PMKT SURVEILLANCE← Market reviewMETHODOLOGY

Reading guide

What the alerts mean

The tool groups stored Polymarket trades by market and highlights patterns for human review. It does not identify insiders, prove coordination, or estimate the probability of misconduct.

Choose a market →

Annotated market review

This screenshot uses a saved sample alert. The numbers identify what to examine in the interface; it is an illustration, not a live market finding.

Market review interface showing the selected market, observed trade metrics, alert list, and alert details.1234
Saved sample data, captured from the tool. Labels correspond to the reading steps below.
  1. Market selection: confirm the exact question, condition ID, and scope classification before reading signals.
  2. Coverage: read the feed freshness and coverage window. Counts describe stored activity; a stale or incomplete feed can hide activity.
  3. Alerts: a saved score ranks rule matches. Opening an alert reveals the actual reasons.
  4. Context: inspect wallet history and trade details, then compare with public information and ordinary strategies.

Detection method

Market eligibility

Sports and election-result markets are excluded. Ambiguous political decisions and questions without an explicit eligible rule require operator review. Eligible questions concern a discrete decision, announcement, or information release. Scope uses event tags and question rules; a recorded operator override can change it. An excluded or unreviewed market produces no insider review alerts. The scanner checks a bounded selection of eligible markets for both individual trade alerts and synchronized trading patterns. Sports, election-result, and unreviewed markets are outside that selection unless their scope changes. The recent review leads list shows a small set of saved results, not complete market coverage.

Individual trade score

Rules add up to three raw points for unusual value, two for concentration, two for thin observed flow, and one for repeated entries. The eight possible raw points are scaled to 0–10. Wallet age contributes no points. The threshold is configurable; the score ranks rule matches and does not estimate the probability of misconduct. Each short insider explanation paraphrases saved rule reasons and states this limit. Older saved scores are adjusted once to remove the former scanner-age rule and use the new scale.

Synchronized trading patterns

At least three distinct wallets must take the same outcome and direction within a rolling two-minute window, with at least $1,000 combined value by default. Duplicate observations are removed and fills from one wallet in the same transaction are grouped. With at least ten other observed actions spanning at least thirty minutes, a burst must have at least three times the baseline action rate. Otherwise it is explicitly labeled as having an insufficient observed baseline, and timing rarity is not estimated.

Repeated wallet pairs are compared with retained patterns from the prior thirty days, using episodes separated by at least thirty minutes. Direct token transfers and common senders are shown only from cached explorer records dated at or before the detected burst; common senders can be exchanges or services and are weak context. Each short pattern explanation reports the stored wallet count, action count, direction, time span, and value when available. These observations never establish beneficial ownership or confirmed coordination.

Wallet evidence

First public Polymarket activity uses the earliest returned full-history record, including deposits and withdrawals. Verified contract deployment is a separate Polygon transaction. Neither establishes that the owner is a new trader. EOA private-key creation cannot be dated from public records. Failed checks and missing records stay unknown; evidence shows when it was checked.

Common false positives

New account, experienced owner

A newly deployed account contract can belong to an established trader. An old address may only recently start trading on Polymarket. Scanner first-seen time adds no score.

Large or concentrated trade

Hedging, rebalancing, market making, or topic specialization can look unusual. Sparse collection can distort the comparison.

Thin observed flow

Stored trade value is not market liquidity. Missing or delayed ingestion can exaggerate this signal.

Shared timing or funder

News, bots reacting to public signals, exchange withdrawals, and routine strategies can produce aligned trades or a common sender. Check alternatives before inferring a relationship.

How to interpret an alert

  1. Read the reason strings. They show which rules contributed. Do not treat the score as a probability.
  2. Check the timeline. Compare the trade timestamp with public news and market price movement. The current UI does not verify that chronology for you.
  3. Check the baseline. Compare trade size with observed flow and other activity in the same market. The visible recent trades are a limited sample.
  4. Look for alternatives. Consider hedging, arbitrage, market making, and repeated routine trading.
  5. Record uncertainty. Note missing coverage and the evidence needed before drawing a conclusion. A high score means “review first,” not “confirmed.”

Coverage and limits

By default, discovery looks at up to 300 open events per refresh and scans up to five selected eligible markets per pass. Selection favors higher-volume eligible markets within the configured range. These caps and polling intervals can be changed by the operator; the dashboard cannot infer unseen market activity. The worker refreshes discovery about once an hour and rotates its five-market scan batch through the cached eligible set between refreshes. A scan pass covers only its reported batch; the candidate set can change at each refresh.

Trade collection uses Polymarket Data API v2 cursor pages, locally bounded to the collector’s checkpoint. Public market feeds serve taker-side rows once per fill; this is not a complete view of every participant or private order. Identical returned trade records are deduplicated, so fills without distinguishable identities can be collapsed. The visible trade cards show only the latest thirty stored trades in the last twenty-four hours. Saved alerts use the server’s configured alert window.

The source freshness check is cached and may itself be old; compare its check time and reported lag with the trade timeline. The market view examines funding context for up to twenty stored member addresses per displayed pattern and only the latest one hundred explorer token-transfer rows per checked wallet, not a complete funding history. Thirty-day repeat evidence covers patterns observed by this installation, not all historical relationships. No record or no alert does not clear a market.

The tool does not verify private information, intent, beneficial ownership, or all cross-market positions. A substantive claim requires source records, a public-information timeline, and independent review.

Sources: Data API migration · Wallet activity · Contract deployment records