RSI Trading Engine

Automated quantitative trading system for QQQ using RSI mean reversion with multi-timeframe confirmation.

---

Executive Overview

A production-style automated trading engine that executes long and short positions on QQQ based on RSI(14) oversold/overbought signals. The system uses a layered, multi-timeframe approach: 45-minute bars trigger entries, while 1-hour and 4-hour confirmations adjust position sizing. ATR-based trailing stops, breakeven rules, and risk gates (daily loss limit, max trades, cooldown after loss) manage downside. Designed to run autonomously via macOS launchd during market hours.

Key characteristics: One position at a time, RSI-only signal logic, configurable risk parameters, dry-run and paper-trading support, structured logging, and preflight checks before execution.

---

Architecture

┌─────────────────────────────────────────────────────────────────────────┐
│                              MAIN LOOP                                   │
│  (main.py) → preflight → engine.run_cycle() every 15s                    │
└─────────────────────────────────────────────────────────────────────────┘
                                      │
         ┌────────────────────────────┼────────────────────────────┐
         ▼                            ▼                            ▼
┌─────────────────┐        ┌─────────────────┐        ┌─────────────────┐
│    PREFLIGHT    │        │  TRADING ENGINE │        │     EXECUTION   │
│  Market hours   │        │  Position mgmt  │        │  Alpaca orders  │
│  DRY_RUN check  │        │  Entries/Exits  │        │  Position sync  │
│  Mode vs URL    │        │  Trailing stop  │        │  Close position │
└─────────────────┘        └─────────────────┘        └─────────────────┘
         │                            │                            │
         │              ┌─────────────┼─────────────┐              │
         │              ▼             ▼             ▼              │
         │     ┌──────────┐  ┌──────────┐  ┌──────────┐           │
         │     │  SIGNALS │  │  SIZING  │  │   RISK   │           │
         │     │ RSI long │  │Confidence│  │ Daily PnL│           │
         │     │ RSI short│  │ Risk pct │  │ Max trades│          │
         │     │ Exits    │  │ Shares   │  │ Cooldown │           │
         │     └──────────┘  └──────────┘  └──────────┘           │
         │              │             │             │              │
         └──────────────┴─────────────┴─────────────┴──────────────┘
                                       │
                        ┌──────────────┼──────────────┐
                        ▼              ▼              ▼
                 ┌──────────┐   ┌──────────┐   ┌──────────┐
                 │   DATA   │   │INDICATORS│   │  CONFIG  │
                 │ Bars API │   │ RSI(14)  │   │ Thresholds│
                 │ Retries  │   │ ATR(14)  │   │ Risk params│
                 │ Validate │   │ NaN safe │   │ Market hrs│
                 └──────────┘   └──────────┘   └──────────┘

Flow: Each cycle, the engine checks exits (trailing stop, RSI-based) then entries. Entry requires 45-min RSI oversold (long) or overbought (short); 1H/4H RSI adds confidence tiers for position sizing. Position state is the single source of truth, synced from the broker on startup and after operations.

---

Design Goals

Goal	Implementation
Single source of truth	Position state synced from broker; one open position at a time
Fail-safe defaults	DRY_RUN=True, market-hours only, preflight blocks live trading on config mismatch
Modularity	Signals, sizing, risk, execution, data, and indicators are separate modules with clear boundaries
Observability	Structured logs per cycle (status, indicators, block reasons, decision summary)
Robustness	Bar fetch retries, NaN/inf handling in indicators, validation before entry
Reversibility	No strategy changes in “safe improvements”; config-driven thresholds

---

Project Structure

Module	Responsibility
main.py	Entry point, market-open wait, preflight, main loop, market-close shutdown
engine.py	TradingEngine and Position: cycle orchestration, entry/exit logic, trailing stop, breakeven
signals.py	check_long_entry, check_short_entry, check_long_exit, check_short_exit (RSI-based)
sizing.py	calculate_confidence_long/short, calculate_risk_percent, calculate_position_size
risk.py	RiskManager: daily PnL, max trades, cooldown after loss, daily reset (market TZ)
execution.py	get_current_position, submit_buy_order, submit_sell_order, close_position
data.py	get_bars_for_timeframe (retries), validate_bars_for_indicators
indicators.py	get_latest_rsi, get_latest_atr (Wilder RSI, ATR)
preflight.py	Market hours, DRY_RUN safety, mode vs URL consistency
config.py	Strategy and risk constants (RSI thresholds, ATR mults, risk %, market hours)
alpaca_client.py	Alpaca REST client for bars, account, positions, orders

---

Strategy Summary

• Instrument: QQQ
• Signal: RSI(14) on 45-minute bars; Wilder smoothing
• Long entry: 45-min RSI ≤ 30 (oversold)
• Short entry: 45-min RSI ≥ 70 (overbought)
• Long exit: RSI ≥ 70 or trailing stop
• Short exit: RSI ≤ 30 or trailing stop
• Position sizing: Confidence 1–3 from 1H/4H confirmation; risk = base + confidence tier, capped
• Stops: Initial 2× ATR, trail 1.5× ATR, breakeven at 1× ATR
• Risk: Max daily loss %, max trades/day, 60-min cooldown after loss

---

Setup

1. Install dependencies

   pip install -r requirements.txt

2. Environment

   Create .env with:

   • ALPACA_API_KEY – Alpaca API key
   • ALPACA_SECRET_KEY – Alpaca secret key
   • ALPACA_BASE_URL – https://paper-api.alpaca.markets (paper) or https://api.alpaca.markets (live)
   • ALPACA_DATA_FEED – sip or iex
   • MODE – paper or live

3. Strategy config (optional)
   Edit src/config.py for thresholds, risk %, and market hours.

---

Running

python -m src.main

The engine polls every 15 seconds, checks entries/exits, manages positions with trailing stops, enforces risk controls, and logs to logs/trading/.

DRY_RUN: Default DRY_RUN=True; no orders are sent. Set DRY_RUN=False for live trading.

Scheduling: Use the macOS launchd plist or cron examples in docs/SCHEDULER_SETUP.md for automated runs at market open.

---

Testing

python -m unittest discover -s tests -v
python -m src.test_preflight

---

Logs

Path	Purpose
logs/trading/trading_YYYYMMDD.log	Daily trading log
logs/trading/bot_scheduler.log	Scheduler output
logs/trading/launchd_*.log	launchd stdout/stderr

---

License

Unlicensed. Use at your own risk.