How I reached 100% accuracy in my Bitcoin Price Prediction Algorithm

Using Machine Learning in Cryptocurrency Algo Trading

Just over a month ago, I built a crypto trading algorithm that makes predictions on the price of Bitcoin based on machine learning, and trades based on its predictions. I admit that I approached this project in a very experimental / no-expectations sort of way, so imagine my surprise when I realized that this could actually work.

First of all, here’s how the strategy works and why I went for this approach. This is the 4th article in a series of articles that explore this particular trading strategy, you don’t have to read the previous articles, but if you want to get the whole picture, I suggest you start here.

The algorithm has two parts, the prediction layer and the trading layer. The prediction layer works by Taking the Open price for the current 1h Bitcoin candle as an input, and based on this input it tries to predict the High price of the same candle. The prediction layer was trained on hourly Bitcoin data so it make sense to work with the same timeframe when inputting data for a prediction.

(If you’re unfamiliar with the concept a candlestick or Kline shows the following data: Open price, Close Price, High, Low and Volume for its timeframe.)