TxLINE's vig-free fair leads; the prediction market book lags. When the gap opens past the threshold, the cheap side is underpriced: a divergence. Two things say the signal is real: whether the book then travels back to TxLINE's price (the delay closing), and whether the cheap side is a positive-edge buy at resolution. Both are measured on the real fills, pooled across matches.
the signal, across 16 matches · θ = 5pp divergence
same-minute dedupe
81%
reached TxLINE (the delay closes)
+1277%
ROI · Kelly-sized, take-profit at fair
90
calls · $75,074,982 size available
why Kelly + take-profit:+1277%take profit at fair-63%same bets, held to the result
Reach = did the prediction market price travel to TxLINE's line before full time. ROI = the compounding return of Kelly-sized bets (f = gap / (1 − price), capped at 30% per call) that exit at TxLINE's fair on reach, else mark out at the close; never held to resolution. Same signal, same bets: taking profit on convergence pays; holding to the result loses on a coin-flip. The compound is concentrated in a few high-volume matches, so treat it as a pilot that firms up as matches accrue. Size available = the liquidity you could exit into at the take-profit price (TxLINE fair or better); nil when the price never reached fair.
sort
▮ gap on the cheap sidereach ✓ = price travelled to fair90 calls · click a row to read it
This measures the mispricing, not a trading strategy: how much someone stakes, and any price they move doing it, is their own execution cost, not part of the signal. Fills and outcomes are the same on-chain data the ledger settles against.