Poll

Is This RNG Broken?

Yes, this output is completely unnatural.
Yes, obviously this casino has altered the results and is cheating.
No, roulette produces these kinds of results as part of normal output.

Author Topic: IS ONLINE RNG BROKEN!?  (Read 1384 times)

0 Members and 2 Guests are viewing this topic.

MickyP

Re: IS ONLINE RNG BROKEN!?
« Reply #15 on: May 15, 2019, 02:25:58 PM »
LOL, It is for this reason that you never test a system in fun mode.
 
The following users thanked this post: GIAJJENNO, Astutillo

Re: IS ONLINE RNG BROKEN!?
« Reply #16 on: May 15, 2019, 03:26:30 PM »
LOL, It is for this reason that you never test a system in fun mode.

MickyP yes this is so true.
Fun mode is a non sense.
 

Third

Re: IS ONLINE RNG BROKEN!?
« Reply #17 on: May 15, 2019, 03:52:05 PM »
the best selection to play RNG is play few numbers.

If the RNG is rigged, the best selection to play is the following:

NO NUMBERS

If a RNG is rigged you will LOSE period.
 

Re: IS ONLINE RNG BROKEN!?
« Reply #18 on: May 15, 2019, 04:01:51 PM »
the best selection to play RNG is play few numbers.

If the RNG is rigged, the best selection to play is the following:

NO NUMBERS

If a RNG is rigged you will LOSE period.

It' not so straightforward.
I made profits betting few numbers, while it's quite impossible to avoid heavy drawdowns betting large sectors or many numbers.
I mean the software behaves differently.
In the second case it clearly turns on a "defensive mode". Same if you play on sunday.

 

Third

Re: IS ONLINE RNG BROKEN!?
« Reply #19 on: May 15, 2019, 04:06:48 PM »
Well, I agree that its not straight-forward but in the experiences I have had with what I suspect was a rigged roulette, I was receiving "reverse odds"; so like, if I would be playing 2 Dozens, I would get the odds of 1 Dozen.  What would happen is my bankroll would go down by 33%, up by 20%, down by 33%, up by 10% and bust.  So that's what I mean.
 
The following users thanked this post: Astutillo

Joe

Re: IS ONLINE RNG BROKEN!?
« Reply #20 on: May 15, 2019, 06:44:27 PM »
I have a serious question for you Joe.  How did you calculate the chances of 5 numbers appearing in 10 spins?
First, I would like to say how much I admire what you're doing Third. It's no mean achievement to make a living from roulette, especially playing systems online, and with an RNG too! Kudos to you. I hope you continue to find success because I'm drawing inspiration from it.

The calculation is a little tricky, but I will try to explain it; let me know if anything's unclear.

I'm assuming you're aware of the binomial distribution formula? it calculates the chances of the number of "successes" in a number of trials. So in this case I took the number of trials to be 8, not 10, because we want to calculate the probability of the cluster occurring in the minimum number of trials, which means taking the end of the sequence at the point where the cluster is confirmed to end (which happens on the 8th spin in your screenshot).

There are 5 of #28 in the sequence, but we don't calculate the chance of exactly 5 #28's in 8 spins, rather we want the probability of 5 or more #28's. Why? this is where it can seem paradoxical; the chance of exactly 5 hits is actually lower than the chance of 5 or more, because in the latter case we are adding up the probabilities of exactly 5 hits, 6 hits, 7 hits, and 8 hits. This is just a consequence of the law in probability which says you that you add probabilities which are mutually exclusive.

Rather than calculate those separate chances and add them up, it's easier to use what's called a Cumulative Distribution Function or CDF. The CDF is actually defined as the chance of less than or equal to some outcome, which in this case would be the probability that you would get 0..5 hits of #28 in 8 spins. But we don't want that, we want the probability of at least 5, and to get this we need to subtract the CDF from 1.

Why? because it has to be true (ie 100% chance) that either #28 hit 0 to 4 times inclusive, or it hit 5 times or more. ie, in terms of probability :

chance that it hit 0 to 4 times + chance that it hit at least 5 times = 1

But the chance that it hit 0 to 4 times is the CDF (using 4 as the parameter), so

CDF + chance that it hit at least 5 times = 1, and rearranging we have

Chance that it hit at least  5 times = 1 - CDF

Calculating the CDF is tedious but there are calculators online to do it. Unfortunately none of the ones I've seen give you the degree of precision needed for this calculation (because the probability is very small). I use a statistical software package called Gretl (free) which can do this very easily and a lot more besides.

