News for hackers. Stuff that matters.
Re:If you want to start a startup, go work for someone else by lutusp on Saturday September 06, 2014 @12:44PM (#8278743)
> According to the author: - 18% success rate for time founders - 20% success rate after that ... so just trying 5 times should assure 1 success.

Not according to the Binomial theorem. For a success probability of .2 per independent trial and a binary outcome (i.e. success or failure, no intermediate states):

     Trials  Probability of >= 1 successes
     -------------------------------------
       1      .20
       2      .36
       3      .38
       4      .48
       5      .67
      10      .89
     100      .99
But the probability of success never equals 1.

Source: http://arachnoid.com/binomial_probability

Re:If you want to start a startup, go work for someone else by gfodor on Saturday September 06, 2014 @01:49PM (#8278937)
These aren't independent events. The OP may be right, but for the wrong reasons.
Re:If you want to start a startup, go work for someone else by lutusp on Saturday September 06, 2014 @04:47PM (#8279420)
> These aren't independent events.

I agree. There are a number of ways this could be true -- a person has more experience, is more likely to succeed in later tries because of being seasoned -- or might even be so discouraged by a first failure that he is then programmed to fail. But to a first approximation, the tries are independent. It only shows the limited applicability of theoretical analyses to real-world events.