We needed a methodology for triggering
trades, even though we were and remain convinced that this is not the most
critical part of designing the strategy. That distinction belongs to the risk
reduction and capital retention aspects of it. One reason for this is that
price movement can very often be counter-intuitive, even for an algorithm.
Example: The RBNZ (New Zealand central bank) made an announcement in July 2014
to the effect that it was raising interest rates, which had been expected. This
should have been a positive for the Kiwi (New Zealand dollar), and in fact it had
been strengthening in anticipation. However, at the very moment of the
announcement the rate for the currency against the US dollar plunged instead.
We found out soon enough that the authorities in NZ wanted a lower exchange
rate, and had gone into the markets to sell $NZ in order to make this happen.
The best algorithm in the world, barring the possession of insider knowledge on
the part of its developers, would not be able to envisage this type of circumstance.
There are many other examples of a similar nature in the Forex trading business.
Notwithstanding the above, we needed
a trigger for Mandelbrot. In our opinion a good place to start looking for this
was in the area of moving averages of price over time. A great number of the
popular Technical Analysis indicators already make use of these. Examples
include MACD (Moving Average Convergence Divergence), Moving Average Envelope
indicator, Bollinger Bands, Ichimoko Cloud, and Moving Average Crossover
indicators. There are others, and even those that do not have MA as their
primary base use it anyway. Stochastics, for example, uses it to smooth the
results of the bar highs and lows it is based on, in order to make its triggers
less volatile.
However, there is one problem with
all of these: they are lagging indicators. This means that, very often, the
information they give comes after the time at which it would have been
beneficial to place a trade.
Acceleration:
the rate of change of a rate of change
The objective of trading Forex at our
level of investment is to be able to synchronise our buying and selling
decisions with those of the larger institutions that actually move the market
(Trillions of dollars are traded each day and moving the market is something we
cannot do at this point in our development). Exchange rates wander about all
the time, going up a little and then down a little. We are looking for trends,
or relatively large movements in one or other direction.
It seemed to make sense, therefore,
to try to distinguish between the rate changes during those quiet periods and
their own rate-of-change when buying or selling picks up, which might have a
better probability of signalling the start of a trend. In relation to
determinate non-linear functions mathematicians call this the second derivative.
For engineers it is the definition of acceleration in bodies that move relative
to a reference point. It was developed by Isaac Newton and is notated as
follows:
Where E (in our case) is the exchange
rate and t is the time over which it changes. The letters d in each case
indicate an infinitesimal change (a change in the limit).
The only issue here is that price
action in Forex is not the smooth, continuous process that lends itself to
Newtonian principles of mathematics and mechanics. While it might approximate
these, there are also elements involved that manifest as sheer apparent randomness.
One way to resolve this problem might be, we thought, to overlay Newtonian
principles with chaos theory, which had been developed by Edward N. Lorenz for
use in weather forecasting. We adopted the acceleration idea and to deal with
the chaos strove all times to make sure that probability was well and truly on
our side. This meant designing the trades so that there was always a positive
expectation: a typical loss had to be considerably less in value than a typical
win, and wins had to, ideally, outnumber losses. Each trading day must be
meticulously analysed to determine if the trading software parameters, or even
the software itself, should be adjusted or modified. Included here is the
aspect of timing. For optimum position sizing we were guided by the Kelly
Criterion. This was developed by J. L. Kelly in the 1950s for use in
probability theory, and has been applied largely to gambling scenarios.
All of the above might seem like a
lot of work, and indeed it is. All those adverts you see promising the ability
to learn to trade Forex in a short time and by only devoting a few hours per
day to it, in between doing something else, are making a false claim.
Benoit Mandelbrot was a scientist who
did much work on the theory of fractals. His ideas have been used too, in order
to move our trading from higher level time frames (4-hour and daily) where
trades were few and far between, to lower level time frames (hourly and down to
10 minutes) so that we could have a larger number of trades. We have named the
Mandelbrot routine in his honour.
As developed, the Mandelbrot strategy
will not give a signal to buy or sell while the exchange rate is moving at a
normal, uneventful pace. It is only when it accelerates from the casual rate
that a trade will be considered. It uses a series of exponential moving averages
to anchor what is regarded as normal behaviour. In practice, price can move up
or down as much as it likes when going at a pedestrian pace, and it is only
when its velocity moves outside of the norm that action will be triggered. A
trade is likely to be terminated when the activity diminishes (although not in
all cases), a condition that is classified as reversion to the mean. In the
meantime the algorithm has a built in ability to reposition a dominant stop
loss (there is normally more than one – recall our commitment to risk
mitigation) so that it moves relatively closer to price over time as profits
build. This gives it the best chance to get the maximum benefit from each
trade. Mandelbrot also takes off some profit as a profitable trade progresses,
and moves the trade into a break-even position as soon as possible, in
accordance with our capital retention policies.
To be continued.
Coming soon:
Defining the trading strategy
· Give the trade enough space to become
successful but ensure that if this is ultimately not to be the case, the loss
will be the smallest that it is possible for it to be
· Risk control
* Eliminate the danger
of taking on too much risk
* Lock in some profit as
soon as it is worthwhile to do so, and move the main stop loss order to break
even.
* Control of Stop Loss
levels at the start of the trade and as it progresses
Coding and debugging the strategy
· Some programmable platforms are
better than others, but this is still a new industry outside of the major banks
and hedge funds
Optimising the parameters
· Historical research on the price
action of different currency pairs
· Simulated trading with historical
data, and real-time trading
· The place of Fundamental and
Technical Analysis
No comments:
Post a Comment