Methodology, in plain language.
Panoptes watches large trades on Polymarket's political, business, and cultural markets and asks one question of each one: does the size, timing, and price of this trade look like someone who knows something the rest of the market doesn't? It never answers that question with certainty -- it answers it with a score, computed the same way for every trade, so you can see the pattern for yourself instead of taking our word for it.
A single large bet on a long-shot outcome isn't unusual by itself -- people do that for fun, or because they're wrong. What's worth noticing is the combination: an unusually large trade, at unusually confident odds, placed on a market where that kind of advance information plausibly exists, sometimes alongside several other accounts moving the same way at the same time. No single signal proves anything. Panoptes combines several of them and lets you see the combination directly.
Every trade gets scored by three separately-tuned strategies, shown side by side rather than collapsed into one number. They agree often enough to be useful and disagree often enough to be informative -- a trade all three flag is a stronger signal than one only the most aggressive strategy catches.
A smooth 0-100 score built from a statistical z-score across the signals below, gated by a $500 minimum trade size. This is the default view on the Flagged Trades feed and what the leaderboard's accuracy numbers track first.
Weighs the actual dollar amount at risk more heavily than the other two -- built to surface "someone just bet a large, specific sum on a longshot," which a purely statistical z-score can under-weight for an account that trades often.
No continuous math: a trade has to clear a $500 floor and trip at least two of three secondary signals. Less sensitive than the other two, but every flag it produces can be explained in one sentence, which makes it the easiest to audit.
Trade size, measured relative to that specific market's own typical trade size, not a flat dollar threshold -- a $5,000 bet is unremarkable in a market where that's normal and striking in one where it isn't.
Contrarian odds -- how far the bet goes against the market's own implied probability. This only counts for a YES bet on an unlikely outcome. Betting NO on one specific long-shot outcome in a market with many possible outcomes wins trivially often on its own -- it isn't a clean signal of advance knowledge, so Panoptes treats it as neutral rather than rewarding it.
Information asymmetry ("info score"), a 1-10 rating Panoptes assigns to each market for how plausible it is that someone could have real advance knowledge of its outcome. Most markets score 1-3; a score of 8-9 is rare, roughly the top 5%.
Account history -- whether the wallet placing the trade is new or has a long track record, and whether its resolved trades have actually won more than the odds it bet at implied.
Coordinated clusters -- multiple distinct accounts betting the same side of the same market close together in time. Typical clusters run 3-34 accounts; a cluster by itself isn't proof of coordination, but it raises the same question the rest of the score asks: why did this many people move at once?
Every flagged trade eventually resolves, one way or another -- the market it was placed on settles. The leaderboard's accuracy numbers come from comparing what each strategy flagged against what actually happened, both live (trades the system actually scored in real time) and via backtest (the same scoring logic applied retroactively to resolved history). We show both, and we show when a number is backtest-only because the live sample is still too small to trust on its own.
Panoptes never executes trades, holds funds, or has custody of anything on Kalshi, Polymarket, or anywhere else. Every score is an automated statistical estimate based on public trade data -- not evidence, not an allegation, and not a claim that any account or trade broke any rule. Nothing here is financial, legal, or investment advice.
Kalshi data is collected but deliberately not shown anywhere on this site -- Polymarket's and Kalshi's terms of service differ, and Panoptes only displays what it has a clear basis to display. See the Terms of Service for the full scope.
Unfamiliar term? The Glossary defines every field and score used across the site.