Do RNGs produce random outcomes?

[happygobrokey]

one day the game will deal you a joker and after that you will be able to sleep easy

 

[ThodorisK]

KasinoKing, you were right on your confusion:

There is no such thing as an event being "due." An event is not more likely just because it has not happened for a long time. For example, many people mistakenly believe that if one color in roulette has won several times in a row then the other color is overdue and they should bet on it. While the ratio of reds to blacks will always approach 50/50 in the long term, it can not be concluded that this will happen in the short term.

This is only regarding theoretically random-independent events (which never happen). But regarding these theoretical conceptions:

The larger the number of the FUTURE draws, the higher the probability for the ratio OF THESE FUTURE draws to be close to 50/50. This conclusion refers ONLY to the future draws. It does NOT approach the 50/50 ratio taking in account the past draws. Because if it did take them in account, then the probability of the distribution of the future draws would be dependant on the past draws, and this definitely is false, as it contradicts the presupposition that these draws are independent.

But as I see, Mr Shackleford implied this mistake, since he said:

While the ratio of reds to blacks will always approach 50/50 in the long term, it can not be concluded that this will happen in the short term

Here he implies that:
the ratio does not approach the 50/50 ratio in the future short term taking in account the past draws, BUT it approaches it in the long term, TAKING IN ACCOUNT the past draws!

This statement is false, either if by long term he meant the future long term, or if he meant the long term including past events.

Most probably he made that mistake because of the rush of oral speach. But one can be confused in these things too.

I have an idea of why one might get confused, as I did in the past:
As mathematicians say, the ratio will definitely EQUAL 50/50 after infinite draws-spins. That might confuse one to say that this approach of the 50/50 ratio includes past draws. No, the approach does not refer to the past draws. But if we suppose that an infinite number of future draws happen, then it can be disputed that this infinity is identical with the infinity which includes the past draws which happened before these future draws. But if you mess with infinity, you will end up in contradictions.
An infinite number is impossible to happen because this would mean that it is the largest number of all other numbers, and as we know, there is always a larger number for any number.

Anyway, all the above comments were under the presupposition of independent draws. But independent events never happen.

 

[mikepipe]

Is this supposed to be an english lesson for people like me?

It is obviously not a lesson in thinking, not in probability, not in mathematics, not in writihng technique.
Hmmmmmm.

 

[ThodorisK]

just a clarification on "gambler's fallacy".

 

[happygobrokey]

you write a lot of words, yet say nothing.

flipping a coin is fifty-fifty.

when you flip a coin, it is fifty percent to land on its head.

when you flip a hundred forty-seven point one two three billion kajillion fafillion smolillion coins all at once, the ratio of heads is going to be ridiculously close to fifty-fifty.

that is just how they are made.

regarding infinity, i thought that was the universe? as the universe expands, so does the value of infinity. everything began as one, and is expanding/diversifying into innumerable things, thus infinity.

 

[GaryWatson]

Im glad I dont know stats

 

[happygobrokey]

there are three kinds of lie: lies, damned lies, and statistics.

 

[Zoozie]

regarding infinity, i thought that was the universe? as the universe expands, so does the value of infinity. everything began as one, and is expanding/diversifying into innumerable things, thus infinity.

That is just some mumbo jumbo using word-play You can even 'prove' god exists this way - Or that you mum is a rock - (Eramus Montanus)
You know better than this

Infinities is one of my favorite mathematical subjects, which is why I have to contribute with some important knowledge, that I find philosophically interesting, but has no real practical usage! (yet)

It is a common misunderstanding that the universe is infinite (besides assuming euclidean geometry which also is wrong) but this is physics and not really important in this discussion.
More important is that there are different types/order of infinities( and an total ordering of these exists).


Example of infinities of growing order:

1.Finite (not infinity): 5 etc. which should be seen as the number of x is this set (x x x x x)
2.Countable infinite: number of elements in these sets (1,2,3,4,5,......) and also (2,4,6,8,.....) (all even numbers). These two sets have the same degree of infinity! (1-1 mapping exists)
3.Uncountable infinite: Any line segment (interval with different end points.) Ie. [1,2] or ]-oo,+oo[ which also also thy have same degree of infinity.
It is shocking that the number of points in a line interval is the same as number of point is a 2 or even 3-dimensional equivalent. (areas or volume)
So a 3-D euclidean universe with no boundries has the same number of points as the interval [1,2]. This was shocking news the math world
when this was proved. ('I see it, but I can not believe it'). Now the universe does not seem to be so big anyway...
4. Number of different finite length curves you can draw on paper without lifting the pen.
5. Number of different sets (also infinite set) you can make from the elements of 4.
6. Number of different sets (also infinite set) you can make from the elements of 5.
...
...
...


