README.md Post Reddit Show HN Topics GitHub À coller directement dans README.md à la racine du repo. Remplace les screenshots par tes propres captures. Les badges sont optionnels.

Copier

Sextant — Event-Driven Backtesting Engine

A local, Python-based backtesting application with a Streamlit interface. Test trading strategies on historical data without depending on third-party platforms.

Features

Strict bar-by-bar event loop — MarketEvent → SignalEvent → OrderEvent → FillEvent. Deterministic: identical dataset + config always produces identical results. Full JSON audit log of every event.

Risk management

• 3 execution modes: Netting, Netting Delay, Hedge (long/short coexist)
• Stop loss, take profit, basket SL/TP (group positions under a shared stop)
• Position sizing as % of current capital

Performance metrics — Total return, CAGR, Sharpe, Sortino, Calmar, max drawdown (amount + duration), win rate, profit factor, avg win/loss, exit reason breakdown (stop / TP / signal), total commissions.

14 built-in indicators — SMA, EMA, RSI, ATR, Bollinger Bands, Stoch%K, Momentum, ROC, Highest High, Lowest Low, VWAP, RAW. Extensible via register_indicator().

No-code strategy builder — Visual IF/THEN condition builder with AND logic, crosses_above/crosses_below operators, and real-time preview of the latest value for any selected series.

Multi-source databank

• CSV import (TradingView export format)
• Yahoo Finance — indices, equities, FX, crypto, VIX, full history from 1990
• FRED / ALFRED — economic time series with vintage support and client-side transforms (YoY%, MoM%, absolute changes)
• Derived series (ratios, custom %)

Quick Start

# 1. Clone
git clone https://github.com/raphaub-hub/sextant.git
cd sextant

# 2. Install (Windows)
install.bat

# 3. Run
streamlit run app.py

What this is NOT

• No live trading / broker connection
• No parameter optimization (grid search, walk-forward) — yet
• No multi-timeframe (daily only for now)
• No options / derivatives
• Simplified slippage: execution at bar close
• No cloud, no SaaS, no REST API — this is a local desktop tool
• No portfolio optimization (Markowitz etc.)
• No tick / intraday data

Stack

Python 3.10+, Streamlit, Pandas, NumPy, Plotly, yfinance, PyArrow (Parquet), openpyxl.

Abstract interfaces: AbstractDataHandler, AbstractPortfolio, AbstractRiskManager, AbstractExecutionHandler, AbstractProvider.

Use case

Personal use / research. For anyone who wants to test strategy ideas on historical data without depending on a third-party tool.

---

Built with Claude Code