Copying whale bets in real time
The founding strategy: watch large bets hit the tape, copy them within seconds at small size. On real money it made +2.1% — and all of it came in one World Cup month.
The hypothesisLarge bets carry information. If a five-figure order hits a market, the bettor probably knows something — copy it fast enough, at small size, and you ride their edge.
The test
A daemon watches the public trade feed (plus a WebSocket trigger that cuts detection latency from ~46s to under 1s), and mirrors qualifying large bets with real orders at the live ask, recording the whale's price, our fill, and the slippage between them. Every copy settles at market resolution. Years of hype posts describe this strategy; the difference here is that every copy is logged with its executable entry price, so the edge can be located rather than asserted. After real copying stopped, we re-scored the whole whale feed month by month to see whether the band we had been copying held up outside the months we traded it.
The result
The full real-money record: 699 copies settled between late May and the end of July 2026 across our two copy lanes, for +$197.13 on $9,413 staked (+2.1%), after $24.47 in fees. Its timing matters more than its size. Copies placed in June made +$388.06 (+7.9%); copies placed in July lost $170.93 (−3.8%). The non-sports lane finished +$221.73 over 362 (+4.8%), the sports lane −$24.60 over 337 (−0.5%). The worst single day, July 9, cost $315.63.
June 2026 was the World Cup. To check whether the band we copied had an edge outside it, we re-scored every whale buy of $5,000 or more priced 35–45¢ — the main lane's band — by month: May −14.7% (248 bets), June +15.5% (496), July −7.5% (680), August −6.2% (530), September −2.0% (700). One positive month, four negative. Across all whale buys, blind copying came out at roughly zero in every period we tested (−0.2% over 6,600 bets, −0.3% over 7,800), and no whale-size tier or price band was positive in both halves when we split the data by period.
The other version of the idea — copy only the whales with the best records — failed too. Wallets that had returned more than +10% through August returned −4.4% over their next 720 bets in September and October. Past whale skill did not carry forward.
Conclusion
Downgraded from "edge confirmed" to no edge. The real-money profit came from one tournament month; outside it, the band we copied lost, and the wallets that looked skilled stopped winning. Copying also costs a point or more in slippage between the whale's price and ours, which on its own is enough to erase a thin edge. We stopped copying with real money at the end of July 2026. Anyone selling "copy the whales" as a general strategy is selling one good month as a method.