Polyloly is a real-time Polymarket whale tracker. Every whale trade above $1,000 hits our live trade alert stream within seconds of landing on-chain, and our detection engine flags volume anomalies, coordinated wallet clusters, and profitable insider wallets so you can read the prediction market the way institutional desks do. Free analytics dashboard, no signup.
Live stream of every Polymarket trade ≥ $1,000. Pagination to 1,000 trades, filter by minimum size, auto-refresh every 5 seconds, and a prominent whale alert when positions ≥ $50k land. Live feed →
Markets where the last 15 minutes of trade volume exceeded the 24-hour rolling baseline by 3× or more. The earliest tell for breaking-news repricing, pre-resolution whale action, or a thin-market sweep. Unusual volume →
Louvain community detection over the last 7 days of whale trades identifies wallet groups that co-trade the same markets within 5-minute windows — almost always one operator behind multiple accounts. Cluster graph →
Live feed of the markets Polymarket wallets with ≥75% historical win rate and sub-$0.60 average entry price have just bought into. Free signal, no signup, updated every minute. Insider picks →
ROI grouped by entry timing × market category × bet size. Shows the cohorts that historically generate positive return (contrarian geopolitical) and the ones that reliably bleed (early-entry crypto). Cohort heatmap →
Ranking of Polymarket wallets by how much the price moves in their favor in the 15 minutes after they enter a position — the clearest latency-edge / news-speed signal in on-chain data. Speed traders →
Claude Haiku 4.5 summaries of why each flagged market is moving — synthesized from current odds, 24-hour whale flow, price trajectory, and top recent positions. Cached 6h, regenerated on anomaly hits.
Per-category rankings (political, crypto, sports, geopolitical, weather) of Polymarket traders by net P&L on resolved positions, with WR and average entry shown. Separates predictors from endgame scalpers. WR leaderboard →
A Node.js WebSocket client connects to Polymarket's live trade feed and writes every trade ≥ $1,000 into our Neon Postgres database within seconds. A watchdog force-reconnects after 3 minutes of silence so stale feeds never mask real activity.
Every 15 minutes a cron rebuilds the coordinated wallet cluster graph via Louvain community detection, computes volume anomaly z-scores, and refreshes insider-watchlist win-rate calculations. All results cache in Postgres and stale-while-revalidate at the edge.
Flagged markets trigger an LLM call via Vercel AI Gateway. Claude Haiku reads the market question, current odds, recent whale flow, and 24-hour price trajectory, then writes a two-sentence explanation of what the signals say. Cached 6 hours per market.
The single-page dashboard pulls the processed feed, cluster snapshots, insider picks, and anomaly banners via read-only APIs. Auto-refreshes every 5 seconds for the trade feed, 45 seconds for anomalies, 2 minutes for cluster membership.
Any single trade with a notional value of $1,000 or more. Every such trade enters our database within seconds of landing on the Polymarket CLOB. Trades below $1,000 are ignored — they are noise for the questions Polyloly answers.
We build a weighted co-trading graph: edges connect two wallets that bought the same market on the same outcome within a 5-minute window. Pairs are kept only if they meet the threshold on at least three distinct markets and $10,000 combined value. Louvain community detection then partitions the graph. Full method in our clustering blog post.
Yes. All live data, cluster graphs, insider picks, cohort analyses, and the blog are free with no signup. If you want to support development you can leave a tip through the heart widget in the header — not required, never gated.
No. Polyloly is an independent analytics service built on Polymarket's public API and on-chain trade data. We are not a broker, counterparty, or gambling operator. Links from Polyloly to the Polymarket app carry our referral code so we earn a small commission if you sign up there, which funds the analytics.
Trade feed streams in real time, with the dashboard pulling every 5 seconds. Volume anomalies recompute every 45 seconds. Cluster snapshots rebuild every 15 minutes. AI market context refreshes every 6 hours per market or immediately on anomaly hits.
Yes. We expose read-only JSON endpoints at /api/trades, /api/top-trades, /api/anomalies, /api/clusters, /api/insider-picks, and more. Cache headers are tuned per endpoint. Please be considerate with request volume; if you need high-rate access reach out.
This page visualises Polymarket whale clusters — groups of wallets that behave as a single coordinated operator. Multiple accounts trading the same markets on the same outcome within 5-minute windows across at least three distinct markets. Louvain community detection runs every 15 minutes on the last 7 days of whale trades and surfaces the resulting clusters with their combined volume and member lists.
We build a weighted co-trading graph — nodes are wallets, edges connect pairs that bought the same outcome within 5 minutes of each other, weighted by combined dollar value. Pairs with fewer than 3 shared markets or under $10,000 of combined value are dropped. Louvain community detection then partitions the graph by modularity optimization, which cleanly separates true syndicates from incidental co-movers on popular markets.
Each coloured island represents one cluster. Node size scales logarithmically with the wallet's 7-day whale volume. Solid edges within a cluster mean the members share 3+ markets; faded grey edges are cross-cluster co-occurrences that didn't meet the threshold. Click any node to isolate its cluster and see every member wallet sorted by 7-day value.
Current top cluster in our snapshot: 36 wallets, $7.37 million combined 7-day volume, 729 distinct markets touched. That is institutional-scale coordination — either one desk running accounts across jurisdictions, or a formalised syndicate. The behaviour pattern is consistent: same markets, same side, nearly simultaneous entries, large tickets.
Prediction markets work because they aggregate independent beliefs. When a meaningful fraction of volume comes from coordinated wallets, the "market consensus" you see is actually a few operators' beliefs repeated. Our cluster view lets you net that out. Read the full methodology post for the algorithm details.