Last week, I was making money with the Commitment of Traders data. But surely, we could change a few things to improve the bottom line?
Directly continuing from last week, I’m going to drop consideration for the two weaker pairs which leaves just AUDUSD and EURUSD. Concentrating on fewer instruments can be wise during the experimentation stage of such tuning.
Actually, leveraging the COT data is more commonly used for other instrument classes like commodities including such classics as gold, crude oil, copper and platinum. Or indices such as S&P500, DowJones and the NasDaq.
Last time, I was just using normalised net position figures, which our Indicator calls Strength. Another type is called Index, which applies stochastics to Strength.
To avoid confusion, other sources also use the term of Index to represent the latest net position expressed as a percentage of the maximal value over the previous period, generally of two or three years.
This is the basic crossover using Index, slightly up from Strength of last week. EURUSD performs better, with an interesting increase in trade count for AUDUSD. That said, COT is only weekly so this is still a mere 65 trades over 66 months.
For Strength, we had good success with varying the threshold required to trade. Basic crossover on a 0 to 100 scale will typically be at the 50 mark, but shorting at 45 and going long at 55 gave superior results.
After sending that through the optimiser, Index did not show any real improvement. But no harm done, and good to try the options. Lets try something else.
Index requires a certain time period for it’s stochastic calculations. The default is 182 days, which is 26 weeks. This example is using 98, or 14 weeks.
As usual, I have to be careful not to overfit my settings on historical data but all shorter time frames were better. Please read a previous article for more on this. Although it’s a little tricky in this case given the low trade trade frequency.
This would be a prime example of when different instrument classes would benefit from different settings. I would be very surprised if metals like copper performed in a similar fashion.
Thus, with a couple of simple tweaks, I have doubled the bottom line figure from last week while using fewer instruments and thus would require less cash to run.
As the last chart shows, the very first trade is the biggest draw down. Yet another example of my luck…