A program that illustrates the likely spread of chance events over a period of time and how past performance does not influence future chances

To have a program in which there are, say, 50 individuals listed which each have equal probabilities of a particular outcome. Say the number generated by rolling two dice. Then, after a given number of 'throws' (number of throws to be able to be varied - typically in the range 50 - 250) to list the sites in order of the number of sites with the greatest number of double sixes. Then, retaining all the data generated so far, to repeat the 'throws of the dice' for another set of throws (number of new throws to be variable and likey to be different form first round of throws). At the end of this second round of throws, to be able to see which sites have the greatest number of 'double sixes'. Background to this is to demonstrate regression to the norm. In the profession tha I work, we try to minimise injuries at work and there is large focus on those sites with the largest number of recorded injuries in the previous year. These sites may indeed have weaknesses that deserve attention but I suspect that a significant number of those with 'higher than average' numbers of recorded injuries are just showing the top end of the random distribution of injuries across sites. I would like to show that when we take the sites showing the highest number of injuries in one year and focus on them, if the high number is purely (or significantly) a chance outcome, then it is likely that, in the following period, we will see a smaller number just by regression to the norm rather than from any perceived beneficial impact of our intervention. I am suggesting the format above based on the number of 'double sixes' in a series of two dice throws as a simple way of presenting this but I'd be very happy to consider any other way of presenting this. Hope this makes sense - please get back to me if you have any questions.