"CCI mean reversion on gold, London session: $10,000 → $46,519"
Source: a popular X day-trader, with a strategy-report screenshot · Tested October 2026 · Gold 1-minute bars, May 2024 – Oct 2025
Before costs, the rule makes almost exactly nothing, in every setting we tried. The claim needs half a point a trade. With nothing there before costs, no cheaper broker or bigger position can rescue it.
The claim vs. our replication
| The post (1 year, gold futures) | Our test (1.45 years, gold CFD) | |
|---|---|---|
| Trades a day | 5.4 | 19.4 |
| Win rate | 64.4% | 51.6% |
| Profit factor | 1.23 | 0.49 |
| Edge before costs, per trade | +0.52 pt (implied) | −0.02 pt |
| Result after costs, per trade | +$26 a contract | −0.42 pt |
| Sharpe ratio | not given | −15.1 |
The post names the indicator, the market, the session and the size, but not the timeframe, CCI settings, entry trigger or exit. We used the charting platform's defaults: CCI(20), levels ±100, 1-minute bars, entry when CCI crosses back inside a level, exit at zero. All positions closed by 08:00 New York.
Every setting sits on zero
Each dot is one of 15 settings (CCI length × level), placed by its edge per trade before costs. The claim needs +0.52 points. Our costs are 0.40.
Show all 15 settings
| CCI length | Level ±100 | ±150 | ±200 |
|---|---|---|---|
| CCI(10) | +0.014 | +0.004 | +0.012 |
| CCI(14) | −0.008 | +0.010 | −0.041 |
| CCI(20) | −0.018 | −0.041 | −0.020 |
| CCI(30) | −0.033 | −0.026 | −0.076 |
| CCI(40) | −0.013 | −0.004 | −0.074 |
No good hour either
Split by the New York hour the trade opened. The light bar is before costs, the red bar after. Every hour loses once costs are paid.
The test battery
Every study runs the same checks. A strategy has to pass all of them to go further.
-
FAIL
Beats random entries at the same timesRandom trades at the same New York hour, direction and holding time do as well. Ours ranks at the 27th percentile of 500 random draws.
-
FAIL
Beats a time-shifted copy of itselfSliding the same positions to other dates scores as well: 26th percentile of 300 shifts.
-
FAIL
Both halves of the data make moneyFirst half −0.40 pt a trade, second half −0.43.
-
FAIL
Survives higher costsAlready negative at normal costs. At 1.5× costs it is −0.62 pt a trade.
-
FAIL
Nearby settings agree0 of 15 settings are profitable after costs. The median Sharpe is −8.9.
-
FAIL
Beats buying and holding goldStrategy Sharpe −15.1 against +1.66 for simply holding gold over the same period.
-
FAIL
Survives the luck of trying many variantsDeflated Sharpe 0.000 over the 23 variants this study ran. The pass mark is 0.95.
So how did the post make $36k?
- One year of results, in a strong gold rally.
- One futures contract on $10,000 is about 40× leverage, so small swings look huge.
- An unpublished filter: the post takes 5.4 trades a day, the literal rule takes 19.
- The report's two drawdown figures come from different episodes. The deeper one took about five weeks to recover, not 17 days.
What would change our verdict
The exact rules: timeframe, settings, entry and exit. With those we would re-run the author's own year on gold futures 1-minute data. Until then, the published rule has nothing before costs.
Test details
- Data
- Gold CFD, 1-minute bars, 20 May 2024 – 31 Oct 2025, 377 sessions
- Market during the test
- Gold +66% (2,414 → 4,004)
- Costs
- 0.40 points round trip, measured at a prop-firm broker. Also checked at the post's own 0.257.
- Fills
- Signal on the bar close, fill at the next bar's open
- Held-back data
- Nov 2025 onward kept unseen for any future re-test
- Variants counted
- 23 in this study, all priced into the deflated Sharpe
Got a strategy that "keeps working"?
We'll run it through this exact battery and send you the report.
This is a replication of a public claim using our reading of its rules, not an accusation.