Case triage for event-market teams with account-level trading records
Public records show how trading developed around an event. Authorized partner records can add the account link, restrictions, and prior activity that public feeds cannot. Those evidence layers feed one ranked queue, with the supporting record attached to each case. The public replay uses real Polymarket trades; the partner workflow is implemented and has so far been evaluated only with synthetic records.
Public activity alone cannot identify a person or escalate a case. ReventLens prioritizes review; it does not determine misconduct.
Reconstructs and ranks public trading for replay. Without private attribution, it cannot identify a person or support escalation.
Checks the supplied account link, restricted-list status, baseline position size, and prior activity, then records what changed and why. Context can raise priority, lower it, or block escalation. Demonstrated with synthetic records; local file import, nothing uploaded.
Stale or conflicting partner context blocks escalation until resolved.
Public records show how trading developed around an event. A partner may separately supply the account link, restrictions, and prior activity needed for deeper review.
Reconstruct trading around a defined event, then rank its timing, concentration, direction, and source completeness.
Local CSV or JSON intake can map an account link, restricted-list match, baseline position size, and prior activity. The current examples use synthetic records.
Evidence gates control the handoff. Public-only activity cannot escalate, and the investigator decides what the record means.
Public trades show behavior. A partner's records may link an account to a known subject and show whether it was restricted. The prototype does not infer one from the other.
The public layer reconstructs trades, ranks unusual activity, records provenance, and prepares a review packet.
Fictional records show how an account link, restricted-list status, baseline position size, and prior trading activity can move a case into deeper review or suppress it from the review queue. Suppression is a triage decision, not a finding that the activity was proper.
The synthetic control raised a warning. When the same simulated attribution packet was applied to ordinary control wallets, 8 of 18 that could be evaluated would also have been routed for deeper review. That is not a false-positive rate. It shows that attribution alone is not proof and that real partner validation remains necessary.
Inspectable output
The packet makes the engine's reasoning reviewable. Here one fictional case is scored twice: first without partner records, then with synthetic partner context supplied. The relevant fields, gates, and routing change are inspectable.
Priority bands: Low 0–69, Medium 70–84, High 85 and up. Routing also depends on attribution and evidence gates, not the score alone.
Pass 1 · Without partner records
Same case, partner context added
Pass 2 · With synthetic partner context
A prototype score and severity, not a probability of wrongdoing.
Public replay evidence
ReventLens reconstructed the public Polymarket trading behind the Google Year in Search episode: 1,132 pre-reveal trades and 201 wallets positioned on the eventual winner. Among the twelve largest wallets the replay scored, each holding at least 1 percent of the pre-reveal volume, the engine ranked one anonymous wallet first, mostly on timing: it built its position from about 1.2 to 1.0 hours before the public reveal. It was also the most concentrated wallet in the replay, at 14.7 percent of the analyzed winner-side volume. Neither timing nor concentration implies misconduct.
The engine also held that wallet below escalation because public trading did not provide account attribution. That restraint is not a disclaimer added after the result. It is enforced in the scoring engine.
We selected this episode after the allegations became public. ReventLens did not surface it first; the replay shows how the workflow ranks known public trading, not a discovery. This is a retrospective workflow demonstration, not measured detection performance. The underlying conduct is alleged in the Justice Department announcement and CFTC complaint.
Designed restraint
A surveillance tool creates risk when it is more confident than its data. ReventLens makes the handoff point explicit.
A scoped next step
ReventLens is looking for event-contract venues and the compliance teams that lawfully hold account, restriction, and trading-history records. Scoping starts with a 30-minute review of the public replay, the synthetic account fields, and the records a historical pilot would need.
Scope a controlled pilotBring a real surveillance process. Review what public trades show, what partner records add, and which questions still require an investigator.