Comment Re:Read the article. (Score 1) 136
You should read the article. There's a little section about "wisdom of crowds" and then the balance of it is about particular people they've selected as being super accurate, such as the pharmacist they use as an example. If you take enough people and ask them to guess randomly, some of their guesses will line up very nicely with the answers to any questions. Purely by luck. If you cherry pick these randomly lucky guessers and don't properly allow for your cherry picking in your calculation of expected performance, you will be misled badly.
A different version of the same phenomenon confuses people who try to write classifiers. I have a friend who was trying to classify patients who did or did not have a disease. He put a bunch of measurements into the classifier, trained it, and look, it was 100% accurate! He was suspicious, so he put in a bunch of randomly generated numbers, trained it, and look, it was 100% accurate! Of course, neither version did any better than chance on data it hadn't been trained on.