Mean Reversion waits for gold to stretch unusually far from its own recent average, and then for price to actually turn back before it acts. It aims at the average as it stood at that moment, and it stands aside entirely when gold is trending hard.
What it does: Measures how far price has stretched from its reference average against how much gold has been moving lately. Waits for price to recover back through the level where the stretch began rather than guessing at a turn. Aims at the average as it stood when the stretch was identified, so a drifting average can never flatter a losing position.
When it works: Balanced, rangebound gold weeks with no dominant macro driver. Quiet hours, where gold often oscillates around a level. After a spike that overshot on thin liquidity and was immediately rejected.
When it struggles: Trending markets — the stretch simply keeps growing, which is why a strong hourly trend blocks entry outright. Breakouts out of long compression. High-impact news, when the average is irrelevant for hours.
What the trader controls: Risk per trade, as a percentage of the account or a cash amount. Daily loss limit — once hit, no new entries for the rest of the London trading day. Maximum open positions on gold at any one time. Sessions it may trade, and whether high-impact news blocks new entries. You can switch the model off at any time, and reducing or closing risk is never blocked.
The method behind Mean Reversion is documented under Gold mean reversion at https://goldhunts.com/gold-strategies/gold-mean-reversion. Gold Hunts publishes win rate and sample size for closed XAUUSD trades only — never profit or loss figures, and never a projection.
GHX-04 Mean Reversion v2.0 is designed for a specific gold market behaviour and records a plain-English reason for every decision. Coming to forward testing. It is not yet available for account activation. During its trading window, it assesses each completed 15-minute candle and never makes a decision from a candle that is still forming.
Its approach to gold: It first reads the hourly picture and stands aside completely when gold is running hard in one direction — a stretched trend is not an overreaction. It then measures how far price has stretched away from its reference average, judged against how widely gold has been trading recently, and requires that stretch to be genuinely unusual rather than merely large. It does not act on the stretch alone. It waits for a later completed 15-minute candle to recover back through the level where the stretch began, with real conviction in that candle rather than a single spike. The average it aims at is fixed at the moment the stretch is identified and never moves afterwards, so an average drifting toward a losing position cannot be mistaken for price reverting successfully. Every trade carries a stop beyond the extreme of the stretch, set before the order is sent, and the trade is only taken when enough room remains back to the fixed target to be worth that risk. It leaves early if price closes back beyond the original stretch, if the trade has made no progress after the first part of its life, or when its time limit or session cutoff arrives. If any condition is missing, it records a wait with the single reason that stopped it.
Its dedicated public ledger is at https://goldhunts.com/gold-models/ghx-04-mean-reversion/ledger. The record begins with the first eligible demonstration forward-test trade and will include every closed result, whether it wins or loses. Nothing is backfilled with simulations. Customer trades are never published. Results are shown in net R with gross R alongside, and sample size is always visible. Demonstration results do not represent live trading.
GoldHunts — the AI-powered gold trading operating system.