Reinforce your trades with trades

reinforce-trader is an open-source trading journal app that provides a comprehensive solution for trader to track, analyse and improve their trading strategies. The best way to learn anything is to learn from your own mistakes. reinforce-trader aims to provide a solution for traders so that they can maintain a systematic approach to "reinforce" their trading strategies.

reinforce-trader emphasizes the use of technical pattarns when traders make trading decisions. Traders often use technical patterns to find entry points and exit points. e.g Drawing lines of resistance levels and supporting levels, Drawing a triangle pattern, ...etc. Therefore, reinforce-traders allows you to upload an image to capture the moment of trade. I personally use TradingView to draw lines and you can easily download the chart image using Tradingview's "Take a Snapshot" function.

OS

Currently the application was only tested on MacOS.

How to start

1. Make a copy of .env.sample and rename it as .env

cp .env.sample .env

2. Start with docker

# `docker compose up -d --build` if you want to rebuild it
docker compose up -d

3. Open the app

You can start adding new trades in your trading journal: localhost:8050/add-trade

How to use

Add a new trade

You can add a new trade on localhost:8050/add-trade

View your trades

You then can view all your trades on localhost:8050/trades

Study the chart pattern

View the trade records

(Coming soon) View your entry and exit points on the candlestick

Roadmap

• Gather publicly available data as much as possible -> Datalake
• Extract information and knowledge from the publicly available data as much as possible -> research
• Build trading strategies using the information and knowledge from data -> trading system

Research pipeline

A flexible and modularized pipeline to build machine learning models for financial data

1. prepare the datalake template

python generate_datalake_template.py

2. download historical market data

import os

from reinforce_trader.research.datalake_client import DatalakeClient

datalake_dir = os.path.abspath(os.path.join(os.path.dirname(__file__), 'data'))
template_file_path = os.path.abspath(os.path.join(os.path.dirname(__file__), 'datalake_template.yaml'))
dl_client = DatalakeClient(datalake_dir, template_file_path)

dl_cleint.download_all()  # download all historical data of tickers from the template

3. run the classifier_v1 pipeline

python classifier_v1.py

TODO

• need sanity checking for the feature pipeline, e.g need to know the distribution of the labels in case of inbalanced target.
• need to analyse the feature (input, output, input vs output) to understand the effect of the featurizer.
• add regularizers to tackle overfitting
• add samplers to tackle data inbalance
• seperate data layer (including feature pipeline) from the model? (YES, you need 2 models from the same feature pipeline)
• refactor code for data layer and clustering

References

1. https://medium.com/@masteringllm/mastering-react-prompting-a-crucial-step-in-langchain-implementation-a-guided-example-for-agents-efdf1b756105
2. suggestions on fastapi application: https://www.youtube.com/watch?v=HTSK6eRwyGM
3. why do we need to detrend?: https://www.svds.com/avoiding-common-mistakes-with-time-series/
4. classifying sequences using unsupervised learning: https://datascience.columbia.edu/wp-content/uploads/2020/12/33_JPMorgan_Identifying-Trading-Opportunities.pdf
5. wiki of entropy: https://en.wikipedia.org/wiki/Entropy_(information_theory)