- Overview
- Why
bbstraderStands Out - Trusted by Traders Worldwide
- The
bbstraderEdge: Uniting C++ Speed with Python Flexibility - Key Modules
- What's New in 2.3.0
- Getting Started
- Installation
- CLI workflow
- Community & Support
- Professional Services
Imagine having the raw, blistering speed of C++ for your high-frequency trades, combined with Python's ecosystem for lightning-fast prototyping, advanced AI models, and seamless data analysis. That's bbstrader, not just a library, but a game-changing toolkit designed for quants, algo traders, and institutional pros who demand an edge in volatile markets. Whether you're scalping forex pairs, backtesting complex strategies, or copying trades across accounts in real-time, bbstrader empowers you to build, test, and deploy with unmatched efficiency.
Forget the frustrations of slow Python bottlenecks or MQL5's rigid sandbox. bbstrader bridges worlds: C++ for mission-critical performance and Python for intelligent orchestration. It's open-source, battle-tested across platforms, and ready to supercharge your trading arsenal.
In a crowded field of trading libraries, bbstrader is architected to solve the most challenging problems in algorithmic trading: performance, flexibility, and platform limitations.
- Blazing Speed with C++ Core: Compile your strategy logic in native C++ for deterministic, low-latency execution. Perfect for HFT, arbitrage, or compute-heavy models that Python alone can't handle.
- Python's Powerhouse Ecosystem: Leverage
NumPy,pandas,scikit-learn,TensorFlow, and more for research, ML-driven signals, and backtesting, all seamlessly integrated with your C++ core. - Institutional-Grade Architecture: From its event-driven backtester to its modular design,
bbstraderis built with the principles of professional trading systems in mind, providing a robust foundation for serious strategy development. In today's hyper-fast financial landscape, every microsecond counts.bbstraderisn't another lightweight wrapper, it's an institutional-grade powerhouse engineered to tackle real-world trading challenges head-on. - Break Free from MQL5 Limits: Ditch interpreted code and ecosystem constraints. Build multi-threaded, AI-infused strategies that execute orders in microseconds via MetaTrader 5 (MT5) integration.
Flexible Interface: CLI & GUI
bbstraderadapts to your workflow.- Automation Fanatics: Use the CLI for headless scripts, cron jobs, and server deployments.
- Visual Traders: Launch the Desktop GUI (currently for Copy Trading) to monitor your master and slave accounts, check replication status, and manage connections visually.
- Cross-Platform & Future-Proof: Works on Windows, macOS, Linux. (IBKR integration in development).
With thousands of downloads, bbstrader is trusted by traders worldwide. It's not just code, it's your ticket to profitable, scalable strategies.
bbstrader's hybrid design is its secret weapon. At the heart is a bidirectional C++/Python bridge via client module:
- C++ for Speed: Core classes like
MetaTraderClienthandle high-performance tasks. Inject Python handlers for MT5 interactions, enabling native-speed signal generation and risk checks. - Python for Smarts: Orchestrate everything with modules like
tradingandbtengine. - The Data Flow: The result is a clean, efficient, and powerful execution loop:
Python (Orchestration & Analysis) -> C++ (High-Speed Signal Generation) -> Python (MT5 Communication) -> C++ (Receives Market Data)
This setup crushes performance ceilings: run ML models in Python and execute trades in C++. The backtester is an event-driven simulator with a replayable, columnar data feed built for fidelity (faithful order state and accounting), paired with a fully vectorized research fast-path (vectorized_backtest(...)) that screens millions of bars per second for fast "does this have alpha?" hypothesis testing.
MetaTrader 5 is a world-class trading platform, but its native MQL5 language presents significant limitations for complex, high-frequency strategies:
- Performance Ceilings: As an interpreted language, MQL5 struggles with the computationally intensive logic required for advanced statistical models, machine learning, and rapid-fire order execution.
- Ecosystem Constraints: MQL5 lacks access to the vast, mature ecosystems of libraries for numerical computation, data science, and AI that C++ and Python offer.
- Architectural Rigidity: Implementing sophisticated, multi-threaded, or event-driven architectures in MQL5 is often a complex and error-prone endeavor.
bbstrader eradicates these barriers. By moving your core strategy logic to C++, you can unlock the full potential of your trading ideas, executing them with the microsecond-level precision demanded by institutional trading.
bbstrader is modular, with each component laser-focused.
