VWAP Reclaim Lab
Features at a Glance | Folder Guide | Key Tech
Introduction
An intraday options backtesting framework - built around VWAP stretch-and-reclaim signals. Built with clean boundaries for data, signals, option selection, exits, and reporting.
Includes simulations for contract scaling and real-world friction (latency, slippage, fees).
Fully cacheable, sweep-ready, and designed for reproducible backtests and Bayesian optimization.
(With an obnoxious amount of prints for data integrity and auditability.)
Features at a Glance
Signals: Detects a stretch away from VWAP beyond a threshold - and a partial reclaim towards it. The stretch-and-reclaim movement must occur in a specific window.
Option Selection: Sweepable parameters for ITM/ATM/OTM and strike depth. Buys calls and puts based on stretch direction. Same-day expiry enforced by default.
Exits: Take-profit, stop-loss, max hold duration, end-of-day, and emergency failsafe exits.
Trading Friction: Latency, slippage, commissions/fees.
Bayesian Optimizer: optimize/smart.py utilizes Optuna for parameter sweeps over a discrete search space. Built for concurrency and parallelism.
Key Parameters (Subset):
| Parameter | Example | Description |
|---|---|---|
stretch_threshold |
0.003 |
Min percent move away from VWAP. |
reclaim_threshold |
0.0021 |
Percent move back toward VWAP (after initial stretch). |
cooldown_period_seconds |
120 |
Throttle multiple signals per stretch event. |
option_selection_mode |
itm |
itm/otm/atm selection. |
strikes_depth |
1 |
1 = closest depth to ATM within the chosen mode. |
take_profit_percent |
80 |
Exit on gain. |
stop_loss_percent |
-25 |
Exit on loss. |
max_trade_duration_seconds |
600 |
Time-based exit. |
Full details in strategy/README.md.
Folder Guide
| Path | What’s Inside |
|---|---|
strategy/ |
Core logic: parameters, data loading, signals, option selection, exits, backtest, reporting. See strategy/README.md. |
optimize/ |
Parameter sweeps with Optuna and Postgres-backed storage. Prunes low-trade trials, deduplicates param combos, and reports best trials. See optimize/README.md. |
quickstart/ |
One-command run with bundled synthetic data. See quickstart/README.md. |
tests/ |
Basic correctness and regression tests against synthetic data. |
NOTE: Synthetic data is used in quickstart/ and tests/ to adhere to Polygon's terms.
Key Tech
| Tech | Description/Utilization |
|---|---|
| Python | Install packages via requirements.txt. |
| Polygon.io API | Stocks and Options Developer plans (or higher) are required to pull historical data at one second aggregates. |
| Optuna | Bayesian optimization engine used by optimize/smart.py to sweep parameters. |
| PostgreSQL | Persists Optuna studies and deduplicates parameter combos; configure PG_* env vars. Useful for parallel and/or multi-run parameter sweeps. |
| Docker | Reproducible environment and fast builds for virtual/ephemeral compute. |