(actually you can start from 1. and also make 2,3 by this recursive definition)

So there are infinity many kind of infinities. I my example this I only make up an countable infinite (ie 2. in my list) number of infinities, however there are many more. There are more different degrees of infinities than the size of any of the infinites that can be constructed recursive as I did. (some mathematicians claim this is due to using set-theory on infinite sets which is wrong - so this conclusion is controversial). Do not think too much about this or you might go insane as the mathematician who discovered this did -

You do not have permission to view link Log in or register now.

 

[GrandMaster]

Sheeesh this is one of the most boring threads I ever tried to read!

I echo Spear:-


So are you saying casino software is rigged, RNG's are not 100% truly random (which no one can detect & doesn't matter a toss to the person playing the game anyway), it's impossible to beat the house edge, or maybe something which we have not already heard 1000+ times before...?
(That question is to Thodorisk, not Spear!)

You do not have permission to view link Log in or register now.

Please understand this: There is no such thing as an event being "due." An event is not more likely just because it has not happened for a long time. For example, many people mistakenly believe that if one color in roulette has won several times in a row then the other color is overdue and they should bet on it. While the ratio of reds to blacks will always approach 50/50 in the long term, it can not be concluded that this will happen in the short term.

In my mind those two bold bits totally contradict each other.
It's OK - please no-one try to convince me otherwise!
I am quite happy to carry on winning in my ignorance and using my gamblers fallacy...

I will try to explain it anyway. What it means that the long term normal behaviour will dominate any kind of short term deviation. Let's stick with red/black on the roulette wheel, ignoring 0's. Let's assume that you get 100 reds on the first 100 recorded spins. The expected number of reds in the next 999900 spins is 499950, and the probability that it will be within 2000 of this number is about 0.999937. So the total number of reds in the first 1000000 spins will be between 498050 and 502050, and the proportion of reds will be between 0.49805 and .50205 with probability 0.999937. There is nothing about black being due or occurring more frequently after the first 100 spins.

 

[happygobrokey]

so the universe is NOT infinite, but the number of points on the line segment underneath these words i just typed, IS infinite? i'd say there are infinitely many more points in all of the universe than just on that line, if as you say, there are infinitely many infinities?

mind-bottling, isn't it?

did you just say "mind-bottling"?

yeah, mind-bottling. you know, when things are so crazy it gets your thoughts all trapped, like in a bottle.

(identify the movie for a free thank)

 

[Zoozie]

I will try to explain anyway. What it means that the long term normal behaviour will dominate any kind of short term deviation. Let's stick with red/black on the roulette wheel, ignoring 0's. Let's assume that you get 100 reds on the first 100 recorded spins. The expected number of reds in the next 999900 spins is 499950, and the probability that it will be within 2000 of this number is about 0.999937. So the total number of reds in the first 1000000 spins will be between 498050 and 502050, and the proportion of reds will be between 0.49805 and .50205 with probability 0.999937. There is nothing about black being due or occurring more frequently after the first 100 spins.

Can it be explained this way GM?

You start with 100 red spins and THEN do x-spins. 0.5x of these new spins will be red and 0.5x black. So for these new spins #red/#totalspins= (0.5x)/x=0.5

However the total number of reds including the first is now:
(100+(0.5x))/(0.5x) -> 0.5 for x-> oo

So short term deviation does not matter. It does not have to be 'evened out' in anyway - the math will still hold.

 

[ThodorisK]

It was my mistake to talk about infinity. We dont have to mess with infinity to analyse "gambler's fallacy". Also, the "gambler's fallacy" topic is already known, there is no point to discuss it further.