- Purpose: Simulate strategies with historical data across multi-asset portfolios, with metrics like Sharpe Ratio, Drawdown, and CAGR and pluggable (commission, partial fills, latency, and per-bar swap/overnight funding).
- Features: Event queue for ticks/orders, a replayable columnar data feed (re-run the same data for parameter sweeps and walk-forward), and integration with models for signal generation. The engine runs in two modes behind one strategy API: an event-driven path for fidelity and a
vectorized research fast-pathfor high-throughput screening. - Example: Backtest a StockIndexSTBOTrading from the example strategies.
# Inside the examples/
from strategies import test_strategy
if __name__ == '__main__':
# Run backtesting for Stock Index Short Term Buy Only Strategy
test_strategy(strategy='sistbo')The engine ships a batteries-included research stack on top of the shared strategy API:
- Execution realism (opt-in, defaults unchanged): pluggable slippage (fixed-spread, percent, volatility, volume-participation), square-root market impact, commission models, partial fills, time-frontier (next-bar) fills, order-to-fill latency, and per-bar swap/overnight funding costs, so backtests survive the jump to live.
- Vectorized research fast-path:
vectorized_backtest(...)screens entry/exit signal arrays across the whole history at once for fast "does this have alpha?" hypothesis testing, alongside the high-fidelity event engine. - Built-in indicators & strategy templates: vectorized SMA/EMA/RSI/ATR/ Bollinger/MACD/z-score and ready-made trend / mean-reversion / breakout templates on the same API used for live trading.
- Optimization & validation: parallel
optimize(...)parameter sweeps and walk-forward (replayable columnar data +reset()), plus overfitting diagnostics, deflated/probabilistic Sharpe, PBO (CSCV), combinatorial purged CV. - Risk analytics: historical & parametric VaR/CVaR, Monte Carlo equity-curve confidence bands, volatility-regime detection, and factor/beta exposure.
- Multi-strategy & multi-timeframe: several strategies sharing one portfolio/clock, and on-the-fly higher-timeframe resampling (e.g. daily signals on a 1m feed, no look-ahead).
- Reproducibility: a cached data catalog (Parquet) and an
experiment store that persists params, metrics and equity curves for
leaderboard-style comparison. A
benchmarks/script backs the performance claims. - Broker abstraction: a venue-neutral
Brokerinterface (with an in-memoryPaperBroker) so strategies can target MT5 today and other brokers later.
- Purpose: High-speed MT5 integration. C++ MetaTraderClient mirrors MT5 API for orders, rates, and account management.
- Features: Bidirectional callbacks, error handling, real-time tick processing.
- Strategy Patterns: Two main patterns to build strategies:
This is the recommended pattern for latency-sensitive strategies, such as statistical arbitrage, market making, or any strategy where execution speed is a critical component of your edge. By compiling your core logic, you minimize interpretation overhead and gain direct control over memory and execution.
Use this pattern when:
- Your strategy involves complex mathematical calculations that are slow in Python.
- You need to react to market data in the shortest possible time.
- Your production environment demands deterministic, low-latency performance.
C++ Side (MovingAverageStrategy.cpp):
#include "bbstrader/metatrader.hpp"
#include <numeric>
#include <iostream>
class MovingAverageStrategy : public MT5::MetaTraderClient {
public:
using MetaTraderClient::MetaTraderClient;
void on_tick(const std::string& symbol) {
auto rates_opt = copy_rates_from_pos(symbol, 1, 0, 20);
if (!rates_opt || rates_opt->size() < 20) return;
const auto& rates = *rates_opt;
double sum = std::accumulate(rates.begin(), rates.end(), 0.0,
[](double a, const MT5::RateInfo& b) { return a + b.close; });
double sma = sum / rates.size();
double current_price = rates.back().close;
if (current_price > sma) {
std::cout << "Price is above SMA. Sending Buy Order for " << symbol << '\n';
MT5::TradeRequest request;
request.action = MT5::TradeAction::DEAL;
request.symbol = symbol;
request.volume = 0.1;
request.type = MT5::OrderType::BUY;
request.type_filling = MT5::OrderFilling::FOK;
request.type_time = MT5::OrderTime::GTC;
send_order(request);
}
}
};This C++ class would then be exposed to Python using pybind11.
