🧠 Real-Time Cryptocurrency Trading Bot Using SMA Strategy on BTC Futures

This repository implements a modular framework for researching and deploying quantitative trading strategies in real-time environments. The design separates research, productionization, and runtime system architecture into clearly defined layers.

---

🏗️ Project Structure

Below is the updated architecture of the codebase:

This diagram illustrates the flow from strategy research to event-driven integration and validation.

---

⚡ How to Run

Gateway

Add the keys in the gateways/binance/vault/binance_keys file Run the run_binance.py

Engine

Add the telegram keys in engine/vault/telegram_keys Run the main.py

Run in Windows

Execute the run.bat

---

🧪 Research

Exploratory and experimental phase where strategies are initially tested in a vectorized notebook format using historical data.

Components

• research/datasets/: Raw historical price/time-series data
• research/: Strategy prototyping and analysis

---

🚀 Productionization (Model Integration)

Bridges research models into real-time production-ready event-driven strategies.

Steps

1. Implement a class that inherits from the base Strategy and wraps vectorized logic.
2. Use MarketDataClient and CandleAggregator to simulate or stream tick-level market data.
3. Compare vectorized vs real-time strategy signals via the ValidationTestingSuite ~ model_integration/tests.

---

⚙️ Runtime System Architecture

System design for executing trades in a live setting using real-time data and event streams.

Data Flow

Exchange
  ↓
MarketDataClient (e.g. CCXT, RemoteFeed)
  ↓
CandleAggregator → MidPriceCandle → Strategy
  ↓
→ StrategyManager → on_signal()
  ↓
→ TradeExecutor → RiskManager.validate_order()
  ↓
→ OrderExecutor → OrderStore
  ↓
→ AccountStateStore → MarginCalc → NotificationCenter

---

🧩 Key Components

Component	Description
MarketDataClient	Streams raw order book data from an exchange.
CandleAggregator	Constructs event-driven candles from raw ticks.
Strategy	Real-time strategy class (e.g., SMACrossoverInflectionStrategy).
StrategyManager	Manages strategy lifecycle and dispatches signals.
TradeExecutor	Listens to strategy signals and sends validated orders to executor.
RiskManager	Validates signals and orders against capital and risk rules.
OrderExecutor	Places or cancels orders through broker or mock client.
OrderStore	Tracks live and historical orders.
AccountStateStore	Updates current positions, PnL, and available balance.
MarginCalculator	Computes margin usage for active trades.
NotificationCenter	Pushes alerts to the notification platform.

---

🧪 Backtesting & Pine Parity Validation

Strategies are validated against TradingView Pine Script reference trades to confirm execution timing and signal logic are identical.

Execution Model

The backtest engine runs execution_timing = next_bar_open: orders queued on signal bar → filled at next bar's open (matches Pine's process_orders_on_close = false, calc_on_every_tick = false).

Current Parity (post deferred-fill refactor, March 2026)

Strategy	Parity	Status
CCI	100%	✅
ROC	100%	✅
ULTOSC	100%	✅
TRIX	99.9%	✅
BBAND	99.8%	✅
CMO	99.7%	✅
RSI	99.5%	✅
PPO	97.5%	✅
MOM	6.6%	❌ systematic 1-bar max-hold offset (Pine internals)

See engine/backtest/PYTHON_PINE_COMPARISON.md for detailed root cause analysis and planned fixes.

Running Validation

cd engine
python -m backtest.validate_runner --config backtest/configs/validate_pine_parity.json

---

✅ Development Workflow

1. Research a new idea in notebooks using historical datasets.
2. Productionize the idea by wrapping logic in a real-time Strategy class.
3. Use the Validation Suite to ensure fidelity between vectorized and live logic.
4. Plug the new strategy into the Runtime Engine for simulation or live deployment.

---