Forex Trading System- Swing Summary

OVERVIEW

We have introduced quite a lot of new and technical information in the preceding chapters. Our goal in this chapter is to unify the salient estimating routines into a viable tool that traders may incorporate into their existing trading systems.

SELECTING SWING PARAMETERS

In his very informative book, Point and Figure: Commodity and Stock Trading Tech-
niques (Traders Press, 1997), author and trader Kermit C. Zieg Jr. suggests the use of a constant reversal amount, which he sets at three boxes. He prefers to vary the box size according to the underlying security. Table 31.1 is our recommendation based loosely on Mr. Zieg’s idea.

An alternative method of setting swing reversal parameters is the converse of Zieg’s
suggestion; that is, set the box size to the constant minimum price fluctuation (one pip in currency markets), then vary the reversal amounts according to the influence of the parity rates of the individual currency pairs. We recommend that the transaction cost (the bid/ask spread) multiplied by 3 be used to initialize the swing reversal algorithm.

The second column in Table 31.2 is the average transaction cost of several major cur-
rency dealers. However, if Zieg’s method is preferred, then set the box size to the transaction cost, and a constant three-box reversal amount can be used. Essentially, we recommend that currency traders experiment with different combinations based on their different trading objectives.

COMPOSITE AVERAGE ESTIMATES

The idea here is very simple and direct: to generate multiple discrete estimates using the different methods described in earlier chapters and combine them to create one average forecast. In all instances we use one pip as the box size and three-, five-, and nine-box reversal amounts. We also use a three-wave cycle where the heights of the last three waves in the swing data are 15, –12, and 9 (see Figure 31.1):
A = 1.0002
B = 1.0017
C = 1.0005
D = 1.0014

Linear Regression Method

In this section, we apply an ordinary least squares (OLS) linear regression to the same three waves with heights 15, –12 and 9. Technically, the regression method requires only the final two waves and our model becomes:

Wave 4 = (A Wave 2) + (B Wave 3) + C

The estimate for the three-box reversal amount is:

Wave 4 = 0.7119(12) + 0.0871(–9) + 0.0110
Wave 4 = 8.5428 – 0.7839 + 0.0110
Wave 4 = 7.7699

A five-box reversal amount produces:
Wave 4 = 0.0140(12) – 1.1137(–9) + 0.0047
Wave 4 = 0.1680 + 7.8863 + 0.0047
Wave 4 = 8.0590

The estimated height using a nine-box reversal amount is:
Wave 4 = 0.2167(12) – 0.8128 (–9) + 4.1850
Wave 4 = 2.6004 + 7.3152 + 4.1850
Wave 4 = 14.1006

Convert pips to discrete prices and subtract from last vertex D:
1.0014 – 0.00078 = 1.00062
1.0014 – 0.00081 = 1.00059
1.0014 – 0.00141 = 0.99999

CAVEAT

The various mathematic and statistical approaches illustrated in this book are, to our knowledge, new and therefore experimental. The rationale behind averaging multiple discrete forecasts (nine in the example in this chapter) is very logical, though, and should theoretically enhance the validity of the desired output. Nonetheless, the authors would like to emphasize that a thorough testing period and ample paper trading are advisable before traders incorporate any one method into their existing trading systems.