Here is the output from computing the above calculation :

Code: [Select] [nofollow]
p = 1 - cdf(b, 0.027, 8, 4)
Generated scalar p = 7.50545e-07

But remember this is the probability of a particular number (in our case #28) hitting at least 5 times in 8 spins. We need the probability that any number will do it. That means we have to multiply the probability by 37 (because there are 37 ways it can happen, one for each number).

So the final probability is 37 * 7.50545e-07 = 0.000027770, which is 1 in 36,010.

Ok so that was quite tough for those who don't know basic probability. Let me know if you want anything clarified.
 
The following users thanked this post: kav, Third

Joe

Re: IS ONLINE RNG BROKEN!?
« Reply #21 on: May 15, 2019, 06:58:12 PM »
It's true that strange steaks become much more suspicious because (when) they are most often against us instead of favorable to our bets.
Kav that's a good point. You would never hear someone say : "this casino must be fixed because I'm winning too much!". That might seem silly because obviously if the casino is fixing outcomes it's going to be in their favour not ours, but it does draw attention to our confirmation bias.
 
The following users thanked this post: kav, Third

Third

Re: IS ONLINE RNG BROKEN!?
« Reply #22 on: May 15, 2019, 07:56:23 PM »
LOL.  Ya it would make no sense the casino sitting around saying, "Darn it!  We were fixing the results but the guy didn't bet the way we wanted!".  ROFL.

Thank you Joe.  Excellent work btw.  Being of an ignorant sort here, I am strongly inclined to argue that the chances of a PARTICULAR number doing something, is the same as ONE (out of 37 numbers) doing something and we cannot calculate the chances of a particular number doing something without factoring in the 37 number set, right?  So, I don't understand how you needed to make a separate calculation for the 37 numbers?
« Last Edit: May 15, 2019, 07:58:44 PM by Third »
 

Joe

Re: IS ONLINE RNG BROKEN!?
« Reply #23 on: May 16, 2019, 04:07:50 PM »
So, I don't understand how you needed to make a separate calculation for the 37 numbers?
It might help to show why you have to do it by describing a simpler calculation. If you want to calculate the chance of getting two hits in row on a single number you might reason like this :

The chance of getting a hit on one number is 1/37, so the chance of getting 2 consecutive hits on a number is 1/37 * 1/37 = 1/1369, because you multiply the probabilities to find the probability of a streak.  But there is an ambiguity here; does "a number" mean a particular number, such as #19, or any number? If the latter, the above calculation is not correct because it applies only to some particular number. If you think about it, it's much harder to pick a particular number specified in advance to repeat than it is to predict that some number will repeat.

In fact the chance that any number will repeat is 1/37 and not 1/37 * 1/37. This because you are saying that :

#0 will repeat or #1 will repeat or #2 will repeat .... all the way up to #36

And because these events are mutually exclusive you can add their probabilities (as I explained above). So each probability of a repeat is 1/37 * 1/37 and the calculation for all numbers is :

 1/37 * 1/37 +  1/37 * 1/37 +  1/37 * 1/37 + ... up to 37 times.

But this is just a long way of writing 37 * 1/37 * 1/37 = 1/37 (the 37 cancels the 1/37), so the probability of any number repeating is 1/37.
The same logic applies to more complex calculations using the binomial formula; you still have to multiply by 37 to get the probability of the event for any of the 37 numbers.

If you're still not convinced, you can write code to find the probability. In fact I would recommend doing this for any complex calculation because it serves as check that your maths is correct. And actually some probability calculations are so complex that there isn't any way of calculating them; you have to write a simulation.

I don't have time right now but I'll write some code to verify it later this evening.
« Last Edit: May 16, 2019, 04:42:43 PM by Joe »
 
The following users thanked this post: Third

Joe

Re: IS ONLINE RNG BROKEN!?
« Reply #24 on: May 16, 2019, 08:20:50 PM »
So according to my calculation, the probability of at least 5 repeats in 8 spins is 1 in 36,010. This means that if I take 10 million samples, each consisting of 8 spins, then 10,000,000/36,010 = 278 (on average) of them should result in at least 5 repeats.

