A Wednesday London-drift session with no scheduled US data. Gold had crossed session VWAP four times before 10:00 UTC, and the 15-minute chart showed overlapping candles with no run of consecutive higher highs — the regime filter passed before the stretch was even measured.
Worked example · winner
London drift reversion — short from 2.4 ATR above VWAP on fading volume
By 10:40 UTC, session VWAP sat at 4,481.30. Price pushed to 4,489.80, a $8.50 extension on a 15-minute ATR of $3.55 — 2.39 ATR — with the three push candles printing below the session's average volume, consistent with thin flow rather than genuine participation.
- Anchor (session VWAP)
- 4,481.30, confirmed by the 20 EMA at 4,481.90
- Extension
- $8.50 above the anchor, 2.39 × the 15m ATR of $3.55
- Regime check
- Four VWAP crosses by 10:00 UTC, overlapping candles all session — rotational, passed
- Volume check
- Push candles at 78% of the session average — declining, passed
- Rejection candle
- 10:55 UTC close at 4,486.90, extreme beyond the prior candle's extreme
- Entry
- 4,486.70 short, on the rejection candle close
- Stop
- 4,491.30 (extreme 4,489.80 + $1.50 buffer) — $4.60 risk per ounce
- Size
- £150 ÷ ($4.60 × $100) ≈ 0.32 lots
- Target
- 4,481.30, the anchor
- Result
- Filled 4,481.50 at 11:32 UTC, inside the four-candle time stop — +$5.20, +1.13R
Every item on the hub checklist passed in order: regime first, then stretch, then volume, then the rejection candle. Nothing about the entry required a discretionary read — the four prior VWAP crosses alone were enough to rule out a directional session.
The extension size, 2.39 ATR, sat inside the workable band described on the risk-management page, and the declining volume on the push candles was the clearest evidence that the excursion was thin flow rather than the start of a trend leg.
Price covered the full distance back to the anchor in thirty-seven minutes, comfortably inside the four-candle time stop, and the order filled two ticks beyond the anchor with no management required.
What it teaches: A clean reversion win looks unremarkable at the point of entry: a respected anchor, a normalised stretch, fading volume on the push and a rejection candle that closes back inside its own range. None of the four checks required a judgement call.
Worked example · loser
Every rule followed correctly — the anchor moved before price arrived
A Monday New York afternoon session, rotational all day with three VWAP crosses recorded by 18:30 UTC. At 19:05 UTC, with VWAP at 4,538.60, price pushed to 4,547.90 — a $9.30 extension, 2.51 ATR on a 15-minute ATR of $3.70 — on volume in line with the session average.
- Anchor (session VWAP)
- 4,538.60, confirmed by the 20 EMA at 4,538.10
- Extension
- $9.30 above the anchor, 2.51 × the 15m ATR of $3.70
- Regime check
- Three VWAP crosses, overlapping candles all session — passed
- Volume check
- Push candles at 96% of session average — flat, passed
- Rejection candle
- 19:20 UTC close at 4,544.60
- Entry
- 4,544.40 short
- Stop
- 4,549.40 (extreme 4,547.90 + $1.50) — $5.00 risk per ounce
- Size
- £150 ÷ ($5.00 × $100) = 0.30 lots
- Target
- 4,538.60, the anchor
- What happened
- A late unscheduled Treasury-auction headline lifted real yields at 19:45 UTC, and gold-wide selling pushed the VWAP anchor itself down to roughly 4,536.00 while price stalled at 4,545.50 rather than continuing towards the original target
- Result
- Time-stop breach at 20:05 UTC — under 30% progress after four candles — closed at 4,545.90, -1.06R
Every checklist item passed at entry: regime, stretch, volume and the rejection candle all met the stated standard, and the time stop was applied correctly rather than being held past the four-candle window in the hope of a late fill.
The failure was not in the entry — it was in the assumption baked into the strategy that the anchor stays roughly fixed for the duration of a 25 to 90 minute hold. On this occasion, the anchor itself moved during the trade because of new information, which is precisely the calendar risk the hub page's invalidation rules exist to catch, and which the entry checklist could not have flagged in advance because the headline was unscheduled.
The post-trade review confirmed there was no tier-one release inside the expected hold at entry, so the calendar check had genuinely passed. This trade is filed as a valid loss: the process was correct, and an unscheduled input moved the goalposts after the position was already open.
What it teaches: A reversion trade can pass every pre-entry check and still lose when the anchor itself relocates mid-trade. The time stop did its job by capping the loss near one unit rather than allowing the position to sit through a widening gap to the new anchor.
Worked example · loser
The trade that should never have been taken — stretch fired inside a trend day
A Friday NY overlap session following a stronger-than-expected NFP print at 13:30 UTC. Gold sold off in a single direction from 4,512.00 to 4,468.00 between 13:31 and 15:10 UTC, printing consecutive lower highs and lower lows on every 15-minute candle — a textbook trend day with the calendar-check condition already flagged as a no-trade window at entry time.
- Anchor (session VWAP)
- 4,491.40, itself falling continuously through the session
- Extension at trigger
- 4,468.00 was $23.40 below the falling anchor, 2.9 × the 15m ATR of $8.10 — well past the 2.0–2.5 trigger band
- Regime check
- Failed — six consecutive lower highs with lower lows, zero VWAP crosses since 13:35 UTC — a trend, not a rotation
- Volume check
- Failed — push candles at 190% of session average, expanding sharply — the signature of real flow, not thin excursion
- Trigger taken anyway
- 15:15 UTC — a shallow one-candle pause was read as a rejection candle and treated as valid
- Entry
- 4,470.30 long
- Stop
- 4,466.50 (recent low − $1.50 buffer... treated as extreme + buffer for a long) — $3.80 risk per ounce
- Size
- £150 ÷ ($3.80 × $100) ≈ 0.39 lots
- What happened
- Price continued falling to 4,452.00 by 16:00 UTC, stopping the position out and continuing well past the abandoned anchor
- Result
- -1.0R on the stop, filed as a rule violation rather than a valid reversion loss
The extension size alone made the setup look attractive — 2.9 ATR is a larger stretch than either of the first two trades — and that is exactly the trap the hub page warns about: the stretch condition is satisfied repeatedly all the way down a trend, and checking stretch before regime is how this strategy destroys an account.
Two separate checks should have stood the trade down before the rejection candle was even evaluated: zero VWAP crosses since the NFP print, and volume expanding to nearly double the session average on the down-candles. Either failure alone is sufficient under the hub checklist to void the setup.
The post-trade review recorded the reasoning at the time as "the stretch is huge, it has to snap back soon," which is the exact size-of-move fallacy the strategy exists to filter out — a large extension on a trend day is not a bigger reversion opportunity, it is confirmation that the move is not a reversion candidate at all.
What it teaches: A trend day satisfies the stretch trigger more convincingly than a rotational session does, which is precisely why regime is checked first, not stretch. This loss belongs in the violation column, not beside the two trades above it.