GenStrat

A quantitative strategy generation & backtesting system. With several data API provided.

Introduction

Simple Quantitative strategies such as SMA, EWM, BBand, RSI, etc. are always the very first step for beginners to get into the Quantitative world. However, it is not easy to find a platform that can provide a simple way to backtest these ideas and stratgies. GenStrat is designed for

1. Someone who have ideas but unfamiliar of how to implement strategies & backtest systems.
2. Someone who want to generate lots of startegies for ML / RL.

Project Structure

Installation

git clone

git clone https://github.com/AbnerTeng/GenStrat.git

cd GenStrat

Install through uv (recommended)

uv init && uv venv

source .venv/bin/activate

uv pip install -r requirements.txt

Install through pip

python -m venv .venv

source .venv/bin/activate

pip install -r requirements.txt

Install through docker (tbd)

Usage

1. Download dataset (optional)

Modify the config file config/stock_id.yaml to get the stock id you want to download.

python -m src.data.get_state

2. Cutomize your configuration in config/gen_strat.yaml

• Brief explanation of configuration

- DataAPIYahoo: Yahoo Finance API kwargs
- DataAPITAIFEX: Taiwan Future Exchange API kwargs
- DataLocal: If you have local data, or you don't have proxy to access any data API, the example data of SP500 ETF is in `tick_data`

- task:
    - syn_spsl (synthetic stop loss / profit experiment)
    - hpo (hyperparameter optimization among single strategy)

- trials
    - number of trials for hyperparameter optimization

- Class:
    strat: The strategy class you want to backtest
    params: The hyperparameters you want to optimize (please refer to `doc/strat_naming.md`)

Settings:
    - initial_cap: Initial capital
    - trans_cost: Transaction cost
    - stop_loss: Stop loss threshold
    - stop_profit: Stop profit threshold

1. Run single strategy backtesting with below command

uv shell
uv run python -m src.rolling_main --data_source <data_source> --plot

# Example
uv run python -m src.rolling_main --data_source yahoo --plot

(Alternative) Run multiple strategies backtesting with below command

chmod +x script/syn_traj.sh

bash scirpt/syn_traj.sh

Note that the whole project is still under development, everything is executable but not fully tested & optimized.