An open log of who’s actually winning, how we score them, and how often last quarter’s leaders are gone by next.
| # | Alias / Wallet | 7D ROI | 30D ROI | Vol | Trades | Last Trade | Last Seen |
|---|---|---|---|---|---|---|---|
| 01 | mintblade0x96cf…1acb | — | +52% | $0 | 166 | 60d ago | |
| 02 | fishalive0xed64…217d | — | +68% | $0 | 39 | 60d ago | |
| 03 | endlessFate0x5e4c…dba2 | — | +28% | $0 | 138 | 60d ago | |
| 04 | frostrizz0xbc11…eec6 | — | +39% | $0 | 92 | 60d ago | |
| 05 | GRIMDRIP0x3f87…742c | — | +56% | $0 | 93 | 60d ago | |
| 06 | supersob0x9d36…0e68 | — | +31% | $0 | 36 | 60d ago | |
| 07 | 0x8139E8F571eb4372ecc8498771f6a319Cb5bBe43-17678284671160x89dd…f662 | +170% | +170% | $973.6K | 197 | 60d ago | |
| 08 | 0x5966Db1fE50763C9e3C014d756369BAd07E1F804-17776485342410x5966…f804 | +138% | +14% | $986.6K | 518 | 60d ago | |
| 09 | NiNo9990xcd65…ee96 | +95% | +95% | $1.31M | 2,013 | 60d ago | |
| 10 | YolYo7890x97cb…e4d0 | — | — | $0 | 27 | 60d ago |
Historical ranking only. Past performance does not predict future results — see methodology for the regression-to-mean evidence.
Of 1707 ranked wallets with ingested trades, here’s how many actually traded inside each window. The hero question gets a number.
Every prediction-market product we looked at — Polymarket Analytics, Polywhaler, Nansen, eToro, Public — hides the same thing behind a signup wall: who is actually winning, by how much, and on what time horizon. The leaderboard is the headline feature, gated. We think that’s backwards. If your product is “follow the smart money,” the smart money should be the page.
The research that motivated this log keeps surfacing the same finding. A recent CoinDesk study on prediction-market accuracy reports that roughly 3% of traders drive price discovery; about 12% of frequent participants show measurable skill above noise; and about 60% of last year’s top performers regress out of the top decile within twelve months. Those three numbers should sit above any leaderboard, not behind it.
So this site is the leaderboard, the methodology, and the persistence test, in that order. We rank by the Wilson lower-bound interval rather than raw ROI so a 3-trade lucky streak doesn’t crowd out a 200-trade modest edge. We sample every 15 minutes from the public Polymarket API. We will publish the persistence-decay number quarterly, even when it cuts against the “follow the leaders” pitch — especially then.
Most recent ingested trades from ranked wallets. Refreshes with the 30-minute trade-ingest cadence; not real-time.
Snapshots come from Polymarket’s public leaderboard API (lb-api.polymarket.com). We poll on a 15-minute interval across four timeframes — 1d, 7d, 30d, and all — and persist the top n=50 wallets per timeframe per snapshot. No private data, no off-platform positions, no access beyond what any user can see in a browser. Every row on the leaderboard above is reproducible by hitting the same endpoint.
Raw ROI rewards small samples — three lucky trades in a thin market can outscore two hundred careful ones. We rank by the lower bound of the Wilson score interval at α=0.05, which behaves correctly at small n and at extreme proportions. The implementation lives in worker/src/insiderbet/analytics/significance.py; it’s a 40-line module, the conservative one-sided variant, and we’d rather under-claim a winning trader than over-claim one. If you see a wallet near the top with n < 30, the lower bound is doing most of the work to keep them honest.
The honest version of “follow the leaders” is a persistence test: of the top-decile wallets at quarter t, what fraction is still in the top decile at quarter t+1? The literature says roughly 40% — the rest regress. We have not yet computed our own version of this number; v0 is a collector, not a longitudinal study. The first quarterly persistence verdict ships with v1 and will be linked here whether or not it flatters the thesis.
InsiderBet is not investment advice. It is not a copy-trading product. It cannot see private trades, off-platform positions, or wallets that Polymarket excludes from the public leaderboard. Past performance does not predict future results — that is the whole reason we run the persistence test. Prediction-market participation is illegal or restricted in some jurisdictions; nothing here is a recommendation to trade.