Therefore, the correct way is to trade a system alongside a clone system. The clone takes all signaled trades while the original system takes the real trades allowed after a given date. The clone system is used to filter these real trades as follows:
- If the clone system is long on our start date, we do not allow any
real long trades until the clone system goes flat or short.
- If the clone system is short on our start date, we do not allow any
real short trades until the clone system goes flat or long.
Doing this means that we only take the new trades and the first signals that have been tested. This approach is termed "sync technology."
If we do not adopt this approach, then we have one of two options. We have to use simple money management, which can be calculated manually, or we assume the risk of taking secondary untested signals that could lose money.
Sync technology is so important that it has become a standard operating procedure that I generically implement on any system I employ. We also address another issue. Because of money management, we must decide what to do if a trade’s initial signal is skipped and we later decide to allow the trade. This would give us the same secondary signal problem we have when we just started trading. We easily can resolve this by applying the sync technology rules to these trades.
Spreading the risk
Many commodity trading advisor (CTA) programs trade a single trend-following strategy on a basket of markets. This is why the Barclay CTA index produces the number it does, as follows:
From January 1980
Compound Annual Return 11.60%
Sharpe Ratio 0.42
Worst Drawdown 15.66%
Correlation vs. S&P 500 0.01
Correlation vs. U.S. Bonds 0.11
Correlation vs. World Bonds 0.00
This gives us a minimal acceptable return (MAR) of 0.74 and a Sharpe ratio of 0.42. These results don’t look that good, but they are better than the average because of survivorship bias of the index. CTAs that perform badly are removed from the index.
The problem is that all commodities correlate during times of major events, so the correlation risk is greater than implied. To really diversify, you must combine multiple systems of different core logic, such as combining a trending strategy for a commodity basket with an arbitrage system for Treasury bonds or a countertrend stock index approach. Developing multiple system trade plans, in which each system can trade a basket of markets, creates money management strategies that are so complex that a computer is required for implementation. You also might want to allow different allocation units between the systems.