Dynamic statistical trading strategy
- 1. Dynamic trading strategy Managed futures fund strategy Andrej Ogorevc
- 2. Dynamic trading strategy Create monthly returns drawn from a desired distribution Liquid strategy – dynamically trades only most liquid assets No lock up period No style drift No need for extensive due diligence No lack of capacity No regulatory complexity No credit of counterparty risk Low transaction costs – trading futures This is a risk control system that allows investors to design futures trading strategies, which generate returns with pre-defined statistical properties, i.e. returns with a pre-defined volatility, correlation with stocks and bonds, etc. Once the desired return distribution is selected, the strategy simply tells the investor every day how many futures contracts to buy or sell. Following these instructions will generate returns with the desired properties. It can be applied to the creation of completely new risk-return profiles, which are not available elsewhere in the market. Investors no longer have to rely on a limited supply of funds or make a guess at a fund’s true risk-return profile. We can structure tailor-made synthetic funds, which produce returns with exactly the properties investors are looking for. Distribution Example 1 Basket: Russell 2000 (RTAA Comdty), S&P 500 Index (ESA Index), Gold (GCA Comdty) The strategy employs an option hedging technology to replicate a monthly payoff drawn from a desired distribution. In our first example we are targeting a distribution where monthly returns lower than -3% will exist only with a very small probability and almost 100% of our monthly returns will be below 4%. This feature controls the volatility of the strategy and the drawdowns. Half of our desired returns will be greater than 1%. In this example the strategy dynamically trades 3 reserve assets from the Basket 1 and cash in order to replicate the distribution of monthly returns equal to our desired distribution of returns. The strategy is testes in the period from April 2001 to April 2013. Given the large size of most futures contracts and the fact that it is impossible to trade less than 1 futures contract, the minimum amount that can be invested in such synthetic fund is around USD 10 million.
- 3. Figure 1: Distribution of monthly payoffs if 100 is invested at the beginning of each month Figure 2: Number of Contracts held on a given day
- 4. Every day the strategy dynamically readjusts the number of contracts in order to replicate the payoff drawn from the desired distribution. The strategy is, with few exceptions, always short S&P 500 and long Russell 2000 and Gold. The number of contract held every day does not change excessively and therefore it does create excessive transaction costs. Round trip transaction costs of 4bps are taken into account. By dynamically trading the reserve asset and cash we obtain the actual distribution of returns very close to the desired distribution. Figure 3: Distribution of monthly payoffs if 100 is invested at the beginning of each month Assuming an investment of USD 1m was made in April 2001 the NAV of the strategy would grow to USD 3m in April 2013 with maximum draw dawn of 9.6%. On average the strategy was only 1.76% below its last observed maximum. In the observed period the strategy had a Sharpe ratio of 1.14.
- 5. Figure 4: NAV of the Trading Strategy Table 1: Descriptive statistics of Strategy’s daily returns Distribution Example 2 Second distribution example has lower targeted volatility and but keep the capital protection feature since it targets to generate monthly returns that are greater than -3% with 99.2% probability.
- 6. Figure 5: Distribution of monthly payoffs if 100 is invested at the beginning of each month Figure 6: Number of Contracts held on a given day
- 7. Every day the strategy dynamically readjusts the number of contracts in order to replicate the payoff drawn from the desired distribution. The strategy is, with few exceptions, always short S&P 500 and long Russell 2000 and Gold. The number of contract held every day does not change excessively and therefore it does create excessive transaction costs. Round trip transaction costs of 4bps are taken into account. By dynamically trading the reserve asset and cash we obtain the actual distribution of returns very close to the desired distribution. The strategy also allows targeting the correlation of the returns with other portfolios and therefore enables to create a good diversifier. Figure 7: Distribution of monthly payoffs if 100 is invested at the beginning of each month Assuming an investment of USD 1m was made in April 2001 the NAV of the strategy would grow to USD 2.34m in April 2013 with maximum draw dawn of 6.75%. On average the strategy was only 1.51% below its last observed maximum. As expected from the choice of the desired distribution, the volatility of the strategy over this period was lower and was at 6.6%. In the observed period the strategy had a Sharpe ratio of 1.05.
- 8. Figure 8: NAV of the Trading Strategy Table 2: Descriptive statistics of Strategy’s daily returns
- 9. Summary This is a risk control system that allows investors to design futures trading strategies, which generate returns with pre-defined statistical properties. It also allows modeling the correlation between the returns of the trading strategy and the returns of other assets thus enabling to create the perfect diversifier. The reserve assets can be chosen from a wide variety of underlyings. The strategy allows for tactical input through the choice of futures contracts to trade. The composition of futures portfolio can be changed whenever and as often as needed, thereby incorporating any tactical views of the fund manager. From a tactical point of view, this strategy is as active or passive as the fund manager needs it to be. By construction, returns are drawn from the desired distribution and, forwardlooking, will therefore have the targeted properties. There can be temporary discrepancies between the target parameters chosen and the sample parameters generated. This is nothing unusual though. When tossing a coin, the chances of heads and tails are 50/50. This does not mean that when tossing a coin 10 times one will always find 5 heads and 5 tails. When the number of observations increases, however, this is likely to be corrected as the sample becomes more representative for the distribution it is drawn from.