Saturday, 16 May 2015

Sector Momentum Model












The above strategy is a sector momentum model which buys the strongest sectors as measured by their weekly returns over a certain look back period and also that are only above the 40 week moving average. So in this way we are trying to identify sectors that are showing momentum and are also in uptrends as defined by the moving average.

Top 1 means we are invested in only 1 sector index which is given a rank of 1 and Top 2 means we are invested in the top 2 ranked sectors. Example, as of 15 May 2015 Pharma and Auto indices were ranked 1 and 2 so the Top 1 portfolio will buy into Pharma only while the top 2 portfolio will buy into Pharma and Auto indices.

The indices considered for this model are :
CNX Auto
CNX Bank
CNX Energy
CNX FMCG
CNX Infrastructure
CNX IT
CNX Media
CNX Metal
CNX Pharma
CNX Realty


The price only returns of both portfolios have managed to beat the Nifty index on a price and total returns basis and also have a much lesser drawdown than the Nifty. Yes there will be periods where the portfolios would lag Nifty returns on a shorter timeframe but on a long term basis I believe they would perform well given the trend following filter that we have applied, and one cannot rule out the benefits of a mechanical strategy over the long term. 

Thanks for reading :) 

6 comments:

  1. Interesting. Can u talk about your backtest process in a bit more detail and what lookback period did you use ? Thanks

    ReplyDelete
    Replies
    1. Hi Anish, lookback period is average returns of rolling 12 and 4 weeks and Top 2 portfolio returns is an average of the 2 sector returns.

      Delete
    2. Used friday closing prices, but obviously the real world is different and unfortunately we dont have sector specific ETF's to implement this if possible

      Delete
    3. True. But if you identify which sectors to back. then 50% of work is done. Have you done work for previous years like 2009-2014. is it possible for you to share raw historical data on the indices if you have it in an excel file ?

      Delete
    4. yes this data is already from 31st dec 2004 onwards, il put a screenshot of the historical rankings from the next post onwards.

      Delete
  2. What are the filters that you apply.

    ReplyDelete