Edge Log · Wave 01
Fade the range sweep, or follow the break?
No validated edgeNo candidate qualified
The idea.
We took the previous H1 or H4 range and asked one thing: after costs, does fading a sweep back inside pay, or does following two closes outside?
One mechanical previous-range rule (CRT-style): confirmed M15 signals using completed H1/H4 ranges, with stop and target fixed 1R and 2R from the signal close. Signals are admitted from 08:00 to before 15:30 New York time; entry is at the next available open.
A simplified illustration of Wave 1's H1 FADE and ACCEPT rules with no EMA context filter (NONE), using simulated 15-minute candles—not market data. The simulation clock is illustrative, not historical New York/DST timing. Setup 1 is the first fade whose sweep reaches at least half an ATR beyond the range, so the sweep is easy to see; setup 2 is the first follow-through setup; the rest follow in time order, whatever they did. Outcomes are gross R before costs; this sequence is not a performance sample.
Read as text
- The previous hour's range: high 100.23, low 99.47.
- Swept below, closed back inside: a fade long.
- Entry at the next open 99.55; stop 99.32 (1R); target 100.00 (2R).
- Stopped out at 99.32: −1.00R before costs.
The test.
Gold, Nasdaq-100 and S&P 500 CFDs · Dukascopy bid prices, 15-minute candles, January 2018 to September 2026 · Assumed round-trip costs: gold 0.40, Nasdaq-100 2.0, S&P 500 0.60 points; stress test at 2×.
The result.
| Rule | Gold | Nasdaq-100 | S&P 500 |
|---|---|---|---|
| H1 · Fade | −0.207 | −0.146 | −0.163 |
| H1 · Fade + trend | −0.210 | −0.108 | −0.157 |
| H1 · Follow | −0.202 | −0.052 | −0.078 |
| H1 · Follow + trend | −0.163 | −0.045 | −0.089 |
| H4 · Fade | −0.236 | −0.132 | −0.170 |
| H4 · Fade + trend | −0.369 | −0.040 | −0.208 |
| H4 · Follow | −0.162 | −0.103 | −0.101 |
| H4 · Follow + trend | −0.094 | −0.050 | −0.086 |
| Rule | Gold | Nasdaq-100 | S&P 500 |
|---|---|---|---|
| H1 · Fade | −0.138 | −0.055 | −0.118 |
| H1 · Fade + trend | −0.113 | −0.086 | −0.072 |
| H1 · Follow | −0.111 | −0.056 | −0.076 |
| H1 · Follow + trend | −0.066 | −0.018 | −0.137 |
| H4 · Fade | −0.156 | +0.048 | −0.035 |
| H4 · Fade + trend | −0.018 | −0.083 | +0.003 |
| H4 · Follow | −0.100 | −0.051 | −0.134 |
| H4 · Follow + trend | −0.105 | +0.032 | −0.187 |
| Rule | Gold | Nasdaq-100 | S&P 500 |
|---|---|---|---|
| H1 · Fade | −0.041 | −0.085 | −0.070 |
| H1 · Fade + trend | +0.002 | −0.083 | −0.176 |
| H1 · Follow | −0.087 | −0.061 | −0.082 |
| H1 · Follow + trend | −0.091 | −0.046 | −0.048 |
| H4 · Fade | −0.058 | −0.074 | −0.034 |
| H4 · Fade + trend | +0.114 | −0.164 | −0.252 |
| H4 · Follow | −0.047 | −0.129 | −0.058 |
| H4 · Follow + trend | −0.014 | −0.096 | −0.041 |
Each cell shows one primary rule on one market in one period. Green means positive mean R per trade at the selected cost assumption; red means negative. n counts closed trades. The sixteen ATR stop-floor variants are not shown here. Mean R per trade.
Dot: base-cost mean R per trade. Bar: descriptive 95% block CI. Ring: mean at double costs. A positive lower bound is only one of the requirements for retrospective follow-up qualification; it does not establish a validated edge.
A recent gold pocket, with its limits.
Sum of base-cost R by period, not account return; periods differ in length. This pocket was selected descriptively after seeing the results. For 2025–Sep 2026, its mean was +0.114 R per trade at base costs and +0.063 at double costs. Its descriptive 95% block CI was [−0.127, +0.405] R per trade; matched random-direction p = 0.136.
Gold H4 fade + trend, Jan–Sep 2026 · 29 trades
Jan–Sep 2026 only: the same 29 trades move from +0.021 to −0.012 mean R per trade when assumed costs double. The full 2025–Sep 2026 window remains positive at double costs.
Gold H4 fade + trend, 2025–Sep 2026
2025–Sep 2026 at base costs: removing the three best trades leaves 62 of 65 trades. Mean R per trade falls from +0.114 to +0.023 and remains positive.
What it taught us.
All eight primary rules lost after assumed costs when pooled across markets in each period. Gold H4 FADE + TREND was positive in 2025–Sep 2026 but negative in earlier periods; its Jan–Sep 2026 mean turned negative at double costs.
Evidence and limits
- Retrospective: every date here had been studied before, so this is not an untouched out-of-sample test.
- Costs are assumptions, not measured broker quotes. No swap or financing.
- Pooled means weight each closed trade equally across markets, not each market equally. They are not funded-portfolio returns. Trades from different rules can overlap and are not independent observations.
- R is the planned signal-to-stop risk, not an account return.
- This tests one mechanical rule set. It does not show that every discretionary CRT method loses.
- The rule film uses simulated candles: an illustration, not a replay of market data.
- The descriptive 95% block CIs use 2,000 moving-block replicates with blocks of 20 observed trading dates. They do not adjust for the full prior research search.
- In 2025–Sep 2026, none of the 24 primary rule-by-market cells beat the matched random-direction benchmark at the nominal 5% threshold. This event-level counterfactual does not enforce portfolio ownership and is not a tradable benchmark.
- In 2025–Sep 2026, every paired daily ACCEPT-minus-FADE 95% block CI included zero. The study did not establish a reliable advantage for continuation over reversal.
Provenance
- Research journal record
RVD-W1-RESULT-IMPORT research/results/study.json· sha256d12d5c5ff9a2…research/results/audit.json· sha256366a9a36fc56…- Claim review record
RVD-W1-WEB-EDGE-LOG-NOTE-001