TradePro

FOR LEARNING ONLY. NOT FOR REAL TRADING.

A paper trading web application built with Python and Flask. It pulls live market data, runs ML price predictions, and lets you execute simulated trades.

Live Demo

https://trading-strategy-nttz.onrender.com

The app runs on Render's free tier and may take around 30 seconds to wake up on the first visit.

Page	URL
Home	/
Trading Dashboard	/dashboard
Paper Trading	/trading
Performance	/performance/
Symbol Manager	/symbols/
System Status	/system

Stack

Layer	Technology
Backend	Python 3.11, Flask 2.3.3
Server	Gunicorn (2 workers)
Database	PostgreSQL (Render Managed)
Market Data	yfinance
ML Models	scikit-learn (Random Forest), statsmodels (ARIMA)
Charts	Plotly.js
Frontend	HTML, CSS, JavaScript
Hosting	Render

Features

Trading Dashboard Real-time candlestick charts with RSI, SMA20, and SMA50 overlays. Generates BUY / SELL / HOLD signals based on RSI thresholds and moving average crossovers.

Paper Trading Execute BUY and SELL orders at live market prices against a simulated account starting at 1,00,000 INR. Positions and transaction history persist in PostgreSQL.

ML Price Predictions 5-day ahead price forecast using an ARIMA + Random Forest ensemble. Results are cached per symbol for 1 hour.

Symbol Manager Search any stock across NSE, NYSE, NASDAQ, ETFs, and Forex. Watchlist is stored in PostgreSQL and persists across restarts.

Performance Dashboard Portfolio value over time, position breakdown, P&L per position, and full transaction history.

System Status Live CPU, memory, and disk usage. Service health indicators and log viewer.

ML Models

ARIMA (40% weight)

Order: ARIMA(5, 1, 0). Uses the last 5 days of closing prices, differences once to remove trend, and forecasts the next 5 trading days. Handles short-term momentum well.

Random Forest (60% weight)

100 decision trees trained on 10 engineered features: 1/3/5-day price returns, 5/10/20-day SMAs, 5/10-day rolling volatility, volume change, and 5-day volume SMA.

Ensemble

Final Prediction = 0.4 x ARIMA + 0.6 x Random Forest

Database Schema

All tables are created automatically on first boot using CREATE TABLE IF NOT EXISTS.

Table	Purpose
paper_account	Cash balance, positions, and transaction history (JSONB)
watchlist	Saved symbols
positions	Open position tracker
transactions	Full trade log
portfolio_history	Portfolio value over time

Project Structure

TP/
    web_app.py                      Main Flask app and paper trading routes
    db.py                           Shared PostgreSQL connection helper
    apps/
        symbol_manager.py           Watchlist CRUD and symbol search
        performance_dashboard.py    Portfolio analytics
    templates/                      Jinja2 HTML templates
        base.html
        home.html
        dashboard.html
        trading.html
        performance.html
        manage_symbols.html
        system.html
    static/
        css/
        js/
    src/
        models/                     ML model implementations
    config/                         JSON config files
    Dockerfile
    docker-compose.yml
    render.yaml
    requirements.txt

Local Setup

Requirements: Python 3.11+ and a PostgreSQL instance (local or remote).

git clone https://github.com/sreeteja2006/TP.git
cd TP

python -m venv venv
source venv/bin/activate

pip install -r requirements.txt

export DATABASE_URL="postgresql://postgres:yourpassword@localhost:5432/tradepro"

python web_app.py

The app will be available at http://localhost:5000. All database tables are created on startup.

Docker

docker-compose up -d

Render Deployment

The app runs as a Render web service backed by a Render managed PostgreSQL database.

1. Create a PostgreSQL database on Render (free plan).
2. Create a web service and connect your GitHub repository.
3. Add an environment variable DATABASE_URL using the Internal Connection String from the database.
4. Set the build command to pip install -r requirements.txt.
5. Set the start command to gunicorn --workers 2 --timeout 120 web_app:app.

The included render.yaml can auto-provision both services if you use Render's Blueprint feature.

Configuration

config/trading_config.json

{
  "max_positions": 8,
  "position_size_pct": 0.1,
  "stop_loss_pct": 0.05,
  "take_profit_pct": 0.15,
  "daily_loss_limit": 0.02,
  "max_trades_per_day": 10
}

Disclaimer

This is a paper trading simulator. No real money is involved. The signals and predictions are not financial advice. Market data is sourced from Yahoo Finance via the yfinance library.

Contact

GitHub: sreeteja2006 Email: gsreeteja25@gmail.com