Reading the result
For the first ten minutes a late copy is a coin toss on price. Thirty seconds after the whale, 40.6% of entries were within half a cent of the whale's price, 29.9% were worse and 29.5% were better. The moves cancel, which is why the return sits between +7.6% and +7.7% through ten minutes.
After that the market starts pricing in the result. An hour after the whale bought, the price was up 13.7 cents on bets that went on to win and down 10.9 cents on bets that lost. A late copier buys the winners dearer and the losers cheaper, and since most whale bets here won, the return falls from +7.6% to -2.5%. Sports carries most of it: +10.3% at the whale's price and -2.4% an hour late across 4,737 settled entries.
The practical reading is narrow. Speed matters less than the popular idea of copy trading suggests, but it is not free past ten minutes, and it matters most for the traders with a real edge, because they have the most to give up. Every name above links to a public record with the same flat-copy simulation.
Population and method
- Entries. Every whale-sized buy on the Polymarket trade tape from August 23, 2026 to August 25, 2026 (UTC), deduped on transaction, outcome, price and shares, with fills by one wallet on one outcome in one second grouped into one entry at the volume-weighted price: 15,976 entries.
- Late price. The average of the first buys of the same outcome by anyone at or after the delay, searched for 5 minutes.
- Scoring. $100 per entry at the price in question, held to settlement. A win returns 100 / price − 100; a loss returns −100. Sells are never scored.
- Scored set. 14,240 entries settled win or loss, 87 were void and 1,649 were unresolved at export. Only the 6,005 entries with a trade after every delay are scored.
- Trader floor. A trader needs 30 scored entries to be ranked.
- Not modelled. Fees, depth beyond the next trades, partial fills, and correlation between entries in the same market.