Decided to throw up a bit of quantitative analysis I did on the weekend.
First of let's start with a chart of BHP, plotted alongside a Bollinger(20,2)
So the first step after acquiring the a years worth of daily closing prices for BHP was to calculate daily returns, as well as the 20 day average and 20 day stddev.
From there I calculated the usual Bollinger %B values and lagged them by one period. Charting this will allow me to see what sort of daily returns are associated with different values of yesterdays %B.
I'm not including all of the analysis here, but happy to share the below charts so people can get some ideas.
You can see in the top chart the "average one day return in %", split into deciles based on yesterdays reading of %B. "-20" for example, means the average returns for all days where yesterdays %B was > -20 and < -10. "90" for example means the average returns for all days where yesterdays %B was > 90 and < 100 (you can see from the first chart that there were no closes above the upper bollinger for the last year so there were no readings >=100).
The bottom chart displays how many occurrences of each %B decile there were.
Interpretation, a few things seem noteworthy immediately: returns at extreme %B readings tend to correspond on average to overbought/oversold. If yesterday %B was >70 and BHP is in a down-trend then intraday returns to the short side are pretty decent. Alternatively, breakouts of the lower bollinger band (%B <0) have provided good long side returns over the last year. You can also see the high frequency of days in the 20-40% range, a definite indication of long term downtrend if most of the bars over the last year are below the 20SMA.