Derren Brown, stock spammers and doodlebugs

I only caught the second half of Derren Brown's The System last night, but as usual it was top stuff. He managed to convince a woman chosen at random that he had a system which could predict the winners of horse races in advance. Each time he sent her a prediction sure enough the horse won. It looked like he really could predict the races in advance.

The thing is he hadn't only contacted one woman. He had contacted thousands of people. In each six horse race he split the target group into six and sent members of each group a different prediction. So one group was bound to be right. For the next race he split the group that had received the correct 'prediction' in the previous race into six and again sent each sub-group a different recommendation. Again, one group would inevitably be given the correct tip. By the time this had been run three or four times you can see why someone might become convinced that there was a genuine system at work as implausible as it must initially seem.

By the end of the programme he managed to convince the woman who had been given seemingly flawless predictions that he did have a system, and as such she borrowed money to take a punt on his last tip. At this point - with the bet apparently placed - he revealed the 'system', much to her shock and annoyance! This was the only bit of the show that didn't quite ring true as at no point did I really think he had stuck her money on the random horse. He's too much of a nice bloke. But in any case the show provided a great example of how randomness can look structured.

The other great thing about Derren Brown is that he does seem to be on a mission to expose con-artists. And in fact this programme is a variation on technique that has been used in the past, only substituting share price movements for horse racing tips. Imagine if you were sent info from a financial adviser that correctly predicted, say, movements in the FTSE for four or five months on the trot. Wouldn't you be tempted to have a punt? As I am currently being targeted by some irritating stock spammers in my blog comments I was particularly grateful to be reminded of this example!

Finally the scene at the end of the programme where he revealed that he had spent all day flipping a coin in order to produce the film sequence where he managed ten heads in a row was also another great example of how we get taken in by randomness. Much as we 'know' that it is possible for such a series to arise in a genuinely random sample somehow we can't quite believe it is random when we see it. We expect randomness to look more, well, random. A great example of this I read about recently was the way that Londoners interpreted V1 and V2 strikes during the war. The Germans were using petty basic guidance technology and were simply targeting 'London'. However, when the rockets fell randomly across the capital the random distribution included a few clusters. It did not look random. This led people to believe that certain areas were more likely to get hit than others.