TRADING SYSTEMS
Calculate Those Results With This Effective Strategy
A Nonparametric Performance Measure
by Guy Brys and Luc Van Hof
Gauge the true quality of your buy and sell signals
using these performance measures.
Many an indicator generates buy/sell signals,
but how effective are they? We can measure the quality of such signals
by embedding them into a fairly simple trading system. We will apply two
performance measures - the classical return on account (ROA) and a nonparametric
one (ACC) - to four indicators to measure the quality of the buy/sell signals,
followed by a walk-forward optimization and a discussion of the results.
IN THEORY
Generally speaking, a trading strategy is a collection of rules triggered
by one or more indicators. Because of this, the quality of an indicator
is often masked by the given strategy. Therefore, we will only vary the
definition of the underlying indicator. This makes it possible to make
an objective distinction between different indicators independent of external
factors.
We used historical tick-by-tick data from the exchange rate of the euro
(EUR) (the US dollar expressed in euros) in this study. Data prior to the
euro's introduction on January 1, 1999, was computed from the Deutschemark
expressed in US dollars (DEM) and other European currencies (see Figure
1).
Figure 1: DATA CALCULATION. Here's how the data for this study
was expressed.
Each tick represents a transaction with the exact time (in seconds) and
price at which the transaction took place. For computational reasons, we
recorded all data without decimals (the measurement unit is one pip). The
data was condensed to 15-minute time bars. Each entry in the new dataset
contains the date, time, open, high, low, and close prices. A trading day
is designated to be around 13 hours long (from 8:01 to 20:46), and in this
setup, no position will be entered outside this range, although exits can
take place.
A strategy's performance is highly influenced by the slippage and trade
size (number of contracts traded). The slippage must be subtracted when
you exit at a price different from the current price - that is, when you
exit on a stop-loss. Slippage may be estimated as 0.03 * dailyrange. Slippage
is not taken into account when exiting at the open or close.
...Continued in the January 2003 issue of Technical Analysis of STOCKS
& COMMODITIES
Excerpted from an article originally published in the January 2003
issue of Technical Analysis of STOCKS & COMMODITIES magazine. All rights
reserved. © Copyright 2002, Technical Analysis, Inc.
