Futures trend following strategy from "Trading Evolved"

Hi all,

I’ve been going through Andreas Clenow’s Trading Evolved book, which I do recommend if you’ve not already checked it out. Lots of interesting strategies, with full zipline implementations.

I’ve been trying to replicate the futures trend following strategy backtest using futures data from Norgate (with a view to applying to to CFDs). I chose Norgate over CSI (used in the book) since Norgate provide a very nice zipline integration library: zipline-norgatedata · PyPI

The trouble I’m having is that my backtest results don’t match up with the results in the book very well. For example. in the first month of the backtest Jan 2001 the book records -4.6%, whereas I see a loss of -1.4%. For 2002 the book records +18%, whereas I see just 1.8%.

The code and settings are identical (different pandas version, but I think it’s unlikely this would have a material impact).

Has anyone else read the book, and tried to reproduce the results?

Thank, Mark

I also tried to replicate the core trend following model from the book using Norgate data. The results were different, but still very similar, to these in the book. However, the book uses some obscure futures symbols and it was very difficult to achieve a 1:1 correspondence in the investment universe. Some of the symbols are also not supported by Norgate data, most notably ‘BL’ which stands for Milling wheat, and is a mistake. Also, Norgate’s database and CSI’s database very likely differ in some ways, not to mention the bundle creation process is very different on both platforms. So it makes sense not to have a 100% similarity in backtest results.

Overall, my replication backtest achieved very similar results to Andreas’ backtest, namely:

  • Annual return 12% (book is 12.12%)
  • Sharpe 0.70 (book is 0.69)
  • Max drawdown -27.9% (book is 25.5%)
  • Annual volatility 18.8% (book is 19.4%)

So I decided the comparison was satisfactory.

1 Like