Custom Search

In the current popular trading systems terminology, a stochastic refers to a measure of the placement of the current price within a recent range. If the current price is near the top part of the recent range, the stochastic will be high; if it is near the low end of the range, the stochastic will be low. Specifically, the stochastic can be defined as follows:

S N = C t В - L / H-L

Ct = current closing price

H = high price during past N days

L = low price during past N days

N = number of days used to calculate the stochastic value

S = the stochastic's value on the Nth day

This definition uses daily price data, but any time interval can be used.

Testing stochastic signals

Stochastics seem to have become relatively popular in recent years, although much of the evidence regarding the forecasting ability of this measure has been anecdotal. To rigorously test the usefulness of the stochastic, the following simple system was devised:

1. Cover long and go short if the short-term moving average of the stochastic moves below the long-term

moving average of the stochastic for a specified number of consecutive days.

2. Cover short and go long if the short-term moving average of the stochastic moves above the long-term moving average of the stochastic for a specified number of consecutive days.

The specific system included four parameters:

N1 = number of days used to calculate the stochastic

N2 = number of days in short-term moving average of stochastic

N3 = number of days in long-term moving average of stochastics

N4 = number of consecutive days for which crossover must hold to provide a signal.

The system was tested for a range of parameter sets. An analogous system was tested using weekly data.

The systems were tested using a hypothetical $1.3 million portfolio—a portfolio size which would only be relevant for pools, funds, or extremely wealthy individuals. The reason for this fund size was to allow for adjustments in position sizes to compensate for extremely wide differences in volatility between markets.

For example, the tested portfolio contains 25 contracts of corn, but only four contracts of coffee. This type of approach provides a much better test of a system than simply assuming that all contracts are traded in a one-unit size. However, based on our experience, the general conclusions at the end of the article would also apply to similar portfolios trading single contracts for all markets.

The simulation program charged $150 per trade for transaction costs. Although this figure may seem high, it is important to realize that transaction costs are much greater than commission costs. On average, there is slippage on both entry and exit of trades. For example, if one assumes only a one point slippage factor in bonds on entry and exit, slippage alone would equal $62.50. Admittedly, the $150 per-trade transaction cost may be a bit conservative, but there is a strong argument for using a figure at least equal to $100 per trade. In any case, any system which cannot make money at $150 per trade should be viewed with some skepticism.