My experiment in predicting celebrity deaths ended in failure last week. No one on the list died in the timeframe I predicted, so I have to wonder what went wrong. I think my method of choosing the sample celebrities was flawed in that I used the first 50 celebrities I thought of (25 men and 25 women). A better way would be to take a larger group of celebrities (2000 or so) and use a computer to pick 50 random ones out of that group. That way, any biases I have toward younger or healthier celebrities could be factored out.