How Polytrack works
A transparent method for screening Polymarket activity.
Polytrack is a screening system, not a verdict engine. It ingests available Polymarket trade and market context, normalizes those records, evaluates several size, timing, market, and wallet signals, and assigns LOW, MEDIUM, or HIGH severity so a researcher can decide what to inspect first. The label describes conditions in Polytrack's monitored dataset; it does not identify a person, prove intent, establish non-public knowledge, or predict the market result.
Reviewed July 14, 2026
1. The unit of analysis is a monitored trade.
Polytrack starts with an execution associated with a Polymarket market, outcome, side, wallet address, timestamp, price, and size when those fields are available. Market metadata and wallet context are attached to that record. A record becomes part of Polytrack only after the monitor receives and processes it, so the database is a monitored subset rather than a claim of complete exchange history.
The wallet address is the research key. It can connect activity, positions, prior flags, watchlists, and profile metrics, but it is not treated as verified identity. Public pseudonyms and user-created labels are context fields, not proof of ownership.
- Trade value is the recorded size multiplied by price when the source fields support that calculation.
- Market and outcome labels come from available upstream market metadata.
- Timestamps describe the source event or the closest available observation, not a promised delivery time.
2. Severity combines independent screening signals.
A single large order can be ordinary, so Polytrack does not use one public dollar cutoff as a universal definition of a whale. The classifier can consider trade size, short-window market volume, change from a baseline, price movement, market sensitivity, directional concentration, coordinated-wallet counts, new-wallet context, accumulated volume, and prior flags. A stronger condition raises the label; lower conditions do not cancel a stronger one.
The exact production thresholds are operational classifier settings and are not published on the public site until the product owner approves that boundary. The public method names the inputs and interpretation so readers can evaluate the logic without mistaking a hidden threshold for scientific certainty.
- LOW means the record is retained with limited alert evidence.
- MEDIUM means one or more stronger screening conditions make the record more useful to review.
- HIGH means the record met the strongest configured alert conditions and may be eligible for Telegram delivery.
3. Aggregates stay bounded by coverage and time.
Analytics group stored monitored trades into selected time windows. Panels may sum notional value, count records, count distinct wallets, group recorded sides or categories, and rank markets or wallets. These are descriptive views of Polytrack coverage. They are not total Polymarket volume, open interest, exchange-wide wallet activity, or risk-adjusted performance.
The leaderboard follows the same rule. It groups monitored records by wallet within a selected window, forms a leading candidate set by tracked volume, and derives the displayed volume and average-size rankings from that set. Changing the window or coverage can change every rank.
4. Human interpretation is part of the method.
Polytrack is designed to shorten the path from a signal to the underlying market, wallet, positions, and recent activity. A careful review asks whether the record is isolated, whether the market is liquid, whether a wallet has enough history to support comparison, whether the position changed later, and whether the source was complete at the time of observation.
No panel should be used alone to infer wrongdoing, identity, causation, or a profitable trade. When a field is missing, Polytrack treats it as unavailable context rather than evidence that the underlying activity did not exist.
5. Changes are dated and open to correction.
Public explanations carry a reviewed or modified date. Material changes to classifier inputs, data sources, calculation scope, or public claims require a new date and a content review. Research notes cite the relevant public documentation and state their limitations.
Questions about a definition or possible correction can be sent to polytrack.cash@gmail.com. Include the page URL, the field or sentence at issue, and a source that makes the discrepancy reproducible.
Sources and provenance
Follow the definitions to their source.
External links point to primary documentation. Internal links identify the Polytrack page that defines the visible claim or boundary.
Frequently asked questions
The short version of the boundary.
Is severity a probability that a trade will win?
No. Severity prioritizes records for research based on configured activity signals. It is not an outcome probability, trader-quality score, or recommendation.
Does HIGH severity mean insider trading?
No. A HIGH label does not establish identity, intent, private information, coordination, illegality, or future market direction.
Why are exact classifier thresholds not listed?
They are operational product settings within the public-data boundary that still requires owner approval. Polytrack publishes the signal categories and interpretation limits now and will date any later threshold disclosure.
Are analytics totals exchange-wide?
No. Analytics and rankings describe records available in Polytrack's monitored dataset for the selected window.
Use the definitions beside the product.
Open the public tools guide or continue into the monitored dashboard.