Gold Chrono Edge
- エキスパート
- バージョン: 2.20
- アクティベーション: 5
Gold Chrono Edge is an Expert Advisor for XAUUSD that trades specific times of the trading day. It is built on a simple question: does the way gold moves in one part of the session tell us anything about how it moves in a later part? Instead of assuming an answer, the EA measures it on your broker's own price history, keeps only the time windows where the relationship holds up under statistical testing, and trades those windows with fixed rules.
Core idea
Academic research on intraday momentum (Gao, Han, Li and Zhou, 2018) found that returns in the early part of a trading session can help predict returns later in the same session. Later studies on commodities showed that gold's predictive periods are different from those of stock markets, so they cannot be copied from equity research and must be measured directly.
Gold Chrono Edge applies this idea to XAUUSD. It divides the trading day into 30-minute windows on New York time, from 18:00 to 17:00, and studies how a price move in one window (the formation window) relates to the move in a later window (the holding window).
Main strategy
For every candidate pattern the EA checks:
- A formation window of 30, 60 or 120 minutes, starting at any half hour of the day.
- A holding window of 30 minutes, 1 hour, 2 hours or 4 hours that starts immediately after it.
- Two directions: momentum (trade in the direction of the formation move) and reversal (trade against it).
- Two strength levels: a normal formation move and a strong one, measured in volatility units.
This produces roughly 1,900 candidate patterns. Each one is simulated on one year of M5 history, using the same stop loss, take profit, breakeven and partial-close rules the EA uses in live trading, and after deducting spread and a slippage allowance.
When a selected pattern appears, the EA enters at the start of the holding window with a stop loss and take profit already set. Every trade has a maximum holding time and closes at the end of its window if neither level is reached. Positions are never held overnight or over the weekend.
The EA does not use martingale, grid, averaging or recovery techniques.
Statistics used to select trading windows
A large search will always find some patterns that look good purely by chance. Most of the EA's design is aimed at filtering those out:
- Shrunk t-statistic. The average result of each pattern is shrunk toward zero according to how many trades support it, then divided by its standard error. Patterns built on few trades are penalised. By default a pattern needs a shrunk t-statistic of at least 3.0.
- Out-of-sample check. History is split into a 70% in-sample part and a 30% out-of-sample part. A pattern must be profitable in both, and its out-of-sample profit factor must be at least 1.15.
- Stability across time. History is divided into four blocks. A pattern must be profitable in at least three of them, so a single good period cannot carry it.
- Decay check. A pattern must also be profitable over the most recent 40 sessions, which removes patterns that worked in the past but have stopped working.
- Minimum sample size. At least 60 historical trades per pattern.
- Walk-forward strictness. On live charts the EA runs a rolling walk-forward test that only ever uses past data, and automatically picks the strictness level (t of 2.5, 3.0 or 3.5) that has performed best out of sample.
- Luck check. The dashboard estimates how many of the tested patterns would pass purely by chance, and shows how credible the best selected pattern is compared with that estimate.
The EA recalibrates every 5 sessions on the latest year of data, so the selected windows adapt as gold's behaviour changes. If no pattern passes all checks, the EA does not trade.
News filter
An optional filter blocks new entries 15 minutes before and after high-impact USD news, using the MetaTrader 5 economic calendar. It works on live and demo accounts. We recommend enabling it for live trading.
The MetaTrader strategy tester cannot access the economic calendar, so backtests always run without this filter.
Designed for small accounts
Default settings are prepared for an account of about 100 USD: 0.01 lots, a profit target of 10 USD per trade and a daily profit target of 10 USD. If you trade a larger balance, send me a message and I will share recommended settings for your account size.
Dashboard
The on-chart panel shows:
- An edge map of the trading day, with the strength of each 30-minute window.
- The active trading windows, their historical statistics and today's status.
- Today's profit and loss, trade count and risk guard status.
- The EA's own track record on your account.
- Walk-forward results and the luck check.
Press R to recalibrate and H to hide or show the panel.
Requirements and recommendations
- Symbol: XAUUSD (or your broker's gold symbol).
- Chart: M5.
- Account type: hedging.
- At least one year of M5 history in the terminal. On first start the EA may need a few minutes to load it.
- Brokers with GMT+2/GMT+3 server time are detected automatically. For other brokers, set the clock mode to fixed and enter the GMT offset.
- A VPS is recommended so the EA can trade at its scheduled times.
