on a random string of digits number 123 appears every 1000 digits, 221 every995digits

[salmo]

I agree that the digits of pi throw up some rather interesting combinations, but you just cannot use these digit combinations to make a statements about probability of digit combinations from a random data set or a gambling table.

Your statements about statistical likelihood assumes that each trial (the selection of the next digit) is indepenent of all other trials, but this is not the case. The next trial (next digit selection) is dependent upon the previous digit in the sequence of Pi since any repention of Pi the nth decimal place is fixed and repeatable. I's be interested to see your results once you've repeated your analysis with a different (random) dataset.