This is what people always say. But testing for randomness requires more than testing for average results over a large sample size. Testing for serial autocorrelation has not, to my knowledge, been formally done. But consider:
9 resists in a row: 0.31^9 = 1 chance in 37,822
I went out testing nukes once and after about 500 I had already observed a 12 resist string.
9 resists in a row: 0.31^9 = 1 chance in 37,822
I went out testing nukes once and after about 500 I had already observed a 12 resist string.