Goldbacktest

Rule-based intraday backtesting engine for XAUUSD using Dukascopy tick data, with reproducible reports and an interactive HTML dashboard.

This repository is designed as a clean public portfolio project that demonstrates:

• Market data ingestion and caching (raw .bi5 + parsed parquet).
• Timezone-aware strategy simulation (including DST edge handling).
• Deterministic strategy rules and trade lifecycle simulation.
• Performance analytics and report generation.
• Artifact generation (charts + dashboard) and test coverage.

• Instrument: XAUUSD
• Data source: Dukascopy tick feed
• Strategy style: first-break intraday wick-break
• Default period: 2025-02-01 to 2026-01-31
• Runtime language: Python 3.11+

• Reference window: 08:00-09:00 Europe/Brussels.
• Build session high/low from bid prices in that window.
• Scan ticks from 09:00:00 onward:
  • bid > session_high -> open SHORT
  • bid < session_low -> open LONG
• First trigger only, max one trade/day.
• TP/SL:
  • TP = 10 points
  • SL = 17 points
  • point_size = 0.01
• Fill and exits:
  • Long entry at ask, short entry at bid.
  • Long TP/SL checks on bid, short TP/SL checks on ask.
• If no TP/SL hit, force-close at final available tick of Brussels day.
• Cost model:
  • Dynamic spread from tick bid/ask.
  • Fixed slippage configurable via --slippage.
  • commission_model=none currently supported.

goldbacktest/
  analytics.py        # metrics + monthly breakdown
  build_dashboard.py  # standalone HTML dashboard builder
  config.py           # validated runtime config model
  dukascopy.py        # hourly download/cache/parse pipeline
  plot_trades.py      # trade-level OHLC chart renderer
  reporting.py        # markdown/json/png outputs
  run.py              # main CLI entrypoint
  strategy.py         # day simulation + full backtest loop
  types.py            # typed domain records (Trade, DayIssue)
tests/
  test_strategy.py
  test_time_and_metrics.py
requirements.txt

py -3.11 -m venv .venv
.venv\Scripts\Activate.ps1
py -3.11 -m pip install --upgrade pip
py -3.11 -m pip install -r requirements.txt

py -3.11 -m goldbacktest.run `
  --symbol XAUUSD `
  --start 2025-02-01 `
  --end 2026-01-31 `
  --timezone Europe/Brussels `
  --tp-points 10 `
  --sl-points 17 `
  --point-size 0.01 `
  --slippage 0 `
  --position-size 1.0 `
  --output-dir outputs

py -3.11 -m goldbacktest.plot_trades `
  --trades-file outputs/trades.csv `
  --output-dir outputs/trade_charts `
  --data-dir data `
  --symbol XAUUSD `
  --timezone Europe/Brussels `
  --price-scale 1000

py -3.11 -m goldbacktest.build_dashboard `
  --outputs-dir outputs `
  --charts-dir outputs/trade_charts `
  --out-file outputs/dashboard.html

Open outputs/dashboard.html in your browser.

Backtest run writes:

• outputs/trades.csv
• outputs/monthly_breakdown.csv
• outputs/metrics.json
• outputs/equity_curve.png
• outputs/data_issues.csv
• outputs/report.md

Chart rendering writes:

• outputs/trade_charts/*.png
• outputs/trade_charts/manifest.csv

Dashboard generation writes:

• outputs/dashboard.html

Main runner:

• --download-only fetches and parses data cache without simulating trades.
• --rebuild-cache ignores existing parsed cache and rebuilds.
• --max-workers controls concurrent hourly download/parse execution.
• --data-dir and --output-dir control local storage paths.

py -3.11 -m pytest -q

Current tests cover:

• Strategy trigger/exit behavior.
• DST hour-count correctness for Brussels calendar days.
• Metric reconciliation sanity checks.

• Uses zoneinfo for timezone correctness.
• Handles 23-hour and 25-hour local days around DST transitions.
• Caches parsed hourly ticks to parquet for rerun speed.
• Emits issue diagnostics (data_issues.csv) for missing/unreadable hours.

Generated artifacts and local caches are ignored by default:

• virtual environments
• local data cache
• generated outputs/charts
• IDE/cache files

This keeps the public repository reviewable while preserving full reproducibility through documented commands.

This project is for research and engineering demonstration. It is not investment advice.