In this report, quantitative analysts Victor Ritseson, Balasurya Sivakumar, and Binh Minh Tran researched and compared traditional trend strategies to machine learning when trading future contacts on commodities. The analysis was conducted on 15 years of data and included for the agricultural commodities: wheat, soybean, and corn, for metal commodities gold, silver, copper, palladium, and platinum, and lastly for the energy commodities crude oil and natural gas.
The method included creating three different approaches, the first approach consisting of solely using the MACD indicator, the second approach utilized the combination of MACD and the EMA of the total traded volume of the commodity, and the third approach implemented a machine learning model. Furthermore, this report investigated the three different approaches on the short-, medium- and long-time periods.
The results varied a lot but the conclusions that could be drawn were that the approaches gave mostly positive results and showed potential to be used in the real market. However, we identified weaknesses in each approach that could be improved if further research were to be done.
To read the full report, please see attached PDF below.