// Inside bindings.cpp
#include <pybind11/pybind11.h>
#include "MovingAverageStrategy.hpp"
namespace py = pybind11;
PYBIND11_MODULE(my_strategies, m){
py::class_<MovingAverageStrategy, MT5::MetaTraderClient>(m, "MovingAverageStrategy")
.def(py::init<MT5::MetaTraderClient::Handlers>())
.def("on_tick", &MovingAverageStrategy::on_tick);
}Python Side (main.py):
from bbstrader.api import Mt5Handlers
import MetaTrader5 as mt5
import time
from my_strategies import MovingAverageStrategy
# 1. Instantiate the C++ strategy, injecting the Python MT5 handlers
strategy = MovingAverageStrategy(Mt5Handlers)
# 2. Main execution loop
if strategy.initialize():
while True:
strategy.on_tick("EURUSD")
time.sleep(1)This pattern is ideal for strategies that benefit from Python's rich ecosystem for data analysis, machine learning, or complex event orchestration, but still require high-performance access to market data and the trading API.
Use this pattern when:
- Your strategy relies heavily on Python libraries like
pandas,scikit-learn, ortensorflow. - Rapid prototyping and iteration are more important than absolute minimum latency.
- Your core logic is more about decision-making based on pre-processed data than it is about raw computation speed.
import MetaTrader5 as mt5
from bbstrader.api import Mt5Handlers
from bbstrader.api.client import MetaTraderClient
# 1. Inherit from the C++ MetaTraderClient in Python
class MyStrategyClient(MetaTraderClient):
def __init__(self, handlers):
super().__init__(handlers)
# 2. Instantiate your client
strategy = MyStrategyClient(Mt5Handlers)
# 3. Interact with the MT5 terminal via the C++ bridge
if strategy.initialize():
rates = strategy.copy_rates_from_pos("EURUSD", mt5.TIMEFRAME_M1, 0, 100)
print(f"Retrieved {len(rates)} rates via the C++ bridge.")The Account class provides a high-level, safe interface to live MT5 account data.
Context manager, automatically shuts down the connection on exit:
from bbstrader.metatrader import Account
with Account() as acc:
print(acc) # Account #12345 on MyBroker | Balance: 10000.0 USD
print(repr(acc)) # Account(number=12345, server='MyBroker', balance=10000.0)
rates = acc.get_currency_rates("EURUSD")refresh(), reload live account data mid-session without creating a new object:
acc.refresh() # re-fetches balance, equity, margin from the terminal
print(acc.balance)Symbol info cache, get_symbol_info() caches results to avoid repeated round-trips to the terminal. Call clear_symbol_cache() to invalidate when needed.
retry_on_disconnect decorator, wrap any MT5 call to auto-retry on connection drops:
from bbstrader.metatrader.utils import retry_on_disconnect
@retry_on_disconnect(max_retries=3, delay=1.0)
def fetch_tick(symbol):
return client.symbol_info_tick(symbol)- Purpose: Manages live sessions, coordinates signals from strategies, risk from models, and execution via metatrader.
- Features: Multi-account support, position hedging, trailing stops.
- Purpose: Build/test models like NLP sentiment, VaR/CVaR risk, optimization.
- Features: Currently Sentiment analysis, and Topic Modeling.
- Example: Sentiment-Based Entry:
from bbstrader.models import SentimenSentimentAnalyzer
model = SentimenSentimentAnalyzer() # Loads pre-trained NLP
score = model.analyze_sentiment("Fed hikes rates, markets soar!")
if score > 0.7: # Bullish? Buy!
print("Go long!")core: Utilities (data structs, logging).
config: Manages JSON configs in ~/.bbstrader/.
api: Handler injections for bridges.
