InwyResearch

Studies · 13 of 13

Copy-trading

Every Hyperliquid position is public. Do the traders at the top of the leaderboard stay there long enough to copy?

Rejected
The question

Does past performance of individual traders persist?

The answer

Rank correlation +0.075; the best fifth does not outperform

Data4,000 leaderboard accounts incl. losers and blown-up accounts
TestQuarterly formation, next-90-day medians, without the best pick
The details

On Hyperliquid, unlike centralised exchanges, every account’s trades and positions are public. That makes a copy-trading strategy easy to build — if past winners keep winning.

Market makers and high-frequency accounts were excluded as uncopyable. Averages proved unusable — individual traders move money in and out erratically and their history-based returns spike — so only medians and rankings are reported.

Median return over the next 90 days, by past-performance fifthRanked by past return per unit of drawdown; 384 eligible traders per month on average
Worst fifth−2.9%
Second+0.0%
Middle+7.2%
Fourth+6.8%
Best fifth+5.0%
Past-performance group Next 90 days, median Lost more than 30%
Worst fifth −2.9% 30%
Middle fifth +7.2% 20%
Best fifth — the ones you would copy +5.0% 19%

Following the top ten each quarter: without each quarter’s single best pick, 100 became 58. With no sign of persistent skill, and copying costs still to subtract, a full trade-level study was not worth running.