Double-7s Channel Dip-Buy

Status: Adopt candidate (paper‑trade first) · Library entry #4 · Found 2026‑08‑06 Instruments: QQQ, XLK (primary) · SPY (near‑miss, see evidence) · US index/sector majors only Market/timeframe: stocks, 1d · Direction: long only · Frequency: ~6–7 trades/yr, ~5‑day holds

⚠ Read this first: same family as entry #2

This is a channel‑based cousin of the RSI(2) dip‑buy — both buy short sharp dips in an uptrending index and sell the snap‑back days later. They fire on overlapping days and draw down in the same panics. Running both at full size is one bet counted twice, not diversification. Pick one, or size the pair as a single risk sleeve. It earns its own entry because the trigger is structurally different (price channel, no oscillator), the exit is symmetric (strength, not a moving average), and it passed the full bar independently — including the 2000–2010 stress decade on every instrument tested.

The edge

Buy an uptrending index ETF when price breaks below its prior 7‑day low (a washed‑out dip), sell when it breaks above its prior 7‑day high (confirmed strength). Connors & Alvarez published the "Double 7s" spec in 2009; the parameters are frozen literature — every year since is genuine out‑of‑sample.

The setup (compilable spec)

Market: stocks, symbol QQQ (or XLK), trading timeframe 1d. Longs only. Entry (daily close): close > SMA(200) AND close < lowest(7) (below the prior 7 days' lows). Exit (daily close): close > highest(7) (above the prior 7 days' highs). No take‑profit, no stop‑loss (the entry‑#2 family rule: a stop sells the flash‑crash low). Fees: 0.03%/side; robust at 0.1%/side.

Stated deviation from the book: Connors uses 7‑day closing extremes; the engine's rolling extremes are the prior 7 bars' lows/highs (slightly stricter entry, slightly later exit). Reported as tested — nothing was tuned.

Validation evidence

Same engine, warm‑up included, 0.03%/side, dividends excluded. DD‑from‑peak (account basis) in parentheses; MC p95 = 95th‑percentile max drawdown across 1,000 trade‑order shuffles.

• Two windows: ✅ QQQ + XLK positive and strip‑best‑surviving in both — and in the 2000–2010 stress decade too (all three instruments).
• Cross‑asset: ✅ QQQ + XLK clean. SPY is a near‑miss: every return/strip‑best check passes, but the 2018–2026 sum‑point drawdown (22.6, the covid crash held with no stop) breaches the ≤ 20 bar — its account‑basis DD is 19.1%. Reported, not rounded away. EEM fails the drawdown bar outright.
• Cost stress (0.1%/side): ✅ QQQ PF 2.43/1.70, XLK 1.57/2.08, SPY 2.33/2.09.
• Monte‑Carlo: the shuffled p95 drawdowns (17–29 points) exceed the observed ones — size for those, not for the path you happened to get.

Guardrails (read before trading)

• Do not stack naively with entry #2 — same mechanism family, same panic exposure (see the top of this page). One dip‑buy sleeve, not two.
• US index/sector majors only. EEM carries 24–32% drawdowns; nothing else was tested.
• No stop means gap risk is real — identical to entry #2: size the sleeve so the worst observed trade is survivable; a stop measurably hurts this family (it sells the low).
• Expect ~65–75% winners with average loss ≥ average win — the mean‑reversion shape. Set signalMode: "early" in the app; the fill‑honesty finding on entry #2 applies here equally (entering near the close preserves the edge; next‑open thins it).

How it was found

Pre‑registered in sweep #5 (research log) as a frozen‑parameter literature transplant, alongside a metals trend‑timing hypothesis that was killed (window‑flip: gold's 2011–2015 bear) and an intraday candidate blocked on data. No neighbors were run — the parameters are the book's. The SPY and EEM misses are reported above exactly as they landed.