Debugging

        Stats, correlation vs cause and effect.  Coincidence (tom and
        the port scan, our line drops).  280 million people in USA;
        every day 280 1-in-a million events happen.  Stats are useful
        for summarizing data, extrapolating from subset, and finding
        correlations.  Avoid superfluous confusing details.  Use
        example of two twins killed within 2 hours on same highway in
        Finland on march 5th. both on bikes.  Ask rather "how
        surprising is it?"  Surprising that 2 men killed <2 hours on
        same day in same town?  no.  fact that they were twins makes
        no difference.  Not unusual for two people to be killed in
        snowstorm on a busy highway on bikes.  Blades of grass in a
        meadow.  Pick any one of them and it's like 1ina billion, but
        it's still a certainty you will pick one.  Talk about feyman's
        example of finding that exact plate on car in lot before
        talk.  Find his example in the book.  You cannot use post
        examination to evaluate probability...it already happened.
        "Six easy pieces".  Never ever can verify an idea with data
        you used to get the idea.  What's te chance I see that license
        plate!  Einstein and the orbit of mercury
        then eclipse.  We are hardwired to explain things and find
        patterns; leads to lots of false positives.