Version 2.3.0 turns btengine into a full research-to-production backtesting stack: new modules for
risk analytics, parameter optimization, overfitting diagnostics, execution realism, and reproducible
experiments, plus a broker-neutral execution layer and a built-in indicator library.
| Module | What it adds |
|---|---|
analytics |
Risk analytics: VaR/CVaR, Monte Carlo simulation, volatility-regime detection, and factor/beta exposure |
optimize |
Parameter optimization with parallel sweeps and walk-forward evaluation |
overfitting |
Overfitting diagnostics: deflated/probabilistic Sharpe, PBO via CSCV, and combinatorial purged cross-validation |
friction |
Pluggable execution realism: slippage, market impact, commissions, partial fills, latency, and per-bar swap/funding |
catalog |
Cached, Parquet-backed data catalog (falls back to CSV when pyarrow is absent) |
vectorized |
Vectorized research fast-path that screens millions of bars per second for quick alpha hypothesis testing |
templates |
Ready-made strategy templates (trend, mean-reversion, breakout) |
timeframe |
On-the-fly higher-timeframe (HTF) resampling |
experiment |
Experiment/results store persisting parameters, metrics, and equity curves for reproducible runs |
| Area | Change |
|---|---|
btengine |
Multi-strategy support over a shared portfolio, multi-timeframe feeds, walk-forward replay, intrabar OHLC fills |
core/broker |
New broker-neutral execution abstraction with a PaperBroker implementation |
core/indicators |
Built-in vectorized indicators: SMA, EMA, RSI, ATR, Bollinger Bands, MACD, and z-score |
| Area | Change |
|---|---|
| C++ docs | New Doxygen documentation workflow with automatic GitHub Pages deployment |
| Benchmarks | Reproducible engine benchmark suite for tracking performance |
| Tests | Greatly expanded coverage across the new btengine modules and the broker abstraction |
- Python: Python 3.12+ is required.
- MetaTrader 5 (MT5): Required for live execution (Windows).
- MT5 Broker: Admirals, JustMarkets, FTMO.
bbstrader is designed for both Python and C++ developers. Follow the instructions that best suit your needs.
Get started in minutes using pip. We strongly recommend using a virtual environment.
# Create and activate a virtual environment
python -m venv venv
source venv/bin/activate # on Linux/macOS
venv\Scripts\activate # on Windows
# Install bbstrader (lean core: numpy/pandas/yfinance + the backtesting engine)
pip install bbstrader
# Add MetaTrader 5 live trading (Windows)
pip install "bbstrader[mt5]"
# Optional extras (install only what you need):
# nlp -> NLP/sentiment stack (nltk, spacy, textblob, vaderSentiment, sumy)
# social -> social-media feeds (tweepy, praw)
# viz -> extra plotting (plotly, seaborn)
# catalog -> Parquet-backed cached data catalog (pyarrow)
# all -> everything above
pip install "bbstrader[all]"Lean core: as of the latest release the base install no longer pulls the heavy NLP/social/viz stacks. If you use those modules, install the matching extra (the code raises an actionable error telling you which one).
To develop your own C++ strategies, you can use vcpkg to install the bbstrader library and its dependencies.
# If you don't have vcpkg, clone and bootstrap it
git clone https://github.com/microsoft/vcpkg
./vcpkg/bootstrap-vcpkg.sh or ./vcpkg/bootstrap-vcpkg.bat
# Install bbstrader
./vcpkg/vcpkg install bbstraderbbstrader shines via CLI, launch everything from one command!
| Action | Command |
|---|---|
| Run Backtest | python -m bbstrader --run backtest --strategy SMAStrategy --account MY_ACCOUNT --config backtest.json |
| Live Execution | python -m bbstrader --run execution --strategy KalmanFilter --account MY_ACCOUNT --config execution.json --parallel |
| Copy Trades | python -m bbstrader --run copier --source "S1" --destination "D1" |
| Get Help | python -m bbstrader --help |
Config Example (~/.bbstrader/execution/execution.json):
{
"SMAStrategy": {
"MY_MT5_ACCOUNT_1": {
"symbol_list": ["EURUSD", "GBPUSD"],
"trades_kwargs": { "magic": 12345, "comment": "SMA_Live" },
"short_window": 20,
"long_window": 50
}
}
}- Read the Docs: Full API reference and tutorials.
- GitHub Issues: Report bugs or request features.
- LinkedIn: Connect with the creator.
If you need a custom trading strategy, a proprietary risk model, advanced data pipelines, or a dedicated copy trading server setup, professional services are available.
Contact the Developer:
📧 bertin@bbs-trading.com
If you find this project useful and would like to support its continued development, you can contribute here:
Disclaimer: Trading involves significant risk. bbstrader provides the tools, but you provide the strategy. Test thoroughly on demo accounts before deploying real capital.