Why settlement data is the only honest surveillance signal in binary markets · September 2026
Binary prediction markets have a property no other market has: every contract confesses. A share of equity never tells you whether the buyer was right. A binary contract does — at settlement, every position resolves to a public yes or no, and every account's track record becomes a falsifiable statistical object. This is why surveillance for event markets should start at the end of the trade, not the beginning.
Most surveillance programs are inherited from equities and futures: order-book patterns, spoofing signatures, wash detection, messaging ratios. Those controls matter — binary markets add their own gateway physics (nothing should ever cost more than $0.99; YES+NO should never exceed $1.00) — but they share a weakness: they look for how someone traded, and informed insiders trade boringly. A trainer's cousin buying 3,000 contracts at fair value fifteen seconds before the injury wire produces a pristine order book. There is nothing mechanically manipulative about the most damaging trade an event market will ever host.
What that trade cannot hide is its outcome — and its pattern of outcomes. Compute, per market, the settled win rate of the whole population: that's your expectation, derived from your own venue, not an assumed 50%. Then z-score every account against it. An account that wins 94% of 38 settled markets against a 52% expectation is six standard deviations from uninformed. That is not a trading style. It is an information source with a login.
Two refinements make it prosecutable rather than suggestive. First, timing: cross-reference each outlier's entries against catalyst timestamps. "Statistically impossible AND positioned before the news in 7 of 9 wins" is a referral, not an anomaly. Second, identity: resolve accounts sharing funding rails, devices, addresses, and wallets into operator clusters before scoring — insiders split across accounts precisely to dilute this signal; the graph reassembles it.
Under CFTC Reg 38.156/38.157, a designated market must maintain automated surveillance capable of detecting abuse — and the staff's posture on sports event contracts makes clear the bar will be event-market-specific, not a futures program with the tickers changed. A program whose flagship control is settlement forensics is complete (every account scored, every day), calibrated to the venue (expectations from your own markets), and falsifiable (thresholds backtested against planted ground truth, changes under maker-checker).
The uncomfortable corollary: if your surveillance stack can't answer "who is statistically impossible on my venue right now?", you are running a sportsbook-shaped risk program in a market-shaped regulatory regime. The data to answer it is already in your settlement table.
DataTap runs this as a service — the proof-of-value is one CSV: for sports venues · for DCM applicants.