 

[ThodorisK]

I am focusing on proving theoretically if there is or if there is not a degree of non-randomness of the outcomes of a RNG, which can be taken advantage of to create an edge for the player, regarding only the below considerations
(Supposing there is no form of cheating of course, e.g. the possibility of REASON No3 forces me to examine only the case in which the future sequences are NOT filtered by stat tests before they come out. So in this NOT case):

I said in a previous post:
In a 1,000,000 digits sample, we expect to see a long run of the same digits. In a smaller sample, we do not expect to see such a long run. In a larger sample, we expect a larger run of the same digits.
So, if the stat tests were testing sequences a RNG produced in the past which are made of only 10 digits, then these tests would reject as non-random, sequences as 1-1-1-1-1-1-1-1-1-1 and if they also rejected the 1-0-1-0-1-0-1-0-1-0, and the 1-1-0-0-1-1-0-0-1-1, then you understand that there arent many sequences left. A great proportion of the possible 10-digit sequences, would be rejected. Actually, the more the different kinds of tests used to test each such sample of 10 digits, the less number of possible 10 digit sequences would be approved by them.
And therefore, if we know the passing requirements of these tests, we have a knowledge of what kind of sequences are produced by a RNG which has passed these tests! The future results of a RNG which passed these tests, would be predictable to a great degree, so a huge edge for the player could be derived!
Actually, the more 10-digit sequences are tested and succeed these tests, the more we can be certain that the RNG produces ONLY sequences that succeed in these tests!!!
Now, as I saw at:

You do not have permission to view link Log in or register now.

The size of a sequence a RNG produced in the past, which is selected for being tested under stat tests, is 1,000,000,000 digits or more.
But no matter how large the samples are, perhaps it is still possible to derive a degree of predictability because of the reasons I mentioned above.

I do doubt my above conclusions, but I cannot prove them wrong so far.
I am trying to see whether the below Aka23s arguments prove them wrong:

RNG tests don't simply report pass/fail. They report the probability of the test results occurring in a random distribution. The confidence level relates to how far off from expected values this probability can be for the test to pass. We arbitrarily designate a certain probability as passing. There is rarely a 100% probability of such a test passing.
The probability of a test passing relates to the specified confidence level, and can be calculated. Some RNG algorithms have a higher failure rate than others. The page at

You do not have permission to view link Log in or register now.

lists the failure rate for some RNGs on the NIST test I mentioned earlier. The author of the page is saying his RNG is good because it has a failure rate of 0.175%
...A RNG is tested with a long sequence of outputs. The result of the RNG test does not use past outcomes of the RNG, only the sequence under test. The outcomes of the RNG test is compared to the the expected outcome in a true random distribution. The confidence interval relates to the distance between the results and the expected results in a true random distribution. These expected results are determined using probability equations."

 

[aka23]

Why do you think RNG tests are only using 10-bit sequences? If you are testing RNGs with 10-bit sequences, all decent P-RNG algorithms will produce all the sequences you listed, including "1 1 1 1 1 1 1 1 1 1." Any sequence should occur during ~1 / 2^10 of the 10-bit sequence trials. If this sequence did not occur over millions of trials, then the RNG would likely be rejected as flawed. It also would be classified as flawed in the all-1 sequence occurred significantly more often than expected in a random distribution. However, a RNG certainly would not be rejected simply because it contained such a sequence. Or are you still suggesting that RNGs have some kind of result filter and don't put out data, if there is a cluster of 1's or 0's?

A more realistic test would be looking at a 1 million bit sequence. One of the NIST tests is to look for the longest run of 1's in the test data. If the longest 1 sequence was 1000 bits, then that particular test would fail. While that trial failed, the P-RNG might pass overall. For example, the most common P-RNG used in casino software seems to be Mersenne Twister, which repeats itself every 4.315425 10^6001 outputs. 10^6001 outputs is long enough to expect to see a sequence of 1000 1's. So a rare failure is normal. However, if the failure rate becomes too high on this test, then it is rejected as a bad algorithm. The page you linked to lists the failure rate on the NIST tests for various P-RNGs. Note that the tests that are failing are generally not the longest 1 sequence test or the runs test.

 

[ThodorisK]

I used the 10-digit example to grasp the matter better.

I am talking about the case there is no filter.

 

[ThodorisK]

Does anybody know what kind of RNG is used by Cryptologic and NetEnt?
Do they use algorithms only (as I think Chartwell does),
or a combination of both algorithms and natural phenomena (atmospheric noise, rarioactive decay etc), as Boss media does?

 

[hobobiker]

Does anybody know what kind of RNG is used by Cryptologic and NetEnt?
Do they use algorithms only (as I think Chartwell does),
or a combination of both algorithms and natural phenomena (atmospheric noise, rarioactive decay etc), as Boss media does?

