TSX Swing Trading System

A fully automated virtual swing-trading system for Canadian (TSX) stocks. Screens the market daily, detects technical entry patterns, executes virtual buys and sells, and emails a report after every trade.

Data source: Yahoo Finance via yfinance. All transactions are virtual — no real brokerage connection.

---

How it works

Three scheduled services cooperate across the trading day:

Time (ET)	Service	What it does
4:30 PM	main.py	Checks market regime → rebuilds universe → runs screener → detects patterns → queues buy candidates
9:30 AM	virtual_buy.py	Reads the candidate queue, fetches live prices, sizes and executes virtual buys → sends trade email
3:30 PM	position_monitor.py	Evaluates open positions against exit rules using intraday data → executes virtual sells → sends trade email

All state — cash, positions, trades, signals, intents — lives in a single DuckDB database at data/trading.db.

---

Installation

Requirements: Python 3.10+, internet access (Yahoo Finance).

git clone <repo>
cd StockScanner

python -m venv .venv
source .venv/bin/activate      # Windows: .venv\Scripts\activate

pip install -r requirements.txt

---

First-time setup

1. Set your starting capital

Run once before anything else. Replace 50000 with your paper-trading capital:

python -c "from db import init_db, set_cash; init_db(); set_cash(50_000)"

Verify the database was created and the balance is correct:

python -c "
from db import init_db, get_cash, get_open_positions
init_db()
print('Cash     :', get_cash())
print('Positions:', get_open_positions())
"

2. Configure email (optional but recommended)

The system sends a trade notification email after every buy or sell. Gmail requires an App Password (not your regular password).

1. Enable 2-Step Verification at https://myaccount.google.com/security
2. Create an App Password → copy the 16-character code
3. Create a .env file in the repo root:

GMAIL_SENDER=you@gmail.com
GMAIL_APP_PASSWORD=xxxx xxxx xxxx xxxx
GMAIL_RECIPIENT=you@gmail.com

If .env is absent the system runs normally — it just skips emails.

3. Run main.py before the first trading day

Run this on the weekend before you want to start:

python main.py

This populates data/trading.db with the first batch of signals and queues any CONFIRMED setups as pending buy intents. virtual_buy.py will consume them Monday morning.

---

Daily operation

Once the system is running, the three services execute automatically on their schedule. To run them manually:

# End-of-day (after 4:30 PM ET)
python main.py

# Next morning (at/after 9:30 AM ET open)
python virtual_buy.py

# Pre-close (around 3:30 PM ET, market still open)
python position_monitor.py --mode pre-close

# Optional post-close informational run
python position_monitor.py --mode post-close

Dry-run mode

Both virtual_buy.py and position_monitor.py support --dry-run to preview activity without touching the database:

python virtual_buy.py --dry-run
python position_monitor.py --mode pre-close --dry-run  # not yet wired — informational only

---

Systemd deployment (Linux)

The system/ directory contains .service and .timer unit files for running the three services automatically on a Linux host.

First-time install

# Copy unit files to systemd
sudo cp system/*.service system/*.timer /etc/systemd/system/

# Reload systemd so it sees the new units
sudo systemctl daemon-reload

# Enable timers so they survive reboots
sudo systemctl enable stockscanner-main.timer
sudo systemctl enable stockscanner-buy.timer
sudo systemctl enable stockscanner-monitor.timer

# Start the timers now
sudo systemctl start stockscanner-main.timer
sudo systemctl start stockscanner-buy.timer
sudo systemctl start stockscanner-monitor.timer

After editing a .service or .timer file

# Copy updated files
sudo cp system/stockscanner-main.service /etc/systemd/system/
# (repeat for whichever files changed)

# Tell systemd to reload its configuration
sudo systemctl daemon-reload

# Restart the affected service if it is currently running
sudo systemctl restart stockscanner-main.service

Useful status commands

# Check whether timers are active and when they next fire
systemctl list-timers stockscanner-*

# View recent output for a service
journalctl -u stockscanner-main.service -n 50

# Check the current status of a service
systemctl status stockscanner-main.service

# Run a service immediately (outside its schedule)
sudo systemctl start stockscanner-main.service

Tickers URL

The main service is started with --tickers-url pointing to a remote file (one ticker per line). To change the URL, edit system/stockscanner-main.service and re-run the install steps above. The default URL is also set in config.py (CAN_TICKERS_URL) and is used when running services manually from the command line without --tickers-url.

