.png&w=828&q=75)
Options are priced with time-to-expiry in calendar days. Variance only accrues on trading days. Those two clocks disagree across every weekend, and a short delta-hedged ATM straddle is positive-EV exactly when IV² × calendar_days > RV² × trading_days. RAMON measures that disagreement from underlying price history alone, prices it in dollars at executable quotes, and sells volatility only where the gap is collectable. Everywhere else it does nothing and says which number stopped it. It ran live at 14:28 UTC on submission day and refused to trade: Monday is Labor Day, and a three-day weekend buys an extra day of decay against a gap that delivers more movement. That gate was written beforehand. At 14:50 UTC it filled a real multi-leg paper order at 0.55 against a 0.575 arrival mid — a 1.67x capture ratio, filled inside the natural. At 13:33 UTC Alpaca's multi-leg close bug (#774) orphaned a leg; the RECONCILE stage caught it in 8 seconds and was flat in 77. Evidence: SPY/QQQ/IWM option chains 2011–2025, 46.1M contract-days, every price the side that would actually be hit. Walk-forward Sharpe 2.29, CI [+1.01, +4.00]. 28 purged, embargoed CPCV paths with probability of backtest overfitting 0.0%. Placebo test: 0 of 200 fake calendars beat the real one. The edge survives prior-art neutralization at 102% Sharpe retention against variance risk premium, market beta and momentum. Verdict: INCONCLUSIVE — and that is the finding. The sign is defended five independent ways; the size is not established, and n=237 weekends cannot establish it. Thirty-one specifications died building this, and the six methods that killed them are wired into the console as an adversarial scanner you can point at your own strategy.
4 Sep 2026