The copy-trading test

Does copying whales work?

Updated , from weekly run 1791223421

No. We picked the best whales from earlier weeks and copied their next bets, entries and exits, at the prices a follower could actually get.

  • Polymarket: copiers made −7.0% over 2,053 positions. The whales they copied made −4.5%.
  • Limitless: copiers made −28.5% over 538 positions. The whales they copied made −9.2%.

How we tested it

  1. Pick the whales the way a follower would. We ranked whales by their return in the earlier two-week periods (Polymarket: Aug 20 – Sep 20 and Limitless: Sep 5 – Sep 20) and took the top 20% of those with enough bets to rank: 249 Polymarket wallets and 38 Limitless wallets.
  2. Copy their next moves. In the period after that (Polymarket: Sep 20 – Oct 5 and Limitless: Sep 20 – Oct 5), every time one of them bought, the copier bought the same number of shares off the order book as it stood moments later (a median 3.9 minutes on Polymarket, 12 seconds on Limitless). Every time they sold, the copier sold what it held into the bids.
  3. Score both on the same trades. Anything still held when the market resolved is marked at the result. Return is profit divided by the money put in.

This is an out-of-sample test: the whales were picked before the period they were copied in, which is the only way a real follower can pick them.

Results

Polymarket

Who was copiedPositionsWhales madeCopiers made
Top 20% from earlier weeks2,053 −4.5% −7.0%
Whales that beat the odds earlier787 −7.3% −13.4%
Every whale, same period24,490 −3.9% −6.4%
Every whale, every period (not out of sample)54,567 −2.7% −5.3%

Limitless

Who was copiedPositionsWhales madeCopiers made
Top 20% from earlier weeks538 −9.2% −28.5%
Whales that beat the odds earlier59 −4.4% −66.8%
Every whale, same period3,294 −6.0% −53.0%
Every whale, every period (not out of sample)14,991 −9.8% −58.7%

"Whales that beat the odds earlier" won more often than their prices implied in the earlier periods (see how we test skill). Return = profit ÷ money put in, entries and exits included.

Why copiers do worse than the whales they copy

They arrive late. By the time a follower sees a whale's bet, the price has moved against them. Buying the same bets moments later cost 1.4¢ more per $1 share on Polymarket and 26.0¢ more per $1 share on Limitless. Thin order books make it worse: there isn't always enough left at the whale's price to fill a copy.

Even when whales do make money, the copier doesn't keep it. On Limitless, the whales that beat the odds earlier bought well in the copy period: held to the result, their buys returned +32.8%. The same buys made moments later, off the order book, returned −19.6% (319 buys).

Don't winning whales keep winning?

Not reliably. Here is how often a whale in the top 20% for one two-week period was still in the top 20% for the next:

VenueFrom → toStayed in the top 20%By chance aloneWhales
PolymarketAug 20 – Sep 4 → Sep 4 – Sep 20 36%20% 698
PolymarketSep 4 – Sep 20 → Sep 20 – Oct 5 22%20% 665
LimitlessSep 5 – Sep 20 → Sep 20 – Oct 5 12%19% 129

Some periods show a small cohort that stays on top. It hasn't been enough to make copying them pay: the results above already include the top 20%.

What this doesn't prove

  • It covers one copy period per venue so far. The whole test reruns with every weekly update, as new weeks come in.
  • A follower faster than our recorder would lose less to arriving late. They would still be copying whales who, in the copy period, made −4.5% on Polymarket and −9.2% on Limitless.
  • On Polymarket we only record trades of $1,000+, so a whale's smaller trades aren't copied or counted.
  • The latest period has the most markets still waiting to resolve, so its numbers move the most between updates.

Use these numbers

Free to quote and reuse with a link back (CC BY 4.0). Cite as: hopium.games, “Does copying whales work?”, updated Oct 5, 2026, https://hopium.games/copy-trading. How every number is made: methodology.

About this data · weekly run 1791223421, computed 2026-10-05 18:03 UTC
Source: hopium_compute run 1791223421 (PG metrics_*)
  • Poly: only >=$1k trades captured -> per-wallet PnL partial
  • Limitless: price/contracts only since 2026-09-05
  • latest window has the most unresolved markets
  • Limitless wash/value-transfer trades excluded (see wash); Poly off-book is report-only
  • fair value covers Limitless crypto markets only (Binance); stocks have no reference feed