Why No SELL Signal: The 28-Year Statistical Bar
28-year back-test report — summary · 26 August 2026
We set one bar for a SELL label: n > 500 signals, win rate > 55% at 20 days, with a margin of error that separates above 50%. The BUY signal clears it comfortably; the exit-side candidates finished below it. Three structural findings explain the picture.
1. Same zone as CONTINUATION: complementary by design
CONTINUATION fires when a sector is in accumulation and closes above the upper band — i.e., already strong. Historically, price is higher 20 days later 60.1% of the time.
A "sell into strength" rule in the same zone is the exact complement: it wins when price falls, i.e., at most 39.9% — the ceiling comes directly from the momentum base rate. Candidates in this family landed at 33.9%–41.8% drop, precisely inside the predicted 38–42% band.
2. A different zone → it's the funnel into BUY
The final candidate (D1) tested a genuinely different zone: the first day a sector flips to distribution (money flow ≤ 40 and relative strength negative) while price is still above the lower Bollinger band — before BUY's zone is even reached.
Result: price rose 61.4% of the time at 20 days (vs 60.1% for CONTINUATION), and 57.5% of those days later printed a BUY. The zone marks the doorway into the dip — an early-BUY at an earlier stage.
3. Rotation answers a different question
Sector Orb is a leaderboard: sectors shift between Leader / In-between / Laggard. "This sector is losing relative ground" is a measurable, real observation about standing among the other 10 sectors and the market — a separate scale from forward price moves.
The rotation metrics on the final candidate: 50.9% ±1.8pp underperform peers (a coin flip), and the sector's own rank improved 69% of the time — these markers track recovery as readily as drift. The strongest relative-vs-SPY number anywhere in the research (59.9% at n=227) speaks to relative pace, not price direction.
4. The deeper reason: the market's asymmetry
Downside mean-reversion is stronger than upside. Markets are notoriously easier to time on the entry side (buy dips) than the exit side (sell tops). Downside skew and persistent upward drift mean the indicators that make a clean BUY read as noise on the way down. This is consistent with the practitioner literature.
Our data says the same: BUY shows a +12.7pp edge over 50% at 20 days; the exit-side family sits −10 to −18pp below coin flip — a mirror image. That asymmetry is the finding.
| Entry side (BUY) | Exit side (every SELL tried) | |
|---|---|---|
| Best spread vs coin flip @20d | +12.7pp | −10 to −18pp |
| Sample size behind it | n=2,403 | 30+ variants, n up to 6,016 |
| Verdict | published | research archive |
5. Where the product stands
- Labels carry their evidence. Every published signal comes with sample size and margin; early-stage findings stay in research until they clear the same bar.
- No "beats the market" claim either. BUY's excess win rate vs SPY is ≈ 49% ±1.0pp — essentially breakeven once the market's own drift is subtracted. What the data supports: timing edges exist on the entry side.
- Methodology is published. Every number carries n and ±SE; samples under n<50 stay out of conclusions; each time window keeps its own analysis; thresholds were fixed before evaluation.