Merger Arb (448)

Cash-merger spread capture — the deal-break-insurance premium (research/448, rules frozen 2026-07-27). A machine-built EDGAR calendar catches every 8-K announcing a definitive CASH merger (1,109 events 2022-2026, 9/10 known-deal validation incl. delisted targets), parses the offer price from the filing itself, and buys the TARGET at the next session's close when the deal trades at a 0.5-30% discount to the offer. Diversified across up to 20 concurrent deals, the book collects the spread as deals close (136 of 190 backtest deals delisted at ~the offer) and eats the occasional break — that asymmetry, not direction-picking, is the edge being sold. Long stock only: RH-executable, ~1-2 entries/week, fits the pre-clearance budget.

• Signaldefinitive cash-merger 8-Ks via EDGAR full-text search (6 query batteries); offer price parsed from the press-release/8-K text; target confirmed by premium band 0-120% vs prior close
• Entry / exitT+1 close; entry spread 0.5-30%, price >= $2, 21d median $vol >= $2M; hold to deal close (delisting) or 378-td timeout; 20 equal slots of $5k, skip when full
• Backtest (TRUE capacity)research/448b at survivorship-free delisted-inclusive prices: +7.4%/yr, Sharpe 1.36, monthly t=3.12, beta 0.20, maxDD -6.0%, 4/4 years positive; per-deal +2.48% (t=3.03), 90% win, ex-top3 survives
• Honest caveatRECENT-REGIME window (2023-08+; the 2022 rate-shock break cluster is NOT in sample); beta-adjusted alpha +3.4%/yr at t=1.98 (marginal); per-deal capture compressing (2023 +5.1% -> 2025 +1.6%); worst deal -44.6% (deal repriced). Paper book starts EMPTY 2026-07-27 and ramps as new deals arrive; kill criteria pre-registered (trailing-12m Sharpe < 0, median captured spread < 1%, or new-deal spreads < 1.5% for 3 months)