Look very closely at my 2 screenshots.  You'll see that it shows a RANGE of numbers,  not a single number.  The ranges are 0 through 10 and 9987 though 10000.  In 0 through 10 you can clearly see that at least 95% if not more of the numbers have all been hit at least once and in some cases even hit MULTIPLE times.  0 was hit only once.  If you're a developer at all,  you can see that all I did was select ALL numbers and then sort them based on ASC (ascending) and DESC (Descending) orders.  The fact that the 9000 range showed 9987 is alarming in and of itself without any other scrutinizing.   But if you look at the upper tier numbers,  you'll notice that 9994 through 10000 are all missing from the data.  This is because they have NEVER been hit.  Ever.  The odds of all of them not being hit in order like that is so small that you couldnt store that number to represent it on modern hardware.  So 9994 - 10000 missing would be like 0 through 6 not existing at all in the lower tiers and from the data you can see how unprobable that would be.  In fact,  you could select ANY range of 6 numbers and you'd see that there wouldnt be any chance of ALL of them missing if you have this many rolls for a sample size.  The sample size we're using is greater than 30000 if you combine my own 18K with all the others that have noticed the same thing.

The casino tries to claim that it's "variance",  but variance would not explain this oddity since it's not just a single number,  it's a complete range.  The only explanation is cheating in the form of not all numbers have the same probability of being rolled.  Some have a higher chance than others.  If that were the case then what you would see is that entire ranges of numbers would show less hits than those with a higher probability and thats exactly what we've documented here.  In the data you can also see that even 9886 - 9985 has a lower chance of being hit than 0 - 9885.  It's not as noticable however because the probabilities differ based on 'tiers'.  There are 6 tiers,  each tier has it's own probability level.  This is easily codeable but VERY difficult if not impossible to prove without very large sample sizes (which we have now since the passage of time has allowed for that data to be gathered).