Thread:Jkalm/@comment-24577221-20150313023719/@comment-24577221-20150314054341

Let's try it this way... Suppose you plan to flip a coin 4 times. The chance of getting 4 heads is 1&frasl;16, right? That's based on the assumptions that each time you flip, there's a &frac12; chance of getting a head, and each flip is independant, so &frac12; &times; &frac12; &times; &frac12; &times; &frac12; = 1&frasl;16.

Now suppose you get on with the coin flipping, and it comes up heads the first three times. According to your model, since the chance of getting heads 4 times in a row is only 1&frasl;16, your chance of getting tails on the fourth flip is now 15&frasl;16.

It isn't. It's still &frac12;. Same as it always was. The error lies in trying to argue backwards from the statistical prediction of probabilites and known partial results to violate the original assumption that the individual trials are independant.