AutraSense 2.0 - March 2021 Live Test Result

Deep learning used to manage selected Index CFD with absolute return


AutraSense 2.0 is the new release and breakthrough of AutraSense Machine Learning trading framework. Deep learning technology was used in the platform design, which enable neural network be able to identify complex market pattern with more reliable result. 

AutraSense 2.0 was put in a live test and forward test environment since 1st March 2021. Following trading target was selected:

  • US500

  • UK100

  • DE30

  • JP225

  • HK50

It is an absolute return investment model and be able to receive 17% in 4 weeks trading. 

High Level Design

Comparing AutraSense 1.0, a multi-dimension sensor was put between market data and neutral network. The ideal was inspired by the convolutional neural network (CNN) design, which is strong to recognize graphic objectives. 

Multi-dimension sensor provide a more reliable recognition market movement. The precision is able to reach between 0.55 to 0.65, which provides a little edge to help accumulate the profit in the long run. 

A fully connection neural network connect to the output data from multi-dimension sensor, which is similar to 1.0 version.  Web service was provided by Azure Machine Learning environment which enable trading program to collect the market date hourly and get prediction from neural network hourly. 

Trading program will make Buy/Sell/No Action decision for every participated CFD index at every hour. 

Currently, the neural network was re-training weekly. There is opportunity to increase the training frequency to daily or every hour, if it was really needed. 

4 Weeks Trading Result

After 4 weeks live/forward test, the portfolio is able to reach 17% result at the end, with 1% stop lose per every trade. 

It is impressive result comparing with other absolute return product in the market, but little bit disappointed as the portfolio reached to 30% at the high end, and going down due to poor performance on HK50 investment. 

Here is the details for this 4 weeks live/forward trading result. The portfolio value starts from 100.  

Next Step

I will do a re-balance of the trading index in the portfolio based on 4 weeks test result. HK50 will be removed and AU200 will be added. After another 4 weeks live/forward trade, I will also move it to a big size account, which enable me to leverage lots size adjustment logic with neutral network. 

If you are interest in the experiment i am doing, please send me note and we can discuss some Proof of Concept project opportunities.