The high level algorithm goes like this :

1. generate 8 spins
2. set the number of "successes" (a sample containing 5+ occurrences of the same number defines a success) to zero
3. determining whether the current sample is a "success"
4. if it is, add 1 to the success count
5. goto step 1 until 10M samples have been collected and analysed
6  print out the number of successes (it should be around 278)

Here's the actual code in C:

Code: [Select] [nofollow]
#include <stdio.h>
#include <stdlib.h>
#include <time.h>

# define SAMPLE 8                  // 8 spins in a sample

int main()
{
  int spins[SAMPLE];
  int i,j,k, spin, successes, hits;
 
  srandom((unsigned)time(NULL));   // seed the RNG
  successes = 0;
  for (i = 0; i <= 10e+6; i++) {   // get 1M samples of 8 spins
    for (j = 0; j < SAMPLE; j++)   // get 8 spins
      spins[j] = random() % 37; 
    for (j = 0; j <= 36; j++) {    // for each # from 0 to 36...
      hits = 0;                    // reset hit counter
      for (k = 0; k < SAMPLE; k++) // look at the current sample
        if (spins[k] == j)         // if a number in the sample is j
          hits++;                  // add 1 to hit counter
      if (hits >= 5) {             // if there were >= 5 hits
        successes++;               // add one to success counter
        break;                     // quit loop (get another 8 spins)
      }
    }
  }
  printf("%d\n", successes);
  return 0;
}
You can test the results for yourself by going to

https://rextester.com/l/c_online_compiler_gcc [nofollow]
Remove the sample code from the editor and copy & paste the above code into it, then click the "Run it" button (or press F8). Obviously you won't get 278 every time (it's called variance!) but the number will cluster around this value.

Here's a screenshot of a sample run:
« Last Edit: May 16, 2019, 08:24:10 PM by Joe »
 
The following users thanked this post: kav, Third, UnlikelySam

Third

Re: IS ONLINE RNG BROKEN!?
« Reply #25 on: May 16, 2019, 09:29:12 PM »
Fantastic! :D
 

sam41

Re: IS ONLINE RNG BROKEN!?
« Reply #26 on: May 16, 2019, 09:34:40 PM »
Well I saw the number 7 had hit 4 times in a row on a live dealer wheel just yesterday, so strange things like that can happen. Last week I saw 25 hit on every other spin 5 times, so it was 25, another number,25, another number etc. Five times in a row! Then finally 25 didn't hit after another number, but guess what? It hit on the next spin!

That said I certainly don't play RNG anymore after many uncomfortable experiences. Once on sky Vegas I was playing dozens 1 and 3 and dozen 2 hit....11 times in a row! Has anyone ever seen the same dozen or column hit on a real wheel that many times in a row? I think 8 is the most I've seen. But just the fact it also happened to be the dozen I wasn't betting, it was laughable.

I took a screen shot and sent it to sky to complain, knowing I was wasting my time but I wanted to at least show them the fix isn't even subtle and is actually so blatant it's funny.

 

Third

Re: IS ONLINE RNG BROKEN!?
« Reply #27 on: May 16, 2019, 11:30:12 PM »
Once on sky Vegas I was playing dozens 1 and 3 and dozen 2 hit....11 times in a row!

.999996 chance that simply won't happen.  This is a 1:250,000 coup attempt event!

I would be watching this casino very carefully!

The max I can simulate is 16.
 

sam41

Re: IS ONLINE RNG BROKEN!?
« Reply #28 on: May 17, 2019, 12:02:14 AM »
Exactly. I told them it wasn't realistic that could happen if it's meant to give results like a real wheel but just got the predictable standard reply about it being random and tested etc.

Makes me happy when I take money off them now via the live dealers. I change casinos every day but I play sky Vegas about twice a month and happened to play there today, shifting another £140 from their account to mine. 😀
 
The following users thanked this post: Third

Third

Re: IS ONLINE RNG BROKEN!?
« Reply #29 on: May 17, 2019, 12:17:33 AM »
I mean they are likely to be correct about it but it does cause one to raise the eyebrow more than slightly...