Types of Diversification

THREE TYPES OF DIVERSIFICATION

Diversification is among the most important and underutilized tools available to traders and investors because it allows improvement of our rates of return without proportionately increasing risk assumed to achieve these enhanced levels of performance. The most commonly employed type of diversification—asset class diversification—has already been discussed in Chapters 3 and 4, where we looked at how diversification among assets that had low correlations improved our overall performance. A review of Tables 3.2 to 3.13 and Tables 4.4 to 4.8 shows that diversification almost always yielded improvements when compared with the performance of individual assets.

This chapter focuses on the two other diversification methodologies: adaptation of different parameter sets for the same trading system and combining of negatively and/or uncorrelated trading systems.


DIVERSIFICATION OF PARAMETER SETS

Assuming that a trading account has adequate equity under management, it is preferable to diversify parameter sets rather than to trade multiple contracts with the same parameter set. Although there maybe strong positive correlations between parameter sets of the same trading system, Tables 7.1 to 7.20 show that even minor modifications to parameter sets can make the difference between an overall profitable or losing outcome. Furthermore, as shown in Chapter 7, because we can never be certain as to which parameter set will outperform in the future, parameter set diversification greatly aids in minimizing regret. Minimization of regret in this context strengthens our psychological ability to adhere to a disciplined and consistent (e.g., systematic and/or mechanical) approach toward trading.[2.]

A comparison of Tables 9.1 and 9.2 exemplifies this final point. Table 9.1 shows the results of various parameter sets on the two moving average crossover system for IMM Swiss franc during the in-sample period of 1993 to 2002. Notice that the best-performing parameter set in this Table was the 10- and 29-day moving average crossover; the second-to-worst-performer was the 7- and 20-day parameter set. Compare this with Table 9.2, which is the same system on the IMM Swiss franc for the out-of-sample year of 2003. Not only is the best-performing parameter set of our in-sample period now the worst performer, but also our second-to-worst in-sample performer has now become the top-performing parameter set.

Table 9.2 is even more instructive in the context of diversification when we compare the performance of the 7- and 20-day and the 6- and 20-day parameter sets. Although these parameter sets retained identical longer-term moving average parameters and the shorter-term moving average parameter was changed only by one step, the 7- and 20-day parameter set was the year’s top performer, while the 6- and 20-day parameter set remained in the bottom half of all parameter sets analyzed.

MECHANICS OF TRADING SYSTEM DIVERSIFICATION

Diversification of negatively and/or uncorrelated trading systems is one of the most effective methods of improving rates of return without proportionately increasing the risk assumed to achieve these enhanced levels of performance. To illustrate this point, let us examine a trend-following system from Chapter 3 (MACD) with our diversified futures portfolio, and a directionally biased intermediate-term mean reversion system from Chapter 4 (RSI Extremes with the 200-day moving average filter) with our mean reversion portfolio, and then compare these results with the combined performance of both trading systems.

In comparing Tables 9.3 and 9.4 to Table 9.5, the first and most important improvement is in the profit to maximum drawdown ratio. This is due to the fact that low correlations between the trend-following and mean reversion systems led to a smoothing of equity drawdowns for the performance of the combined trading system results. Although the maximum drawdown column shown in Table 9.5 was larger than in Table 9.3 or 9.4, it represented an increase only of roughly 17 percent and 20 percent respectively. By contrast, because Table 9.5 took all signals generated by both systems, its total net profits were additive, thereby leading to an overall improvement in performance results.

In addition, combining these uncorrelated trading programs lessened many of the deficiencies of both methodologies as stand-alone systems. For example, one of the drawbacks to the trend-following system as a standalone solution is that it experiences more losing trades than winners. By contrast, by combining these two systems, the winning trade percentage increased from 42.85 percent for trading the MACD system alone to 50.82 percent.

Because these two trading systems are not highly correlated, sometimes both will generate profits; sometimes one will profit while the other loses; and sometimes both will lose. Consequently, the only way to replicate the backtested performance of these combined system results is through consistent implementation of all signals generated by all assets and/or trading systems. In other words, traders should not try to outguess the systems.

Although consistent implementation of all signals for all assets sounds like a straightforward proposition, it is complicated by the fact that both systems could be trading the same asset. In fact, this was the case for the combined trading system results generated in Table 9.5, because both the trend-following and mean reversion portfolios contained the E-mini S&P 500 futures contract. Consequently it is quite possible that these two trading systems could have generated opposite trading signals for the same instrument.

When I first started trading multiple systems with low correlations, I encountered this problem of conflicting trading signals. I failed to take a buy signal in the trend-following system because my mean reversion system had generated a sell signal for the same instrument. During the overnight trading session, my mean reversion realized its profit, which corresponded to what would have been a temporary open equity drawdown in the trend-following system (had I taken that trade). Then, almost immediately after the mean reversion system’s profitable exit, the market reversed, and I awoke to find that I had missed out on one of that year’s most profitable trend trades.

This painful lesson reinforced the fact that a prerequisite to successful implementation of diversified trading strategies is never missing a trading signal. Subsequently I have found that the simplest and preferred solution to this problem of conflicting signals for the same asset is the maintenance of two (or more) separate trading accounts—one for each distinct trading methodology (e.g., trend-following, intermediate-term mean reversion, short-term).

Is this book interesting to you ?, If yes, Grab the book now.
Read cusomer reviews before you get the book

Types of Diversification