pair-trade-screener

---
name: pair-trade-screener
description: Use when screening for pair trading opportunities, statistical arbitrage, cointegrated stock pairs, mean-reversion strategies, or market-neutral portfolios, including correlation and z-score analysis.
category: Screening
version: 1.0.0
tools: []
---

# Pair Trade Screener

Identify and analyze statistical arbitrage opportunities through pair trading — a market-neutral strategy that profits from the relative price movements of two correlated securities, regardless of overall market direction, using rigorous statistical methods including correlation analysis and cointegration testing.

**Core Methodology:**
- Identify stock pairs with high correlation and similar sector/industry exposure.
- Test for cointegration (long-term statistical relationship).
- Calculate spread z-scores to identify mean-reversion opportunities.
- Generate entry/exit signals from statistical thresholds.
- Size positions for market-neutral exposure.

**Key advantages**: market-neutral (profits up, down, or sideways); limited exposure to broad market moves; data-driven rather than discretionary; diversifying versus long-only strategies.

## When to Use

User asks for pair trading opportunities, market-neutral strategies, statistical arbitrage screening, "which stocks move together," sector-hedge ideas, mean-reversion trade ideas, or relative-value trades — e.g. "Find pair trading opportunities in tech," "Which stocks are cointegrated?", "Find mean-reversion pairs."

## Analysis Workflow

### Step 1: Define Pair Universe

Select a sector (Technology, Financials, Healthcare, Consumer Discretionary, Industrials, Energy, Materials, Consumer Staples, Utilities, Real Estate, Communication Services), a custom ticker list (e.g. AAPL, MSFT, GOOGL, META, NVDA), or a narrower industry within a sector (e.g. "Software" within Technology, "Regional Banks" within Financials). Filter to: minimum market cap $2B (mid-cap and above), minimum average volume 1M shares/day, actively traded (no delisted/inactive), same-exchange preference to avoid cross-exchange complications.

### Step 2: Retrieve Historical Price Data

Gather at least 2 years (252+ trading days) of daily closing prices, adjusted for splits and dividends, via web_search or knowledge_base_search against available market-data sources. Verify consistent date ranges across symbols, drop stocks with >10% missing data, forward-fill minor gaps, and note data-quality issues.

### Step 3: Calculate Correlation and Beta

For each candidate pair, calculate the Pearson correlation coefficient (ρ) and a 90-day rolling correlation for stability. Filter to ρ ≥ 0.70. Interpret: ρ ≥ 0.90 very strong (best candidates); 0.70-0.90 strong (good candidates); 0.50-0.70 moderate (marginal); <0.50 weak (exclude).

Compute beta as Covariance(A,B) ÷ Variance(B) — the hedge ratio (e.g. beta 1.5 means $1.50 of B per $1.00 of A). Check correlation stability across 6-month, 1-year, and 2-year windows; flag pairs where recent correlation falls more than 0.15 below historical.

### Step 4: Cointegration Testing

Correlation measures short-term co-movement; cointegration proves a long-term equilibrium relationship — cointegrated pairs mean-revert predictably, non-cointegrated pairs may diverge permanently.

**ADF Test** (Augmented Dickey-Fuller): for each correlated pair, compute the spread = Price_A − (Beta × Price_B), then run the ADF test on the spread series. A p-value < 0.05 indicates cointegration (rejects the null hypothesis of a unit root); extract the ADF statistic for strength ranking. Interpret: p < 0.01 very strong cointegration; p 0.01-0.05 moderate cointegration; p > 0.05 no cointegration (exclude).

Estimate mean-reversion speed via half-life = −log(2) ÷ log(mean-reversion coefficient): <30 days fast (good for short-term trading); 30-60 days moderate (standard); >60 days slow (long holding periods).

### Step 5: Spread Analysis and Z-Score

Two spread constructions: **Price Difference** (additive) = Price_A − (Beta × Price_B), best for similar price levels; **Price Ratio** (multiplicative) = Price_A ÷ Price_B, best for differing price levels and easier interpretation.

Z-score = (current spread − mean spread) ÷ std dev of spread, using a 90-day rolling window. Interpret: Z > +2.0 A expensive relative to B (short A, long B); +1.5 to +2.0 moderately expensive (watch for entry); −1.5 to +1.5 normal range (no trade); −2.0 to −1.5 moderately cheap (watch for entry); Z < −2.0 A cheap relative to B (long A, short B). Also track maximum historical z-score deviations and check for structural breaks (spread regime change).

### Step 6: Entry/Exit Recommendations

**Conservative entry (|Z| ≥ 2.0)**: if cointegration p < 0.05 and half-life < 60 days — Long signal (Z < −2.0): buy A, short B at the beta hedge ratio; Short signal (Z > +2.0): short A, buy B at the beta hedge ratio. **Aggressive entry (|Z| ≥ 1.5)**: more frequent trades, higher win rate but smaller average profit, requires tighter risk management.

**Exits**: primary — mean reversion, close both legs when Z crosses 0; partial profit — exit 50% at |Z| = 1.0, remainder at Z = 0; stop loss — exit if |Z| exceeds 3.0 (possible structural break); time-based — exit after 90 days if no mean reversion, to avoid holding broken pairs indefinitely.

### Step 7: Position Sizing and Risk

For a pair with beta β, size for equal-dollar market-neutral exposure — e.g. on $10,000 allocated with β = 1.2: long $5,000 of A, short $6,000 of B (market neutral, net beta ≈ 0). Allocate 10-20% of portfolio per pair, cap active pairs at 5-8 for diversification, and avoid highly correlated pairs across the book. Risk limits: max loss per pair 2-3% of total portfolio; stop-loss trigger |Z| > 3.0 or -5% loss on the spread; portfolio-level risk across all pairs ≤ 10%.

### Step 8: Report

Produce a structured chat report: executive summary (pairs analyzed, cointegrated pairs found, top 5 ranked by statistical strength); a cointegrated-pairs table (pair, correlation, cointegration p-value, current z-score, signal, half-life); detailed analysis of the top 10 pairs (metrics, current spread position, entry/exit, sizing, risk); a text-based historical z-score summary with entry/exit levels marked; and risk warnings (deteriorating correlation, structural breaks, low liquidity).

## Quality Standards

**Minimum requirements for a valid pair**: correlation ≥ 0.70 over 2 years; cointegration p-value < 0.05 (ADF Test); confirmed spread stationarity; half-life < 90 days; no structural breaks in the last 6 months.

**Red flags (exclude)**: correlation dropped >0.20 in the last 6 months; cointegration p-value > 0.05; half-life increasing over time (weakening mean reversion); significant corporate events (merger, spin-off, bankruptcy risk); average volume < 500K shares/day.

**Practical considerations**: assume ~0.1% round-trip transaction cost per leg (≈0.4% total for entry + exit on both legs) — the entry z-score threshold should clear this cost; verify shortability and factor in borrow fees and short-squeeze risk; enter and exit both legs simultaneously to avoid leg risk, and use limit orders to control slippage.

**Common pitfalls**: survivorship bias (universe excludes delisted/failed names, inflating apparent correlation); look-ahead bias (using data not actually available at the decision date, e.g. restated financials or post-hoc universe selection); overfitting (tuning z-score thresholds or lookback windows to a specific historical sample rather than a robust range).