We all know that Random Number Generators in computers don't generate true random numbers, but instead generate pseudo-random numbers. Also, some RNGs are better than others, and some are implemented better than others.

What are some examples of when a poor RNG has been used, or an RNG poorly implemented, and it has been exploited ?

Examples that I have found are

  • 2
    $\begingroup$ There is another related set of issues, of scientific/statistical monte carlo studies done using poor RNG that were later found to be bogus. Sadly I'm not contributing much because I can't remember the reference, but it has definitely happened... $\endgroup$ – Corone Feb 22 '16 at 8:29
  • 1
    $\begingroup$ From memory, the UK game show Deal or No Deal originally used (pseudo-)random numbers generated in Excel, and a viewer was able to crack the problem of what box contained what prize. But I don't think the prize was exploited as such. $\endgroup$ – Silverfish Feb 22 '16 at 9:08
  • 3
    $\begingroup$ See also this Wired article about exploiting pseudo-random numbers on lottery scratchcards. $\endgroup$ – Silverfish Feb 22 '16 at 9:17
  • 5
    $\begingroup$ For people voting to close: I think this post is safely on-topic here. Both (pseudo)random number generation and the history of statistics are clearly on-topic, and I'm not sure what aspect of this intersection would render it off-topic. Even if one were to argue "ahh, but only the mathematics of RNG is on-topic here" (which would be very reductionist in my view), a really good answer to this question - the type I am hoping to read - will explore what mathematical details allowed the exploit to take place. $\endgroup$ – Silverfish Feb 22 '16 at 12:00
  • 1
    $\begingroup$ @Corone: Are you perhaps thinking of RANDU en.wikipedia.org/wiki/RANDU ? $\endgroup$ – David Cary Feb 23 '18 at 1:34

A lottery scheme in Ontario used poorly designed random generation, which was spotted by a statistician, Mohan Srivastava of Toronto, Canada, who notified the Ontario Lottery and Gaming Corporation of the issue, rather than making a hefty profit out of this loophole.

  • 1
    $\begingroup$ There exists a wider set of issues that RNGs feed into. For instance, the sampling used in any experimental design (simple to complex) whether RDD phone surveys, online river samples, election polls, prediction markets, etc. The polling and prediction market shortfalls in giving a high likelihood to a Clinton victory over Trump echoed the 1948 pre-election polling errors giving Dewey the victory over Truman. Related to this are the decennial hoops the Census Bureau has to jump through with each national census to plug and/or pad values for the thinly estimated or missing information. $\endgroup$ – Mike Hunter Nov 18 '16 at 11:56
  • $\begingroup$ @DJohnson: first, this story is mostly anecdotal, I agree. Second, I have trouble seeing the connection between the recent poll failures and RNGs. Or with the statistical correction made by the Census Bureau (and INSEE here). $\endgroup$ – Xi'an Nov 18 '16 at 12:25
  • $\begingroup$ Yup. I understand these difficulties which is why I made this a comment rather than a response. Actually, I now wish I had not placed it in the airstream following your thread rather than as a more general note immediately after the OPs query. My position is that concerns about randomness (or the lack of it as in selection bias problems) either underlies or undermines the adequacy of any quantitative answer to a question. Therefore, I chose to widen the net from the very narrow focus on RNGs in noting these wider concerns. You don't have to agree. $\endgroup$ – Mike Hunter Nov 18 '16 at 12:37

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.