How to create a super predictor


Still remember the legendary World Cup predictor Paul the Octopus? In his short life (May he rest in peace) he has not missed a single prediction of the World Cup games. Now, we have the ape Eli, who’s been right in predicting all the last 6 Super Bowl results. 

These are super natural creatures sent by God, right? Otherwise how could they be so accurate?

Well, don’t be blowed away. In fact by using just a bit of probability common sense, you can create such a super predictor yourself, no mater it’s a cat, a dog or whatever animal.

Let’s start with 64 animals. The first year, you let them predict the Super Bowl and the most likely out come is that 32 animals choose the winner to be team A and the other 32 choose team B. Report the 32 or so animals who are correct to the media, keep them, and discard the other 32 who are wrong. The second year, repeat with the 32 animals, and the most likely result is that you get 16 animals correct so that they are kept. Repeat this procedure year by year, and in the 6th year, it’s most likely that there is one animal left. And you can report to the media: here we have a ape(or whatever animal) who has correctly predicted all the last 6 Super Bowls!

Not that amazing any more, right? Because you know that according to the lows of probability, even if all the animals know nothing about what they are doing and they just pick sides randomly, you still can get this result. That is how this whole thing works! Have you seen any reports on an animal predictor who were right before but is wrong this time? Maybe a few, but the majority will not get reported because they are not interesting anymore. Only those who continuously get them right get reported in the media. With this selection, we can always generate “super predictors” from a bunch of animals out there. But can you trust these super predictors more than others? No. Because they always make the choice randomly. There is nothing special in them at all.

So next time when you want bet on some game (although generally I advice you avoid this practice in which you always lose in a statistical sense), you’d better trust your gut rather than those super predictors.