Insider Pattern Detection · Polymarket · Behavioral Analysis
Leviathan is an automated behavioral surveillance algorithm built to monitor large-scale trading activity on Polymarket — a public prediction market where participants bet on real-world geopolitical and political outcomes using cryptocurrency.
The core question Leviathan asks is simple: does this bet look like someone knew something before the rest of the market did? When a position of significant scale is placed against prevailing odds — particularly by an account with unusual behavioral characteristics — Leviathan flags it for review.
All data analyzed by Leviathan is sourced exclusively from publicly accessible on-chain records. No private data is accessed. No systems are compromised. Leviathan reads only what the blockchain already makes public, and subjects it to pattern analysis unavailable to casual observers.
Leviathan operates in continuous scanning cycles, querying Polymarket's public data infrastructure for geopolitical and political markets exhibiting significant activity. Positions that fail to meet qualifying entry points may still be marked for review should other qualifiers raise enough concern.
Each qualifying position is run through a proprietary multi-signal analysis pipeline. The algorithm cross-references multiple dimensions of on-chain behavioral data to produce a composite Risk Score — a single number representing how anomalous a given position appears relative to expected market behavior.
The signals, weights, thresholds, and combination logic that produce Risk Scores are not disclosed. Publishing the scoring methodology would allow bad actors to engineer positions that evade detection. Leviathan's analytical framework will remain proprietary for as long as it needs to in order to remain effective.
Prediction markets have demonstrated a consistent ability to price geopolitical and political risk more accurately than traditional polling or expert analysis. This makes them valuable — and it makes anomalous activity on them potentially informative in ways that extend well beyond the markets themselves.
A large bet on an unlikely outcome that subsequently occurs is not necessarily evidence of wrongdoing. Markets move. People are sometimes right. But certain patterns of on-chain behavior — visible to any observer with the right tools — are difficult to explain through luck alone.
Leviathan does not make accusations. It surfaces patterns. What those patterns mean is a question for journalists, researchers, and the public to evaluate.
Leviathan's flagged results are published to a public watch feed on a delayed basis. Results are withheld for a fixed period before publication to prevent the feed from being used as a real-time trading signal — which would defeat its purpose and create the very information asymmetry it exists to expose.
The watch feed shows flagged wallets, market questions, bet sizes, and Risk Scores. It does not explain how scores are derived. It is intended for transparency, accountability, and public interest research — not speculation.
Access the Watch FeedLeviathan is published in good faith as a public interest and whistleblowing instrument. The goal is not disruption — it is transparency. Large-scale prediction markets touching geopolitical outcomes carry a legitimate public interest dimension that warrants independent scrutiny. This project is our contribution to that.
This is not insider trading. Leviathan does not access, solicit, or act on non-public information of any kind. It performs pattern analysis on data that is already visible to anyone with the technical means to read a public blockchain. The algorithm infers — it does not intercept. Drawing conclusions from publicly observable behavior is the same work performed by financial journalists, academic researchers, and on-chain analysts every day. Leviathan systematizes that work and makes the output publicly available.
A note on where we think this is heading, from Elodine at Tenebrous Tales Interactive:
Watch algorithms like Leviathan occupy a space that feels familiar to me. Four years ago, automated stock trading bots were rare, expensive, and proprietary — the kind of thing only well-capitalized firms could build and run. We built some ourselves: paper trading bots that outperformed the S&P 500 by 3x over four months. At the time that felt like an edge. It wasn't long before that edge evaporated, as the tools proliferated and the strategies raced toward the mean.
Market watch algorithms are at a similar inflection point. Right now they're uncommon. That won't last. And when they become widespread in pari-mutuel environments like Polymarket, the consequences fall unevenly — it's the average participant who loses, as information asymmetry further fragments the pool. Sophisticated actors extract value; everyone else subsidizes it.
Polymarket profits on volume regardless of outcome. That structure gives them little direct incentive to monitor for — or advocate against — the kind of activity Leviathan is designed to surface. The house always wins.
For media inquiries, research collaboration, or press contact:
✉ [email protected]Leviathan analyzes only publicly available on-chain data. No private systems are accessed. Flags indicate statistical anomaly only — not evidence of illegal activity. This tool is intended for research, journalism, and public interest transparency. Nothing published by Leviathan constitutes legal, financial, or investigative conclusion. All wallet addresses analyzed are pseudonymous public identifiers recorded on the Polygon blockchain.