---

Querying the database

Use DuckDB directly for ad-hoc queries:

import duckdb
conn = duckdb.connect("data/trading.db")

conn.execute("SELECT * FROM transactions ORDER BY trade_date").df()     # full ledger
conn.execute("SELECT * FROM positions").df()                             # open positions
conn.execute("SELECT * FROM trades ORDER BY sell_date").df()             # closed trades
conn.execute("SELECT * FROM account").df()                               # cash balance
conn.execute("SELECT * FROM intents WHERE intent_status = 'PENDING'").df()

conn.close()

Database tables

Table	Contents
account	Current cash balance
positions	Open virtual positions (ticker, entry date, price, shares)
trades	Permanent closed-trade log with full P&L
transactions	Unified ledger — every BUY and SELL in chronological order
signals	Pipeline signal state machine (rebuilt each run)
intents	Buy candidate queue with execution history

---

Screener

canadian_stock_screener.py scores each ticker in the TSX universe with a weighted factor stack:

• Weinstein Stage II alignment
• Relative Strength vs XIU.TO
• MACD momentum
• OBV slope (volume accumulation)
• ADX trend strength
• Volatility-adjusted momentum (VAM)
• 52-week high proximity / breakout pressure

Universe: loaded from CAN_TICKERS_URL in config.py (one ticker per line).

Output: out/screener_out/YYYYMMDD_HHMM.csv

Configure via CONFIG dict inside canadian_stock_screener.py: top_n, min_price, min_avg_volume, weights, lookback_days.

---

Entry pipeline

auto_pipeline.py reads screener CSVs, tracks tickers across days, and runs three pattern detectors:

1. VCP — Volatility Contraction Pattern
2. EMA pullback reclaim — EMA21 and EMA50 variants
3. Base breakout — tight range + volume confirmation

Signal state machine

FORMING  →  AT_PIVOT  →  CONFIRMED  →  ACTIVE
                     ↘              ↘  FAILED / EXPIRED

CONFIRMED setups are written to the intents table as PENDING buy candidates. virtual_buy.py processes them the following morning.

Market regime filter

Before generating new signals, main.py checks whether XIU.TO is above its 200-day SMA. In a bear regime, signal generation is skipped — no new positions are opened. Existing positions continue to be monitored.

---

Exit rules (position_monitor.py)

Each position is evaluated against four exit rules (defaults):

Rule	Trigger
Initial stop	Entry − 1.5 × ATR(14)
Chandelier trail	Highest high since entry − 2.5 × ATR(14)
Profit giveback	Max profit ≥ 6% and current profit drops ≥ 3% below peak
Time stop	≥ 20 trading days held with profit < 0%

---

Backtesting

Run a historical backtest over any date range:

# Single run
python run_backtest.py --start 2022-01-01 --end 2024-01-01

# Custom tickers (URL or file)
python run_backtest.py --tickers https://example.com/tickers.txt --start 2022-01-01 --end 2024-01-01

# Exit-param sweep: time_stop_days × stop_atr (4×4=16 combinations)
python run_backtest.py --start 2022-01-01 --end 2024-01-01 --sweep

# Walk-forward gap filter optimization (find best GAP_FILTER_PCT via sliding windows)
python run_backtest.py --start 2022-01-01 --end 2025-01-01 --walk-forward-gap
python run_backtest.py --start 2022-01-01 --end 2025-01-01 --walk-forward-gap --wf-in-days 84 --wf-out-days 21

python run_backtest.py --help

The backtest simulates the live gap filter when gap_filter_pct is set in BacktestConfig (default None = no filter, matching historical behaviour before May 2026).

Output files are written to out/:

• backtest_DATES_TIMESTAMP.html — HTML report with equity curve and trade log
• backtest_trades_TIMESTAMP.csv — full trade log
• backtest_equity_TIMESTAMP.csv — day-by-day equity curve
• backtest_wf_gap_DATES_TIMESTAMP.csv — walk-forward gap optimization results

---

Directory layout

.
├── main.py                  # end-of-day service (4:30 PM)
├── virtual_buy.py           # morning buy execution (9:30 AM)
├── position_monitor.py      # pre/post-close monitor (3:30 PM)
├── auto_pipeline.py         # pattern detection + signal state machine
├── canadian_stock_screener.py
├── run_backtest.py          # backtest CLI
├── db.py                    # DuckDB persistence layer
├── send_report.py           # Gmail email sender
├── config.py                # path constants + trading parameters
├── data/
│   └── trading.db           # all live state (auto-created on first run)
└── out/
    ├── screener_out/        # daily screener CSVs
    ├── alerts/              # daily alert CSVs + HTML report
    ├── logs/                # service run logs
    └── locks/               # process lock files

---

Running tests

pytest tests/ -v

# By phase gate (backtest refactor)
pytest -v -m phase1    # clock injection
pytest -v -m phase2    # MarketDataProvider
pytest -v -m phase3    # PortfolioState
pytest -v -m phase4    # BacktestRunner
pytest -v -m phase5    # HTML report
pytest -v -m phase6    # CLI entry point

# Specific files
pytest tests/test_db.py -v           # database layer
pytest tests/test_integration.py -v  # service integration

---

Notes / limitations

• yfinance depends on Yahoo Finance endpoints — intermittent rate limits or missing data can occur.
• The screener and benchmark (XIU.TO) are designed for TSX tickers.
• The pipeline caps tracked tickers (default 40) to limit API calls.
• All transactions are virtual — this system does not connect to any brokerage.

---

Disclaimer

For research and education only. Does not constitute financial advice. Trading involves risk of loss.

---

License

MIT