Module: Quant::Mixins::FisherTransform

Defined in:
lib/quant/mixins/fisher_transform.rb

Overview

Fisher Transforms • Price is not a Gaussian (Bell Curve) distribution, even though many technical analysis formulas

falsely assume that it is. Bell Curve tails are missing.
– If $10 stock were Gaussian, it could go up or down $20 – Standard deviation based indicators like Bollinger Bands
and zScore make the Gaussian assumption error

• TheFisher Transform converts almost any probability distribution in a Gaussian-like one

– Expands the distribution and creates tails

• The Inverse Fisher Transform converts almost any probability distribution into a square wave

– Compresses, removes low amplitude variations

Instance Method Summary collapse

Instance Method Details

#fisher_transform(value) ⇒ Object

The absolute value passed must be < 1.0



30
31
32
33
34
35
 
# File 'lib/quant/mixins/fisher_transform.rb', line 30

def fisher_transform(value)
  r = 0.5 * Math.log((1.0 + value) / (1.0 - value))
  r.nan? ? 0.0 : r
rescue Math::DomainError => e
  raise "value #{value}: #{(1 + value) / (1 - value)}, e: #{e}"
end

#ift(value, scale_factor = 1.0) ⇒ Object

inverse fisher transform www.mql5.com/en/articles/303



17
18
19
20
 
# File 'lib/quant/mixins/fisher_transform.rb', line 17

def ift(value, scale_factor = 1.0)
  r = (Math.exp(2.0 * scale_factor * value) - 1.0) / (Math.exp(2.0 * scale_factor * value) + 1.0)
  r.nan? ? 0.0 : r
end