get to participate, reveal they have it, etcetera the more accurate the results will be.

@Aviri

I am after boxes sold as a magnitude (i.e. not 5vs 6, 100 vs 1000, or 1000 vs 2000), and to find this I am using the number of unique gear halo/key styles in existence.

I'm predicting in the latter half of this that the odds are 1/3 key, 1/3 pylon and 1/3 gear halo, because that makes the most sense. Aside from that, there is absolutely no evidence that any of the math would work out- but there is evidence that it IS 1/3, 1/3, 1/3 because that would be the most sensible thing to do w/out thinking about it.

@Trats-Romra

The only problem with being psuedo random is that you aren't truly random and your data might follow a pattern of some sort, or be exploitable. All of what I've said should hold up if SEGA has done a good enough job of ensuring that the numbers used for prize boxes are at least vaguely random, and I am going to just assume that they are because that's the best I can do. Aside from that, you're definitely right in that Bopp and Aviri raise legitimate points, so you should read their responses above.

@Lordasir

I'm not talking about luck or probability though, I'm talking about statistics. Not high level statistics either; I am no mathematician.

While it is possible for this to happen, how likely is it?

@Octaslash

Ok ok ok this is exactly like in those ridiculous high school drama shows, where there is the mean chick and she goes by and is like "GEEZE NERD, where did you get that purse? It doesn't match your outfit at all!" and then the friends are all like "oooh, burn!" except the actual intelligence factor is being reversed- "GEEZE FEHZOR, where did you learn your statistics? It doesn't reflect a fundamental understanding at all." But are you really any better than them, if you're still the one coming by to hate on me?

Like, I get it, you don't like me because I don't agree with you. That's fine, but that doesn't mean that you have to disagree with me, and if you are ONLY disagreeing with me, you don't have to gauge my eyes out with a stapler in the process. If you do that, then we'll all go blind and I'll have won regardless because I'll be the one eyed naked lady leading everyone around.

But that's really beside the point.

You used the example of a dice roll, and there are two things that you're missing-
The EXPECTED VALUE of the dice roll is that there will be 1 of any given value present after 6 rolls. This is by no means the actual result, but would be the "best guess", so to speak.

Enter the Central Limit Theorem-

"the arithmetic mean of a sufficiently large number of iterates of independent random variables, each with a well-defined expected value and well-defined variance, will be approximately normally distributed"

Since we are dealing with a large number of box openings, we can predict that the variance will decrease quite a bit. For instance, if you rolled a dice 600 times, it would be very likely to come out at ~100 of any given number, given that the dice was not the ten sided cube from Mario Party. In effect, more trials reduces the offset and brings us closer to the expected value, which I calculated.

If you don't believe me, you can try rolling a dice 600 times and then taking data on it. I'll even do it for you because it only takes a second. I'll measure the number of 6s that pop up, and the total value of all the numbers and we'll see how far they deviate from the expected.

Expected Total: 2100 = (1+2+3+4+5+6) * (600 / 6)
Expected Sixes: 100 = (600/6)

Results of a few runs:

Total: 2109
Sixes: 110

Total: 2169
Sixes: 117

Total: 2168
Sixes: 113

Total: 2194
Sixes: 110

Total: 2037
Sixes: 102

Total: 2169
Sixes: 111

So I always got quite a bit more sixes than I had bargained for, but at the same time I always got around 100 6s. The same thing should be happening here.

Fri, 10/17/2014 - 12:50

#18

Midnight-Dj

@Fehzor

Okay, this test is nice and all (a refreshment to my math class thx). But what is the point of it? Even if the test proof at the end of the day, that SEGA rigged the prize boxes, will they admit it? It is kinda like presenting evidence to a christian that their god doesn't exist, they would simply say this, 'lol, our god exist outside nature, have you been outside nature? No? Then everything you said is crap!!!'. SEGA would do something similar once you had the proof I can bet my last crown on that.

Fri, 10/17/2014 - 14:19

#19

Trats-Romra

@Fehzor

All of what I've said should hold up if SEGA has done a good enough job of ensuring that the numbers used for prize boxes are at least vaguely random, and I am going to just assume that they are because that's the best I can do.
Hum... I comprehend it. Ok, no problem though.

The only problem with being psuedo random is that you aren't truly random
The thing is exactly that. Numbers generated by computer always be pseudo at all. People can even use "seed method" to avoid number repeating and get numbers that are more random. But it always be pseudo, except when are used devices to detect random natural events to generate random number based on them. And how they envolve special devices, it's problably very expensive and not worth expeding money with it, because they can pretty much create pseudorandom generator numbers that easily attend them and much more cheaper. I don't think SEGA would invest on these devices.

Fri, 10/17/2014 - 21:18

#20

Skepticraven

↓

@Fehzor
"It is largely implied that there are equal numbers of each, and not having equal percentages within brackets makes the odds highly ambiguous."

Implied. Perhaps if you want clarification, you could send a support request rather than go on with your conspiracy theory?

If you want to learn a bit more about random number generation, I suggest taking a read on page 298 of the pdf of numerical recipes in c. It talks about biases introduced using the mod function, pros and cons of various algorithms, and even entropy injection for increased randomness. It is interesting how you use java as the testing platform, when the server generates the random numbers for the game. We are not certain the server runs java. Their open source code does suggest this, however.

@Trats-Romra

Random.org uses atmospheric data. Also, if the PRNG is designed correctly... it is indistinguishable from random. In fact, rule 30 cell automata is computationally inexpensive and practically indistinguishably random if you do not know the seed [If you know the seed, it is deterministic].

Fri, 10/17/2014 - 21:35

#21

Fehzor

@Midnight

SEGA doesn't have to admit anything.

@Trats-Romra

Explain why it matters that it could be psuedo-random? And I say could because there are free software packages to retrieve true random numbers.

I've already said that I'm going to assume that SEGA's RNG that they're using for prize boxes isn't busted in some way, and provides perfectly good random numbers, and it may not but again, we cannot test that.

@Skepticraven

That's not a bad idea at all. I sent them a support ticket, and will report the results whenever they respond.

When I wrote the code in Java, it was only to show that given a large number of trials, we could expect a normal distribution centered around the expected value, since there was somehow a disagreement.

Sat, 10/18/2014 - 04:13

#22

Vohtarak-Forum

I'm vohtarak, call me by that name

if only that sk is bankrupt thread changed its name as many times as this

Sat, 10/18/2014 - 04:18

#23

Ironclaw-Mender

I'm sorry OP, but this thread looks like Time Cube.

Search

Estimate of the number of prize boxes sold and mediocre evidence SEGA is lying about prizebox odds; Please report keys/halos