Naming the reference you expect price to return to
Mean reversion needs a defined mean. Three choices are practical on gold: the midpoint of the established range, the volume weighted average price of the current session, or a medium moving average on the timeframe you trade.
They are not interchangeable. A range midpoint works when price is genuinely rotating between two known boundaries. A session average works intraday and resets each day. A moving average lags and will sit well behind price on a trending day, which is exactly when it should not be traded against.
Pick one, write it down, and keep it for a month so that the results mean something. Switching reference after every bad trade produces a record you cannot learn from.
The three conditions gold actually reverts in
The first is an established range on the four hour or daily chart, with at least two rejections at each boundary. Two touches per side is the minimum evidence that both edges are being defended.
The second is the Asian session. Participation is low, the dollar range is narrow, and price tends to rotate inside the overnight boundaries until London arrives. This is the most reliable window and also the one with the smallest payoff per trade.
The third is the hour or two after a scheduled release, once the initial spike has run. Overshoots on thin books frequently retrace a meaningful part of the move, though the first minutes are not tradeable in any controlled way.
A trigger rather than a feeling
Distance alone is not a reason. Price can be far from the mean and still accelerating. The trigger needs two parts: evidence of stretch, and evidence that the stretch has stopped working.
Stretch can be measured, and our note on measuring distance from the mean covers one way to put a number on it. The stop working part is price behaviour: a push into a level that closes back inside, a long wick with no continuation, a sequence of candles that no longer make progress.
In practice the cleanest version is a sweep of the range boundary followed by a close back inside it. The sweep collects the stops beyond the edge, and the close back inside says the move beyond had no support.
Where the stop has to sit
Behind the extreme of the sweep, with a little room for the wick, and not one cent closer. The entire premise is that the boundary holds, so the stop belongs where that premise fails.
This is where the method gets expensive if you cheat. A stop placed just beyond the entry candle will be collected by ordinary noise, because gold routinely extends two or three dollars past a level before turning. The honest stop is wider, which means the lot must be smaller.
Work the size from the stop: cash you accept losing, divided by stop distance in dollars, divided by one hundred ounces per lot. If that produces a position below your platform minimum, the trade is not available at your capital.
Targets and why the midpoint is enough
The realistic destination is the mean, not the opposite boundary. Reaching the mean is the thing the method claims will happen. Reaching the far side is a separate and more ambitious claim.
So the default target is the reference line, taken in full, with the trade closed and the screen left alone. On the occasions where price runs through the mean and continues to the opposite edge, you will have missed part of it, and that is an acceptable cost for a rule that ends trades at a defined point.
- Target one, the reference mean, which is the stated premise.
- Target two, the opposite boundary, only if the range is wide and the session has time left.
- Expiry, the London open, after which the range conditions may no longer hold.
The way this blows up
Trend days. A rate decision, a surprise inflation print or a sustained move in real yields produces a day where gold leaves the range and does not come back. Mean reversion entries on that day are a series of losses in the same direction, each one looking more attractive than the last.
The damage is structural, not accidental. The method wins small and often, so a handful of trend day losses can erase weeks of results. That asymmetry is the defining feature of reversion trading and no entry refinement removes it.
The only real defences are a hard daily loss limit, a rule against adding to a losing reversion position, and switching the method off when the range is broken.
Tools that tell you the conditions have gone
Range Break exists for this. It marks the current range and alerts when price has closed outside it, which is the moment to stop taking reversion entries. We label it a warning tool rather than an entry tool for a reason: across sixty six daily range breaks on gold, close to half were trading back inside the old range again within a week, and the swing that followed went the other way in roughly 65% of cases.
Read that honestly. It says a break is not a reliable breakout entry, and it also says the range you were trading has stopped being a range. Both can be true, and the second one is what matters for a reversion trader.
The underlying numbers are in our write up on whether gold range breaks work.
A routine that keeps the method honest
Before the session, decide whether the conditions exist at all. Mark the range boundaries and the reference, and note the time of any scheduled release. If a major print is due, the reversion playbook is off for the day.
During the session, take only entries that follow a sweep and a close back inside. Record every trade with the condition it was taken under, so that after thirty trades you can separate the Asian rotations from the post news fades.
At the end of the week, check the losses. If most of them came on days where the range had already broken, the entry logic is fine and the discipline is the problem, which is a much easier thing to fix.
FAQ
Does mean reversion work on gold?
In defined conditions, yes: inside an established range, during the quiet Asian hours, and after a news spike overshoots. The win frequency can be high and the average result small. It fails badly on trend days, when a sequence of entries against a sustained move produces losses large enough to remove several weeks of small gains.
What timeframe suits gold mean reversion?
The range itself should be identified on the four hour or daily chart so that the boundaries have been tested more than once. Entries are usually taken on the fifteen or five minute chart, where a sweep of the boundary and a close back inside can be seen clearly. Mixing the two levels of context is the point.
How do you know the range has broken for real?
Wait for a close outside the boundary on the timeframe you defined the range on, not a wick. Even then, treat it as a warning to stop taking reversion trades rather than as a breakout entry, since our own count of sixty six such breaks had close to half of them back inside the old boundaries within a week.
What is the biggest mistake in gold mean reversion trading?
Adding to a losing position because price is now even further from the mean. That converts a bounded loss into an unbounded one on exactly the day the method is wrong. The second biggest is placing the stop close to the entry rather than beyond the swept extreme, which turns ordinary noise into a stop out.
ⓘ See these ideas on real price: open the free XAUUSD live chart.