SUPERALGOS GOVERNANCE

p-pl Team Proof of Value, October 2022

Bitcoin Factory Dashboard / Reinforcement Learning Enhancement

p-pl
Superalgos | Algorithmic Trading

--

During the last month I was able to add the following main points to the Superalgos project:

  • 1) Reinforcement Learning Enhancement
  • 2) Bitcoin Factory Dashboard

1) Reinforcement Learning Enhancement

To give the AI the possibility to make profits in market situations with falling prices, short selling was implemented. Furthermore, the possibility for limits was introduced for all order types. In order to obtain a solid basis for the comparison of the different test results the new QuantStats Portfolio analytics was implemented.

Example QuantStats Portfolio analytics
Example HTML result page of QuantStats Portfolio analytics

After local tests I organized a public test round for the new features. This test round went very well — except for the fact that the minimum hardware requirement for a RL-TestClient is 32GB Ram after all.

2) Bitcoin Factory Dashboard

The Dashboard app was recently introduced to Superalgos. I build a new Component and a new vue for the Dashboard. With this “Bitcoin Factory Dashboard” its easier to monitor the Testserver and the Forecaster. All data is automatically pushed from SA to the Dashboard, even a reload of Browser isn’t needed!

Overview screen of Bitcoin Factory Dashboard

On the “Overview” screen the Bitcoin Factory Operator can easily see how Test-Server runs (unfinished Tasks, avg. Runtime …) and more important: The Forecaster can be monitored. Lets dive a bit into detail here. In the last row you can see the current “Next Action” — in this example its “3 / 25 / 97”. This means:

  • 3 -> sell long position
  • 25 -> use 25% of the base asset
  • 97 -> limit price is 0.97 multiplied with “previous close price”

These data can be directly consumed by a trading system (see next steps).

Server Tab of Bitcoin Factory Dashboard
Forecaster Tab of Bitcoin Factory Dashboard

Beside the “Overview” the Dashboard also shows detailed listing of Test-Cases for Server and Forecaster. I wrote a view Table component for this data, but this need to be improved especially for large number of test cases (it isn’t lucid enough).

Next Steps

Luis and I have scheduled an online meeting for 11/1/2022 to discuss how to connect the Bitcoin Factory with the social Trading / Signal distribution part of Superalgos. After that I will do my best to connect the two parts as fast as possible.

Get in Touch

Feel free to contact me here or in Superalgos Machine Learning Group https://t.me/superalgosmachinelearning

I would be very happy if more SA users who have either knowledge in the field of AI or interest in implementing a specific trading strategy within the AI get in touch with me.

--

--