Copying a profitable wallet sounds obvious. The published evidence, the latency arithmetic and the holding-time data all point the other way.
Copy trading is sold as the shortcut: find a wallet that wins, mirror it, collect. The evidence for it is thinner than the marketing suggests, and the reason is arithmetic rather than opinion.
On free infrastructure, detection takes roughly one to two and a half seconds, and landing a transaction takes another one and a half to two. Paid, co-located setups measure this in tens of milliseconds. The gap is structural: it is where you sit, not how good your code is.
Median holding time for Solana pool tokens fell to roughly 60 seconds in early 2026. If a wallet holds for a minute and you are four seconds behind, you enter after a meaningful share of its hold has already happened and you exit after it has already sold. The wallets with the best win rates are often the worst to copy, because a seven-second median hold belongs to a bot you cannot follow.
A simulation of copying one of the most-followed traders with a ten-second delay and a ten-minute exit returned minus 21.3 percent. No study has been published showing net positive returns for copiers. The academic work that touches the subject documents the opposite mechanism: the followed wallet sells into the demand its own followers create.
Daily leaderboards rank by realised profit, which is dominated by variance and by size. On one day we sampled, several wallets marketed as profitable were down over the trailing month: minus 60,000 dollars, minus 57,000, minus 12,000. A wallet whose entire month is explained by three trades will still appear near the top.
Those are the filters our leaderboard applies, which is why wallets with enormous profits and seven-second holds score badly on it.