I came across this video series over the weekend, an option trader discusses how he trades credit spreads (mainly looks for mean reversion). Most of you will be familiar with bollinger bands as a common mean reversion strategy, essentially you take the moving average and moving standard deviation of the stock. You then plot on to your chart the moving average and an upper and lower band(moving average +/- n*standard deviations).
It is assumed that the price will revert to the moving average hence any price move to the bands is a good entry point. A common problem with this strategy is that the moving average is a LAGGING indicator and is often very slow to track the price moves if a long lookback period is used.
Video 1 presents a technique called “linear regression curves” about 10mins in. Linear regression curves aim to solve the problem of the moving average being slow to track the price.
Linear Regression Curve vs Simple Moving Average
See how tightly the blue linear regression curve follows the close price, it’s significantly quicker to identify turns in the market where as the simple moving average has considerable tracking error. The MSE could be taken to quantify the tightness.
How to calculate the linear regression curve:
In this example you have 100 closing prices for your given stock. Bar 1 is the oldest price, bar 100 is the most recent price. We will use a 20day regression.
1. Take prices 1-20 and draw the line of best fit through them
2. At the end of your best fit line (so bar 20), draw a little circle
3. Take prices 2-21 and draw the line of best fit through them
4. At the end of your best fit line (so bar 21) draw a little circle
5. Repeat upto bar 100
6. Join all of your little circles, this is your ‘linear regression curve’
So in a nutshell you just join the ends of a rolling linear regression.