No i dont know ...........but funny you should mention this because I know of one casino group (but im not allowed to say who it is on this forum im afraid) that use a EDDIE 3000,000 server which is an advanced version of the HAL 9000 invented by Doctor Sivasubramanian Chandrasegarampillai (or Dr. Chandra ) in the late 70's

And every intelligent person now knows that the name H-A-L is derived from Heuristic ALgorithmic

dunno why they call it EDDIE tho ??

radioactive decay eh ....... space ..... the final frontier .......

 

[happygobrokey]

hobo how dare you wear that maple leaf as an avatar? it had better be to show your regrets over your outburst earlier and not to be further taking the piss. elsewise i shall secede from the commonwealth and make a point to deface every coin i ever own (except from pre-1993 when she was still a bit sexy)

 

[hobobiker]

hobo how dare you wear that maple leaf as an avatar? it had better be to show your regrets over your outburst earlier and not to be further taking the piss. elsewise i shall secede from the commonwealth and make a point to deface every coin i ever own (except from pre-1993 when she was still a bit sexy)

thought it wouldnt take you long to notice that one.... not takin the mick really . and you were right in pointing out that using a personal dig in a forum is useless. as for showing regrets ......ive used up my quota for life im afraid.

but i must disagree about 'her' ever been sexy ... and i usually fancy posh types.

am removing the maple leaf now.

 

[happygobrokey]

fair enough. it was a nice picture though. the maple leaf i mean, not that old hag what's on money. just kidding majesty! lmao

 

[hobobiker]

yeah, mind-bottling. you know, when things are so crazy it gets your thoughts all trapped, like in a bottle.

forgot to say before:

Chazz... Blades of Glory .

 

[KasinoKing]

I will try to explain it anyway. What it means that the long term normal behavior will dominate any kind of short term deviation. Let's stick with red/black on the roulette wheel, ignoring 0's. Let's assume that you get 100 reds on the first 100 recorded spins. The expected number of reds in the next 999900 spins is 499950, and the probability that it will be within 2000 of this number is about 0.999937. So the total number of reds in the first 1000000 spins will be between 498050 and 502050, and the proportion of reds will be between 0.49805 and .50205 with probability 0.999937. There is nothing about black being due or occurring more frequently after the first 100 spins.

Yes, I know that, obviously!
But I'm not talking about past events, only future events.
You know what, I think it's about time I explained to this board how my 'Gamblers Fallacy' works

KK's version of the Gamblers Fallacy​

Questions & My Answers (please correct me if I'm wrong!):-
Q. Is the probability of a defined event occurring different after it has happened to before it has happened?
A. No.
If I set out to toss a coin & get heads 4 times in a row, my chances are 1/2 x 1/2 x 1/2 x 1/2 = 1/16 (1 in 16)
If I actually toss 4 heads in a row, I have just witnessed a 1/2 x 1/2 x 1/2 x 1/2 = 1/16 (1 in 16) event!

Lets go higher!
Say I set out to toss heads 24 times in a row!
The chances of that are a staggering 1 in 16,777,216 = less likely than winning the UK Lottery!
Just suppose for a minute that I actually managed to toss 23 heads in a row.
Now the next toss, as GM pointed out, still has only a 50/50 chance of being heads.
I toss the coin & it's heads! 24 heads in a row!
Q. Have I just witnessed a 1 in 2 event, or a 1 in sixteen & three-quarter million event?
A. Both. The last toss was a 1 in 2 event but the last 24 tosses were a 1 in 16,777,216 event!

This is the 'Gamblers Fallacy' GM, The Wizard & others talk about.
If you had just watched me toss 23 heads in a row and you put your mortgage on the next toss being tails - you still only have a 50/50 chance of winning, not a 16.7m to 1 chance.

The only way you could profit from these probabilities on a 50/50 game is by progression betting (also called martingaling), where you are betting against a long sequence happening from the start of it, not half way through.
In this you double your bet each time you don't win. In fact, you really need to more than double your bet, otherwise you could end up risking 1000's just to break even.
This simply doesn't work because for example after just 8 losing bets you are risking 255 x your starting bet just to break even.
So for '50/50' games like Roulette or BlackJack, forget it - it does not work.
But it can work on slots!
Why? Because each bet is not 'all or nothing' - you don't need to double each time.
With every spin you have a chance of winning something back - even a jackpot.
So you can do a very slow progression over 100's of spins looking for that 1 big win which will take you into the black.
When you get it, or reach a satisfactory profit target, move on to a different slot.
Of course, this doesn't work every time - no 'system' does, but as long as your overall gains are bigger than your losses, you're quids in!

