First-Hour Fade
validated price-action edge · fade the first retest of the first-hour HOD/LODAfter the first 60 minutes of RTH establishes the day’s opening high and low, the first time price retests either level tends to fail. Faded mechanically with a small fast take-profit and a tight stop, this produces a clean positive expectancy across 15 liquid US large-caps and ETFs. Locked geometry, walk-forward-validated.
The trade
- Establish: first 12 5-minute bars (09:30–10:30 ET). Record
HOD_est= max high andLOD_est= min low, plus ATR(14) at the end of the window. - Entry: after 10:30, on the FIRST bar where price retests either level:
- bar.high ≥ HOD_est → SHORT at HOD_est
- bar.low ≤ LOD_est → LONG at LOD_est
- Take profit:
0.30 × ATRfavorable from entry. Small, fast, fills ~87% of the time. - Stop:
0.75 × ATRadverse from entry. Tight enough that the math works at this win rate. - Time exit: 15:30 ET — close any open position before the closing auction.
- NO scaling in. The discretionary instinct is to average down when the extreme breaks. The backtest is clear: scaling converts a winning trade into a losing one. Take the stop loss instead.
Headline OOS
n trades 1,619 (out-of-sample, second half of period)
expectancy +0.144R per trade
95% CI [+0.121, +0.166]
win rate 87%
profit factor 2.08
median trade +0.327R
Replication on NEW tickers
discovery set SPY · QQQ · AAPL · MSFT · NVDA (5 tickers)
validation set +10 new (META · GOOG · AMZN · TSLA · AVGO · AMD · JPM · IWM · XLF · XLK)
NEW set OOS +0.156R per trade · CI [+0.128, +0.184]
individually positive 10/10 new tickers
The new validation set outperforms the discovery set. The edge is not over-fit.
Walk-forward stability by quarter
Locked geometry, sliced by calendar quarter. 5 of 5 quarters positive, all with CI95 clear of zero. Magnitudes consistent (+0.11 to +0.14R), win rates stable (86–88%). Not a regime artifact.
| Quarter | N | Mean R | CI 95% | Win % | PF |
|---|---|---|---|---|---|
| 2025Q2 | 411 | +0.117R | [+0.066, +0.163] | 88% | 1.81 |
| 2025Q3 | 838 | +0.109R | [+0.076, +0.140] | 87% | 1.77 |
| 2025Q4 | 835 | +0.140R | [+0.107, +0.171] | 88% | 2.10 |
| 2026Q1 | 821 | +0.137R | [+0.102, +0.169] | 86% | 1.95 |
| 2026Q2 | 273 | +0.128R | [+0.066, +0.182] | 86% | 1.88 |
Per-ticker OOS
OOS half (Nov 2025 – Apr 2026), locked PT 0.30 ATR / stop 0.75 ATR. Sorted by mean R.◆ = CI95 clear of zero
| Ticker | N | Mean R | CI 95% | Win % | PF | Set |
|---|---|---|---|---|---|---|
| ◆ TSLA | 106 | +0.236R | [+0.153, +0.307] | 92% | 3.52 | NEW |
| ◆ XLK | 113 | +0.202R | [+0.133, +0.265] | 94% | 3.67 | NEW |
| ◆ IWM | 117 | +0.187R | [+0.107, +0.255] | 92% | 3.07 | NEW |
| ◆ GOOG | 100 | +0.168R | [+0.070, +0.254] | 86% | 2.28 | NEW |
| ◆ AMZN | 106 | +0.167R | [+0.072, +0.252] | 88% | 2.25 | NEW |
| ◆ META | 103 | +0.158R | [+0.059, +0.244] | 86% | 2.09 | NEW |
| ◆ AAPL | 100 | +0.153R | [+0.059, +0.239] | 86% | 2.16 | OLD |
| ◆ AVGO | 104 | +0.153R | [+0.052, +0.242] | 84% | 2.02 | NEW |
| ◆ AMD | 101 | +0.137R | [+0.037, +0.233] | 82% | 1.83 | NEW |
| ◆ MSFT | 107 | +0.134R | [+0.035, +0.222] | 86% | 1.88 | OLD |
| ◆ QQQ | 116 | +0.119R | [+0.028, +0.202] | 89% | 1.88 | OLD |
| ◆ NVDA | 110 | +0.119R | [+0.013, +0.215] | 84% | 1.70 | OLD |
| ◆ XLF | 113 | +0.088R | [+0.003, +0.161] | 90% | 1.75 | NEW |
| SPY | 119 | +0.086R | [-0.004, +0.161] | 92% | 1.74 | OLD |
| JPM | 104 | +0.056R | [-0.056, +0.161] | 78% | 1.27 | NEW |
How it was found
- Started with a discretionary trader’s playbook: fade session extremes, scale in if they break, exit at first entry on return.(Strategy L)
- Found the scale-in was a destroyer:trades that scaled averaged −0.49R; trades that didn’t (extreme held) averaged +0.12R.(L diagnostic)
- Stripped the scale-in; pure fade at +0.5 ATR PT / 2 ATR stop was at knife-edge break-even (78% win, profit factor 1.04).(L2)
- Swept the (PT, stop) geometry over a 4×4 grid. The whole PT=0.30 ATR row was positive in both IS and OOS at every stop level — structural pattern, not single-cell luck.(L3)
- Validated on 10 new tickers not in the discovery set. All 10 individually positive OOS; NEW set outperformed the original 5.(L4)
- Confirmed stability across 5 quarters. Every quarter from 2025Q2 to 2026Q2 has CI95 clear of zero; magnitudes within a tight range.(L5)
Honest caveats
- Test window is May 2025 – April 2026. The walk-forward shows stability within this period; longer history may reveal regime-dependence we can’t see here.
- Execution assumes realistic but optimistic fills: 2 bps entry slippage, 4 bps stop slippage, $0.005/share commission. Real-money fills will vary, especially on small-cap or low-liquidity names.
- R denominator is 1× stop distance (0.75 × ATR). Per-trade dollar size depends on your account risk per trade. At 0.5% risk/trade, the headline +0.144R implies ~+7 bps of equity per trade before compounding.
- 14 of the 15 tickers tested are mega/large-cap or major sector ETFs. The strategy has not been validated on small caps, single-stock micro-caps, or non-US universes.
Use it
On the chart route, open the ƒx menu and toggle First-hour fade (HOD/LOD retest). After 10:30 ET the chip surfaces the live state — established HOD/LOD, the projected entry / target / stop if the trade fires, and the realized R when it resolves. Open the chart →