Analysis of Monthly Equity Changes

The impact of a given system on your equity curve will depend on sys­tem design and money-management decisions. In this section we look at the monthly equity curve, to understand month-to-month performance. We follow standard accounting procedures and look at the profit and

Analysis of Monthly Equity Changes 195

loss figures at the end of each month. You may wish to look at the equity curves on a weekly basis, but the random noise in the market often com­plicates the analysis of such detailed data.

We saw in the previous section that the standard error of the linear regression provides an excellent measure of the roughness of the equity curve. However, the linear regression approach does not show how much money the system lost over a 1-, 2- or 3-month period, nor does it reveal the maximum cumulative loss. We also would like to know what percentage of the months showed profitable returns, and whether the curve becomes smoother when we add certain markets or change the portfolio mix. Another useful bit of information is how quickly the sys­tem recovers from a losing streak, measured in months between new highs.

"You must remember that this analysis reflects past data, not how the system will do in the future. However, if you use average numbers and standard deviations, you can get a fair estimate of future perform­ance. You can then decide how to capitalize the system, by quantifying the equity swings you can tolerate. Thus, analyzing the equity curve on a portfolio basis gives a deeper understanding of system performance, and you can better prepare for future equity swings in real-time trading.

Most of this analysis was done in a spreadsheet, since the popular system testing software does not provide this information. We first used Omega Research's System Writer Plus™ software to generate the daily equity curve using real contract data with rollovers, because we found that continuous contracts were not giving reliable results. We then used Tom Berry's Portfolio Analyzer™ software to summarize these data into monthly performance numbers. You can do the same using a spread­sheet, or you can write a simple program.

Once in the spreadsheet, we calculated the actual dollar changes in equity over 1, 2, 3, 4, 5, 6, and 12 months. We could then quickly calcu­late the best performance, the worst performance (drawdown) and standard deviation of profits over each period. The advantage of making the dollar equity change calculations was that we could clearly see the effects of a particular exit strategy or of combining different markets in a portfolio. Some sample calculations will give you a feel for analyzing equity curves at a portfolio level.

We used actual DM contracts from automatic rollover on the 20th day of the month preceding expiration. The test period was February 1988 through June 1995. We allowed $100 for slippage and commis­sions and used the 65sma-3cc system with one-way entries to test differ­ent exit strategies. A one-way model does not allow back-to-back entries of the same type, so that you will not see two consecutive long or short

196 Equity Curve Analysis

trades. Thus, the number of entries over a data set is constant, allowing an apples-to-apples comparison.

Figure 6.14 is the monthly equity curve for the 65sma-3cc system with a $5,000 initial stop and no other exits. Thus, the long entry was also the short exit signal, and vice versa. The large stop makes it a better test of the inherent robustness of the entry signals. The system reported a paper profit of $24,900 from February 22, 1988 through June 20, 1995, with a profit factor of 1.34, 35 of 70 profitable trades, and a draw­down of-$11,687.

The monthly equity curve (see Figure 6.14) shows the overall ris­ing trend with many sharp equity retracements, which occurred during trading range markets following a strong trend. Interestingly, rolling over the contract captured most of the profits in an uptrend, better than most exits. However, the system gave up most of the profits in the con­solidation that followed the uptrend, suggesting that filtering this model should smooth out the equity curve. You could have deduced this infor­mation by studying the charts of each contract tested. However, the eq-