Premium Bond Winner ?

Hi, ive had a massive £5 worth of bonds since 1978 and never won anything. My brother got £500 worth of bonds in july 1998 and won £50 that december but not a penny since. Thinking of buying some more, dreaming of the big one

I've had ones since I was little (35+ years) and won about 3 times when I was young, was bought £50 in 1987 and won £50 a month later, nothing since. Have also been toying with the idea that I should buy some more. My 5 year old has £100 and hasn't ever won anything. I am also thinking about buying her another £100.

Can't decide but believe in spreading the risk - she has £2K in an investment fund (pay £30 per month in), £450 in children's bonds, £600 in 'pocket money' account and maybe £300 in a friendly society account, plus one that holds £600 and is added to by £10 per month. Maybe she will have enough to buy a bag of sweets in 20 years time! Still, if I don't do it, she won't have anything.

Actually I think the problem is that people do not understand the statistics at all.


Humans are genetically inclined to try and see patterns in anything. This was so that we could spot the outline of animals or people hidden in grassland or forest, however the same genetics apply when we view the impact of random number generation. We start looking for patterns even though there isn't one. We are also "designed" to apply our past experience to the future, even though it may not be sensible. For example if water was hot last time, you won't put your hand straight in again without testing, even though this water could be hot or cold.

This means that we have a tendency to try and extrapolate from previous events and come up with behaviour patterns even though there is no sensible reason for doing so.

The other problem is that a series of random numbers never looks random to most people.

An example: You toss a coin ten times. If it comes up heads 10 times, you conclude the coin is dodgy. If it comes up 5 heads and 5 tails, you conclude that the coin is "fair". And yet both outcomes have the same probability of being generated entirely at random.

The OP has formed a theory that more recent PBs are likely to win than old ones, based purely on the fact that HIS recent PBs have paid out more than the old ones. However this is a theory extrapolated foward from random number selection. If the PBs that won were the old ones, would the theory change? If so, why? What would be the reason for some numbers winning more than others?

There isn't one. What you have noticed is just random behaviour at work, combined with a very human tendency to try and make a pattern out of it. The same problem has arisen with the study of telephone masts and cancer rates, and autism and the MMR jab. People have concluded based on possibly random data without the backup of a theory as to WHY a link between the to might exist is in place. This is called "data mining": looking for something in data and then forming a theory around it, rather than forming a robust theory and then testing it with data.

If you want a more frightening example of similar behaviour, visit Motley Fool and look at all the idiots who are attempting to come up with foolproof stock market investment methods based on the study of past performance charts.

The way to prove that the selection of premium bond winners is not random is not by collecting data, as that proves nothing. You need to come up with a valid reason as to why it might be happening, then use the data to back up your theory.

Cheers.