What Tally caught
A delayed sample of past detections: closed, suppressed, or older than 72 hours, so it carries zero actionable alpha. Live, tradeable detections live behind the app.
Ether trades like leveraged Bitcoin: across past cycles it hasn't cleared its own highs without Bitcoin clearing its. With Bitcoin-above-$100k priced at 58% and Ether-above-$4k at just 40%, the model reads the Ether market as cheap to Bitcoin and leans long Ether-YES against Bitcoin-NO to capture the catch-up.
An LLM-inferred soft relationship, not a logical lock. The edge is expected value from convergence rather than a guaranteed payoff, so it's sized and ranked more conservatively (note the lower resolution confidence). This is the judgment-heavy edge a plain arithmetic scanner can't see.
The three mutually-exclusive Fed-decision outcomes price to 0.959 in aggregate (< 1). Buying every YES leg locks a guaranteed $1 payout for less than $1, net of fees.
Two-sided books with real depth behind the touch, so the spread survives slippage.
Mid prices across the 40 World Cup outcomes sum to 0.815, a tempting 18.5pt theoretical edge.
Filtered by the executability layer: the longshot legs are one-sided, with a best ask of ≈ $0.95 against a $0.005 mid. The arb is illusory; you can't actually buy those YES legs near their quote.
On the surface a federal indictment implies an indictment, suggesting a 9pt mispricing.
Suppressed by the resolution layer: the markets settle on different underlying facts. One resolves only on a federal charge, the other on any state or federal charge. Not a true contradiction.