This repo contains backtesting scripts for various models(mainly LSTM) using different type of datasets to predict bitcoin price. Upto 98.7% accuracy, but let me tell you it’s not enough to generate profits on a regular basis ;)
Last updated Jun 23, 2026
21
Stars
3
Forks
0
Issues
0
Stars/day
Attention Score
11
Language breakdown
Python 100.0%
▸ Files
click to expand
README
CryptoAITrader
This contains bactesting scripts for BTC price prediction.
Different approaches used here:
- Retraining: Retraining the model w.r.t. the new dataset(moving window).
- NN2: A second neural net that predicts the accuracy of the first neural net. It was surprising to see the effictiveness of the second NN, as it reduced bad trades by 50-60%.
- Pattern training: Identifying similar patterns accross different price ranges.
- Long/Short prediction: Instead of predicting the price this model focuses on predicting market direction(bullish/bearish).
Usage
Simplycd into the respective folder and run .py scripts
