What every number on 0xinsider is made from: provider facts, stored projections, derived metrics, and authored examples, each with its source and completeness limits. Every metric describes what already happened, and none predicts what happens next.
On-chain and provider records feed position reconstruction, split/merge/redeem handling, and daily mark-to-market P&L for tracked wallets. When source history is missing or incomplete, we say so instead of filling the gap synthetically.
Each record moves through normalization, reconstruction, scoring, and grading. Missing or unavailable fields fail closed: every output that depends on a missing field stays empty instead of reporting a stage that did not run.
Pull trades from Polymarket on-chain and provider records. Normalize the records into a single analytics schema so everything downstream reads from one shape.
Raw fills are not positions. Match entries to exits, track splits and merges, handle resolutions, and compute daily mark-to-market P&L per wallet per market.
Headline P&L is Polymarket's own total, not a local re-derivation. Realized cash legs can be reconciled to the available on-chain BUY, SELL, SPLIT, MERGE, REDEEM, and fee events. Open positions are marked at the provider's price, not built from transfer events. Missing source history, and any difference between those figures, stays visible.
Organize the available trading history into wallet profiles, with positions, resolved results and activity over time.
The trading-style model describes stored activity across markets and categories. Mixed patterns and limited history remain explicit; an unprocessed wallet can have no label. Separate models provide performance and anomaly context.
Show wallet metrics, historical grades from S through F, strategy labels where available, and a 0.0 to 1.0 review score for large trades.
Assign a canonical category when the provider's category, the event's league or series, or the market's tags resolve one. Leagues fold into their sport, so NBA and WNBA markets are Basketball. The 14 categories with their own page are Crypto, Politics, Basketball, Football, Soccer, Baseball, Hockey, MMA, Tennis, Cricket, Esports, Pop Culture, Science & Tech, and Geopolitics. A resolved category outside that list, such as Golf or Racing, keeps its name without a page. A market whose metadata resolves no category shows none; it is never guessed from the title.
Graded Polymarket wallets carry a letter from S to F, with S the strongest rating. Read the letter beside the wallet's profit, resolved-market history and risk metrics. A grade describes historical performance; it does not predict the next trade.
Net realized P&L records closed and settled positions, with fees and rebates included. Read it separately from open-position gains, which can change before a position closes.
A high win rate alone does not establish profitable forecasting. Read results alongside the prices paid and the size of the losses.
The resolved-market count puts the results in context. A long record gives you more observations to study; a short one leaves more uncertainty.
The share of resolved markets won. Compare it with the entry prices, profit and sample size before drawing a conclusion.
Drawdown shows the equity lost from a previous peak. The equity curve shows whether returns accumulated steadily or came from a few large moves.
A grade summarizes historical performance. Inspect the underlying record and current positions before acting; a letter cannot guarantee the next result.
Read a grade beside the underlying record. A short history leaves more uncertainty, and gains on open positions can change before they close. Check the trader's results and risk over time. How sample size changes confidence →
The strongest rating. Open the profile to inspect its profit, history and risk.
A strong rating. Read the underlying record alongside the letter.
An above-average rating. Compare the profile's results and risk over time.
A middle rating. The profile gives the context behind the letter.
A below-average rating. Check the results and the size of the record.
The lowest rating. Review the full record before drawing conclusions.
Each large trade gets a review score. A higher score means look at this trade first.
A 0 to 100 suspicion score ranks unusual trades in the feed. A flag is not proof of inside information or misconduct.
Trading styles describe both-outcome activity, recorded trades per stored market, and category shares of recorded positive USD cost basis. Performance remains separate.
Trading styles describe the recorded activity rather than a wallet's intent. A wallet can qualify for several patterns; the model preserves those traits and uses Mixed activity when none clearly leads. Grades, profit, and risk metrics remain separate. Read the sample, thresholds, and historical-label transition on the trading styles page.
Updated hourly. Counts as of Oct 16, 4:37am ET. The position count is an estimate.
A new trade starts a full sync within minutes, and a wallet idle for a week gets a catch-up sync. Median time since the last sync on Oct 7, 2026: 3.2 hours for S, 8.2 hours for A.
The same trade trigger and weekly catch-up, but the catch-up runs on spare capacity and is behind. On Oct 7, 2026 the median was 172 hours, and 3,721 of 6,960 B wallets were more than a week old.
A wallet someone follows has its open positions refreshed on a rotation of about an hour and its full history about every 6 hours. Median ages on Oct 7, 2026: open positions 26 minutes; full history 6.1 hours.
No fixed schedule. Opening a wallet's profile pulls a sync. On Oct 7, 2026, C, D, and F wallets had a median of 24 days since their last sync.
Each entry names something the data cannot show, and why. A methodology that states its scope is the only kind you can check.
We track wallets selected for activity and performance. Traders who blew up and stopped trading are underrepresented. Aggregate metrics skew upward; the real prediction market success rate is lower than this dataset suggests.
Models retrain periodically. We have not published accuracy trends because the validation pipeline that would defend them is still being built. The numbers go on this page once they show predictive skill against a naive baseline.
Losing shares on Polymarket never receive REDEEM events. Our pipeline zeros the losing side once the winning outcome is known, but markets with delayed or disputed resolutions may show temporarily inaccurate P&L.
Every metric, grade, and score describes what already happened. Market regimes shift, strategies decay, and past patterns are not guaranteed to repeat. We show what a trader did, not what they will do next.
The traders we cover were selected for having a meaningful history. This is not a random sample of all participants. Conclusions describe the active, established cohort, not the market as a whole.
A flag marks a statistical anomaly, not confirmed insider trading. A high suspicion score means the pattern is unusual, not that the trader had inside information. Some flagged trades are well-timed trades on public analysis.
We ingest trades from Polymarket and rebuild each tracked Polymarket wallet from its history: positions, P&L, grade, strategy, anomaly context. If a wallet's history is too thin for a metric, the field stays empty. Coverage and refresh timing vary by wallet and page.
Each graded wallet gets a letter from S (strongest) to F (lowest). The grade describes historical performance. Read it beside realized profit, open positions, the resolved-market count and risk metrics. A short record leaves more uncertainty, and a grade does not predict the next trade.
We watch for large trades across Polymarket and enrich each one with trader, market, spread, and anomaly context. The 0.0 to 1.0 review score says which trade to look at first. It is not a prediction, and we do not claim to have every exchange event.
0xinsider's suspicion score helps prioritize unusual trades for review. A higher score indicates a stronger anomaly flag; it does not establish what a trader knew. The 0-to-100 suspicion score is an anomaly ranking, not proof of insider trading or an outcome prediction.
0xinsider tracks Polymarket traders and large trades using on-chain and provider data. Coverage depends on the available history for each wallet and market.
We publish every known limitation on the methodology page: survivorship bias, model accuracy versioning still in progress, backward-looking analysis only, non-random sample, resolution-dependent P&L, and suspicious-trade false positives.
