Wednesday, February 9, 2011

Quanting India: Sensex Mean Reversion study

Markets display a very peculiar phenomenon called Mean reversion. Without this there the prices would be just random. See more on this concept at Wikipedia. Link here.

There can be multiple variations of this concept which are applied to markets. Like closing price relation to high or low etc.

One widely followed is the mean reversion to moving averages like 50, 100 and 200 day MA. There are already well defined trading strategies based on the price and the moving averages.

For today let us consider the 50 Simple Moving Average and Sensex. I am taking Sensex because it has longest price history in Indian indices so the strategies can be back tested well.

To look for mean reversion to 50 DMA lets plot the Sensex closing w.r.t 50 DMA.

We get the following chart:



There are few spikes but more or less the prices have stayed within a band. Doing the statistical analysis to the values we get the following:



The values are consistent with the bullish bias of Indian markets for last 30 years.
Going further I calculated the lows of the data set in the bull phase and high points in the bear phase. These values can tell us the probable points where the buying in done in bull and selling in bear phase. And this particular trade has the highest probability of success till our view of market phase is correct.

Well now how to trade with the above strategy at this point.
Looking at the extreme right hand side of the chart of last 2 years we can have some idea about the current setup. The ratio is at the lows of the bull run where generally the buying comes.


Does this indicate that buying can be done at this point. Well there is still no confirmation in the reversion strategy as of now as there are a few elements missing for triggering buy at these levels.

I will update in future posts if there is any indication in that respect.

The above post is meant for quant strategy discussion rather than any investment advice.


No comments:

Post a Comment