Neural Crossover System

1
Neural Crossover System

• By Kyle Serra

2
General Overview

• Trading Systems consist of three main things
• Entry Criteria
• Stop placement
• Exit Criteria

3
Entry Criteria

• The entry Criteria is the setup which alerts the
  trader to buy/sell a specific stock
• There are many types of entries
• The following is an example of how This System
  Enters trades

4
Entry Criteria
5
Stop Definition

• The stop is the maximum amount of risk the trader
  is taking on any given trade
• Can also be thought of the point where the trader
  will get out of the trade when X amount of money
  has been lost

6
Exit Criteria

• The exit is the point where the trade is gotten
  out of.
• The exit can depend on anything, but is usually
  either a time stop or some sort of profit target
  that has been achieved

7

• Currently this system has a timed exit point, it
  will always exit after two bars have elapsed.
  This exit is not very effective.

8
Neuro-Lab

• Neuro-Lab allows the user to create a customer
  indicator using the criteria that the user wants
• This information is then trained based on data
  selected by the user
• The user can create the network topology by hand,
  can manually set Momentum rate, training rate,
  hidden layers, input and output layer neurons

9

• My goal was to create a custom indicator that
  would allow me to predict future exits to a
  certain degree using Neuro-Lab. These results
  would then be compared to the old method of a
  timed stop to see which was more profitable.
  Multiple data sets were used to avoid curve
  fitting measures.

10
Network Topology of Exit NN

• Blue Dots Input Layer
• Red Dots Hidden layer
• Green Dot Output Layer

11
(No Transcript)
12
Output Script
13
Training
14
Performance EvaluationTraining Set
15
Performance EvaluationValidation Set
16
Creating the Indicator

• The indicator created by the neural net showed a
  correlation that when readings were high in the
  indicator the market was likely to go up and vice
  versa.
• This information was used to set a threshold for
  exit criteria, when the indicator reached 15 or
  below for long positions, the position was then
  executed, there is also a timed to stop so that
  after 6 bars if the indicator has not hit 15 or
  lower and has not been stopped out the position
  is automatically exited.

17
ResultsOld System, Tested on Dow Index from
1/1/2001 12/31/2001
18
Results New System, Tested on Dow Components
Index from 1/1/2001 12/31/2001
19
(No Transcript)