Pairs trading + relative-strength rank.

In a horse race, the disciplined bet isn't on whether the field finishes — it's on which horse finishes ahead of which. Pair trading applies the same logic to equities: long the leader, short the laggard, in two related names. The market can rise, fall, or chop and the spread between them is what you're betting on. The framework's RS-rank methodology computes relative strength across the universe and surfaces high-RS / low-RS candidates daily. Pair trades work when the spread reverts cleanly; they explode when one leg breaks structurally and the relationship the trade was built on fails.

What relative-strength rank actually measures

RS rank is a percentile on rolling-period total returns within a peer group. The framework computes:

• 1-month, 3-month, and 6-month total returns per ticker
• Weighted composite (50% × 6mo, 30% × 3mo, 20% × 1mo) — recent returns matter more
• Percentile rank within sector and within index (S&P 500)

A name with RS rank 92 is in the top 8% of its peer group on the composite measure — outperforming 92% of comparable names over the relevant windows. RS rank 8 is the bottom 8% — chronic laggards by the same measure.

The pair trade structure

A pair trade is two simultaneous positions — long one name, short another, sized for delta-neutral exposure (equal dollar amounts or equal beta-adjusted dollars depending on the setup). The thesis isn't about market direction; it's about relative performance. If you long NVDA (RS 90) and short INTC (RS 25) in equal dollars, the trade profits when NVDA outperforms INTC, regardless of whether both are up, both are down, or one is up and one is down.

Three structural variants the framework considers:

1. Sector pair — same sector, different RS. Long NVDA / short INTC. Both names move with semis broadly; the spread captures intra-sector dispersion.
2. Index pair — same index, different sector exposure. Long high-RS Tech composite vs. short low-RS Energy. Captures regime shifts in sector rotation.
3. ETF pair — long high-RS sector ETF, short low-RS sector ETF. Cleaner than single-name pairs because diversification reduces single-name blowup risk.

When pair trades work

The cleanest case: two same-sector names with high correlation (≥0.65 over 60 days), one with RS rank ≥85 and the other ≤25. The market can drop 4% on a Fed scare; both legs participate, but the leader holds up better than the laggard. The pair makes money on the relative outperformance even though the long leg is down in dollar terms.

Pair trades also work as portfolio overlays: if your equity sleeve is heavily long-biased (Sports Cars + Anchors at 70% deployed), a pair trade adds market-neutral exposure without taking on additional net long risk. The pair contributes return without contributing to drawdown when SPY drops.

When pair trades explode

The blow-up case: the relationship between the two names breaks. NVDA / AMD has been a tradeable pair for years; if AMD wins a major customer that NVDA loses, the spread can move 30-40% in a week. The "long leader / short laggard" trade becomes "long NVDA / short the new leader." The pair stops out at maximum loss because both legs move against you simultaneously.

The framework's defenses against pair-trade explosion:

• Both legs sized to 1% portfolio risk each (so total pair = 2% risk, not 1%)
• Stop on the pair fires when RS spread compresses below 30 points OR correlation drops below 0.4 over a rolling window
• Pair sleeve capped at 5% of portfolio total — even a full-blow-up of multiple pairs is recoverable
• No pairs through earnings on either leg — single-name event risk doesn't compose well with relative-strength bets

The pair-trade gate composition

A pair trade has to pass:

1. RS spread ≥ 50 points between the legs
2. 60-day correlation ≥ 0.50 (better if ≥ 0.65)
3. Both legs liquid enough for the position size (avg daily volume × $price ≥ 10× position)
4. No earnings on either leg within the planned hold window
5. Same sector preferred; cross-sector pairs require explicit thematic linkage (e.g., long high-RS oil major / short low-RS oil major — clean; long high-RS tech / short low-RS retail — incoherent)

The framework's RS-rank cache surfaces candidate pairs daily via the pair-trade screener. The trader doesn't search for pairs manually; the system surfaces them and the audit decides which clear all five gates.

The cargo-cult pair trade

"Long NVDA / short SPY" or "long QQQ / short SPY" gets pitched as pair trades. They're not — they're two directional bets dressed as a pair. The correlation between QQQ and SPY is ~0.95; the spread is mostly noise. A real pair captures dispersion between names that should track each other; QQQ-vs-SPY captures dispersion between two index ETFs that contain mostly the same names. The framework refuses these at the screener gate.

The real lesson

Pair trades are a market-neutral overlay built on relative-strength dispersion. The framework's RS-rank methodology surfaces candidate pairs; the audit composes correlation, spread, liquidity, and event-window checks before approving. Pairs work when the relationship holds; they explode when it doesn't. The size discipline (2% pair total, 5% sleeve max) bounds the explosion risk. The dispersion is the trade — not the direction of either leg.

Related: L29 — Whale Confirmation · L14 — sector rotation