Strategy tester notes
- Use "Every tick based on real ticks" or "1 minute OHLC" for accurate results.
- The tester needs at least one year of M5 data before the start date for calibration.
- In the tester only, if no window passes all checks, the EA trades the strongest candidate so its behaviour can be evaluated. This mode is marked as exploratory on the dashboard and is never used on live accounts.
- Backtest results are historical simulations and do not guarantee future performance.
Input parameters
Research engine
- Calibration lookback: days of history used to evaluate patterns (default 365).
- Recalibrate every N sessions: how often the trading windows are re-selected (default 5).
- In-sample fraction: share of history used for in-sample testing; the rest is out-of-sample (default 0.70).
- Edge types to search: momentum, reversal, or both (default both).
- Min formation move: minimum size of the formation move in volatility units (default 0.5).
- Strong-move tier: a second, stronger threshold searched as a separate pattern (default 1.0).
- Min historical trades per edge: minimum sample size (default 60).
- Base min shrunk t-statistic: significance level a pattern must reach (default 3.0).
- Auto-pick strictness: lets the walk-forward test choose between t-0.5, t and t+0.5 (default true).
- Profitable in N of 4 blocks: stability requirement (default 3).
- Recent sessions decay check: number of recent sessions that must be profitable (default 40).
- Expectancy shrinkage strength: how strongly small samples are penalised (default 30).
- Walk-forward in tester: runs the full walk-forward in the tester; slower (default false).
- Min out-of-sample profit factor: (default 1.15).
- Min expectancy per trade: in R, after costs (default 0.05).
- Max edges traded per session: (default 5).
- Block overlapping hold windows: prevents two selected windows from overlapping in time (default true).
- Tester exploratory mode: see strategy tester notes (default true).
- Walk-forward min training sessions: (default 120).
Exits
- Stop loss: in volatility units, scaled by the square root of the holding time (default 2.0).
- Take profit: same scaling (default 3.0).
- Use breakeven: (default true).
- Breakeven trigger: profit in R at which the stop moves to breakeven (default 0.8).
- Breakeven lock: profit in R locked in when breakeven triggers (default 0.05).
- Use partial close: (default true).
- Partial close trigger: in R (default 1.0).
- Partial close fraction: share of the position closed (default 0.5).
- Use trailing stop: not included in the calibration simulation (default false).
- Trailing start and distance: in R (defaults 1.2 and 0.8).
- Entry grace period: minutes after a window opens during which entry is still allowed (default 10).
Money profit target
- Money target mode: off, per trade, all open positions combined, or whole session (default per trade).
- Profit target in account currency: (default 10).
Daily money target
- Daily profit target in account currency: no new trades for the rest of the day once reached; 0 turns it off (default 10).
- Count only closed profit: if false, open profit also counts (default false).
- Close open trades when daily target is hit: (default true).
Costs
- Round-trip cost in points: 0 uses the current spread (default 0).
- Extra slippage/commission points: added to the cost used in calibration (default 10).
Position sizing
- Lot mode: fixed lot or risk percentage of equity (default fixed lot).
- Fixed lot: (default 0.01).
- Risk % per trade: used in risk mode (default 1.0).
- Max lot per trade: (default 5.0).
- Size stronger edges larger: scales the lot by the pattern's t-statistic divided by 3 (default true).
- Max edge size multiplier: (default 2.0).
Risk guards
- Max open positions: (default 3).
- Max trades per session: (default 8).
- Daily loss limit %: halts trading for the day and closes positions; 0 turns it off (default 0).
- Daily profit target %: (default 0, off).
- Profit lock activates at %: arms once the day's profit reaches this share of balance (default 2.0).
- Profit lock give-back %: closes all positions if the day gives back this share of its peak profit (default 40).
- Max spread: as a fraction of the volatility unit; entries wait if the spread is wider (default 0.25).
- Friday protection: no new trades from 12:00 New York time and all positions closed at 16:00 (default true).
News filter
- Block entries around USD high-impact news: (default false; enable for live trading).
- Minutes before and after news: (defaults 15 and 15).
Clock
- Live clock mode: auto detect, New York aligned, or fixed offset (default auto).
- Fixed broker GMT offset: used in fixed mode (default 0).
- Tester clock mode: (default New York aligned).
General
- Magic number: (default 770100).
- Show dashboard, position and scale: panel display options.
