Last week John Hussman posted a chart that made me take note. Even though I fall in the camp that future market performance cannot be predicted, I do hold that one can forecast probability. Pretty cool concept, I am saying you have more chance predicting a prediction than simply making a prediction. In the chart below Hussman presents a time series of returns that meet a 3% down day filter with a look back of 3 years. As you can see very noticeably is that large drop days cluster around each other. If they were completely random you would see far more sporadic spikes in the chart. What is clear is that during big drawdowns of the Major Correction variety we see many large drawdown days sequentially. This no doubt makes for an interesting discussion for so called randomness.
I really like this chart, note to self and a little bit of homework, see if you can recreate this chart in R.