The common perception is that all prices follow the standard bell-shaped curveknown as the Normal or Gaussian Probability Density Function. This assumes that68% of a range of samples will fall within one standard deviation around the mean.However, this perception has a fundamental flaw and could explain why a lot oftrading indicators do not yield the results many would hope for.
Let’s assume that
prices follow a square wave pattern and we adopt a trading system where we take aprice that crosses a moving average. What we will find is that by the time any pricemovement has been detected, it has already switched to the opposite value. As witha square wave, the price has only two values, so there is be a clear 50% probabilitythat the price will be only one of those values. So, despite any confusion around howto calculate a probability density, the concept is no more complicated thanunderstanding how likely it is that a certain price will be achieved.
Let’s now compare
a Gaussian, or Normal, Probability Density Function to that of a sine wave cycle.The first thing we notice is that with a sine wave, the majority of the data points willbe present at the extremes of the sine wave cycle and more closely resemble that ofa square wave. This therefore means that in terms of trading, unless you are able topredict the turning point of the cycle as per the method used in the Hilbert Sine WaveIndicator,
the probability is that you are more likely to hit one of these ‘extremes’,
which will make successful trading more difficult. So we can conclude that, as theGaussian PDF will have the majority of values around the mean and the sine wavehas the majority of its values at its extremes, the probability density function of a sinewave is not similar to that of the Gaussian probability density function. This brings usto the Fisher Transformation. The main concept of the Fisher Transform is that by
applying it to a sine wave it alters the probability density function of a wave so that itmore closely matches that of a Gaussian PDF.It does this by limiting any input values to within a range confined to no greater than -1 to +1.This means that input data that is close to the mean yields a parallel gain. However,input data at the extremes of the range yields exaggerated results mirroring thelargest deviations away from the mean.The effects of the Fisher Transform can be significant in trading terms. As the Fisher
Transform has standardised prices so that they remain within a -1 to +1 range, theexaggerated price movements become less common. In so doing, it becomes easierto identify those critical turning points.If we compare the Fisher Transform approach to the more traditional indicator
represented by the Moving Average Convergence / Divergence, or MACD, we noticea difference.The MACD shows smoother, less obvious turning signals meaning one cannot becertain as to exactly when the cycle is starting or leaving a new phase. This willinevitably create delays when deciding when to buy or sell.