THE FORWARD TEST · GOLD FILE · TRADE #28 · ENTERED BY THE FROZEN RULE
Gold · crowd long · Mar 2024 never over its bar · no turn confirmed · the crowd was right
fade short · graded at the 13-week horizon · on the record for good
On Mar 5, 2024 the gold crowd sat at the very top of its own three-year range: net long past 99% of it. The frozen rule faded that at the Mar 11, 2024 open, blind. The crowd was right. The fade never reached its 4.7% bar, no turn was confirmed, and the 13-week close was -7.1%: fading this extreme cost 7.1% before costs. The worst point was 11.6% under water, and it was the clock that closed the window, not a stop.
WHY THIS TRADE EXISTS
No one chose it. The rule enters every extreme it sees, at the next session’s open, blind, on purpose. This file answers one question: what was the extreme worth, measured against every comparable extreme in gold’s own record? The published answer is “not much”. Most cases do not turn. That is exactly why the file records what else was present when this market did turn.
Why no stop, no exit? Because this is a measurement, not a managed position. The question is what the extreme was worth under one frozen rule, every time: a stop would smuggle discretion and survivorship back into the record. The adverse sit-through is published for exactly that reason, and the clock, not a choice, closes the window at W13.
gold’s record: 15 of 31 comparable extremes turned inside the window · median wait 8.5 wks · this one: never crossed
THE FILE · NOTES AS IT RUNS
JUN 10 · WINDOW CLOSED
no confirmed turn in 13 weeks · the crowd was right this time. on the record it stays.
JUN 10 · AFTER THE W13 GRADE
under its bar: -7.1% vs +4.7% needed · 0% of the still-under cases on file turned later.
MAY 6 · AFTER THE W8 GRADE
under its bar: -6.2% vs +3.7% needed · 19% of the still-under cases on file turned later.
APR 8 · AFTER THE W4 GRADE
under its bar: -7.5% vs +2.6% needed · 32% of the still-under cases on file turned later.
MAR 25 · AFTER THE W2 GRADE
under its bar: +0.3% vs +1.9% needed · 41% of the still-under cases on file turned later.
MAR 18 · AFTER THE W1 GRADE
under its bar: +0.9% vs +1.3% needed · 41% of the still-under cases on file turned later.
MAR 11 · WK 0 · ENTRY
entered at the open. the extreme alone is worthless; this file is everything else that was present.
● THE READ · PLAIN TEXT, GATE BY GATE
AS OFThe window closes under the bar: -7.1% against the +4.7% needed. The crowd was right this time. That result stays on the record, like every other one.
written by the numbers · the read re-writes itself as each Friday grade lands · click a gate or scrub the chart below · a measurement, not a recommendation
THE DIRECTION▼ SHORT · FADING THE LONG CROWD
The crowd is long; this file bets against it, short from the entry arrow above the bars. Trend-following, the other classic COT read, would ride with the crowd (long here): same tape, mirror sign · where this fade reads −7.1%, riding the crowd reads +7.1%. We pre-registered follow-the-crowd rules across the whole record too: no edge either. the tables →
THE TRADE ON THE TAPE · PRICE + WEEKLY POSITIONING · one numbered entry: this one · gate ticks mark the window
all entries on the full chart →middle pane: who holds what (% of open interest) · bottom: how crowded (red lines = extreme zone) · drag to pan, scroll to zoom · weekly bars · positioning drawn as steps · front-month proxy
THE PATH VS THE RECORD · TWO CORRIDORS · green = where past turns ran · red = where ran-ons drifted · dots = every case at the cursor gate
AT GATE
W13 of 13
hover the chart to scrub
THIS TRADE
-7.1%
its bar here: +4.7%
TURNED BY HERE
15
of 31 on the record
OF THOSE STILL UNDER
0% turned later
a read on late turns, not a promise
GRADED GATES · FORWARD ODDS
latest gate on top · click a gate to move the cursor · forward odds = of past extremes over their bar at that gate, the share still over at W13 · what would mark the turn →
CONDITIONS AT ENTRY · TRADE #28
3 LEAN TURN · 0 CLOSE · 2 RAN-ONOn 3 of 5 conditions, this trade looked like the ones that turned.
Each condition is compared with gold’s own past: what it read when the crowd turned, and what it read when the crowd ran on. Whichever it sits closer to wins the row. A map for your read, never a rule.
◀ looked like a run-ontoo close to calllooked like a turn ▶
Top-4 players' share
share of the long side held by the four biggest traders
Speculator headcount
how many speculator accounts crowd the extreme
VIX (market-wide volatility)
the stock-market fear gauge on entry day
Dollar direction
dollar rising or falling into entry
Crowd’s first cut
did the crowd start cutting within 4 weeks
percentile rows = this market’s own record (0–100) · yes/no rows = how common the state was in each cohort · ◀ / ▶ points to the cohort this trade sat closer to at entry · measured at entry, stated without judgement
a test written down in advance: extremes that passed this check aged +0.50R better in 2006–2018 and +0.22R in 2019→ than those that failed it (R = one unit of risk) · the gap holds inside each market, same direction in both periods, modest and not statistically significant in the later one · it sorts cases, it does not time entries · the table →
trade #28 of 31 on gold · entry at the first session open after the friday release · fade the crowd, no filters, no discretion · graded against 0.75× own volatility, √time scaled, within 13 weeks · continuous front-month file, close-to-close, roll gaps included, no costs/slippage · a turn is only awarded by the weekly pipeline · out-of-sample era · win or lose, this file stays. methodology → · what we tested and buried →
This dossier is the free showcase.
Every numbered trade on every market has a file like this · Pro opens them all, plus every live one · cancel anytime