How my Fallacy worked in August (not a particularly good month due to some careless play )
+$300 - King Solomons - Slots only.
+$125 - Ruby Fortune - Slots only.
+$100 - 32Red - Slots only, except $215 wagered on Video Poker - gained $2!
_+$35 - iNetBet - Slots only.
Evens - Intercasino - slots & cards.
_-$10 - Casino Fantasy - Slots only (Only a $10 deposit + 500% bonus!)
_-$43 - Will Hills - mostly cards
_-$71 - Tote Sport - Slots only.
_-$75 - DaVinchi's Gold - 90% of play was on slots.
===========
+$361 < Total.

Not brilliant, but better than a lojo kick in the rusty tennis_balls!

KK

 

[mikepipe]

Not brilliant, but better than a lojo kick in the rusty tennis_balls!

KK

Nice. But why should Johnsteed give lojo a slosterish lojo kick in simmos rusty tennis balls? maybe because he didnt winbig?
Oh- where is Thelma, still on her swedish zoozie?

 

[lojo]

Excellent post KK. I've always had a hard time putting my fallacy into words. And the reason I niggle and shmiggle again and again over the simple mechanics of how a slot works is that I want to know when it doesn't work, if there is something about the program I don't understand. Phoo Phoo on whether the RNG's are divinely random when it comes to slots, it's in the program where deviations occur.

The very fact that probability has no bearing over the long run is precisely why probability is important session by session ~ volatility=constant opportunity.

And disciplined play = winners
{([do you have an almost 'secret from yourself' tidy little sum you dip into now and then? When I did I was bux-up.] And I'm not talking about chasing, just a cushion) because I don't factor bonuses} where's that fourth paren??? I wanted to add a disclaimer

Back to topic... damn, no interesting 'whos online now'.

Re-iterate: When it comes to slots, the relationship of the RNG to the Program is what matters. I don't care if the RNG goes 1000 times per second and restarts after a billion clicks or a year in service before it is reseeded, nor if it is constantly reseeded. I care about the program and the relationship of the program to the RNG. I can deal with a slot machine from there.

+++++++++++

'ello Joe
'lo

I see mike piped up
'lo mike

K?
K!

 

[KasinoKing]

I am sorry, but when I made my last post in this thread I was very tired, and I've just realised I missed out one of the most important points I was intending to make!
For clarity, I have repeated a chunk of my previous post (in blue below) and added the missing bit! (Black)

I meant to respond fully to this particular part of GM's post:

There is nothing about black being due or occurring more frequently after the first 100 spins.

Questions & My Answers (please correct me if I'm wrong!):-
Q. Is the probability of a defined event occurring different after it has happened to before it has happened?
A. No.
If I set out to toss a coin & get heads 4 times in a row, my chances are 1/2 x 1/2 x 1/2 x 1/2 = 1/16 (1 in 16)
If I actually toss 4 heads in a row, I have just witnessed a 1/2 x 1/2 x 1/2 x 1/2 = 1/16 (1 in 16) event!

Lets go higher!
Say I set out to toss heads 24 times in a row!
The chances of that are a staggering 1 in 16,777,216 = less likely than winning the UK Lottery!
Just suppose for a minute that I actually managed to toss 23 heads in a row.
Now the next toss, as GM pointed out, still has only a 50/50 chance of being heads.
I toss the coin & it's heads! 24 heads in a row!
Q. Have I just witnessed a 1 in 2 event, or a 1 in sixteen & three-quarter million event?
A. Both. The last toss was a 1 in 2 event but the last 24 tosses were a 1 in 16,777,216 event!

So what this means is that GM is correct; regardless of the fact that I just tossed 23 heads in a row, the chance of it going tails on the next toss is still only 50/50.
But in my mind I am also correct! Having already witnessed an extremely rare 1 in 8 million event (23 heads in a row), the chances I will go on to witness a 1 in 16 million event is twice as unlikely - I really have got a better chance of winning the lottery than seeing that!
Therefore I personally see Tails as being more 'due' than not.

KK