Peer-earnings sector sympathy

Thesis (rejected at S&P 500 resolution): When ≥3 same-sector peers report earnings in the trailing 5 trading days with a positive median surprise, the not-yet-reporting names in that sector should drift up over the next 5-10 days as the cohort signal leaks across the sector before analysts revise the non-reporters' numbers.

Code: research/experiments/peer_earnings_sympathy.py Report: research/experiments/peer_earnings_sympathy_report.json (2018-2026, weekly cadence, median 433 non-reporter names per day)

Why it failed:

ICs are within sampling noise across the 5-21d window. Two plausible explanations: (1) cross-sectional dilution — a coarse "median peer surprise" is drowned out by name-specific noise when 300-450 names are eligible on a typical day; (2) sector ETF arbitrage — large-cap S&P 500 sympathy is captured by SPDR sector flows intraday before the close.

Verdict: Not actionable on S&P 500. Worth re-running on S&P 600 small caps where sector ETFs are thinner, and on a tighter eligibility filter (cluster size ≥ 4, name's own earnings within 21 days). Code left in place for that follow-up.