Assuming you use 0 as your magic ball number. Move the magic ball number up just a tiny bit to 0.07 and you would have had a false positive in september 1998. You could also move it down to -0.32 and you would still a 100% success rate, albeit with a sample size of 3...