The 800% APR that became −3%: the annualization trap on Hyperliquid's vault page
Hyperliquid's vault page annualizes a short recent window into a headline APR — 800%, even 2,000%. With three weeks of our own snapshots, here's the same number weeks later: −3%, −44%. Why it's a trap, and what to read instead.
Open the Hyperliquid vaults page, sort by APR, and you're greeted by a wall of triple- and quadruple-digit annual returns. It is the single most-looked-at number in the whole vault ecosystem. It is also the one most likely to cost you money — and I can prove it with three weeks of our own daily snapshots.
This isn't a knock on the platform for showing a number. It's a warning about which number, and what it quietly does to your judgment.
What that APR actually is
The headline APR takes a short, recent window of a vault's performance and annualizes it — projects it forward as if that window repeats, over and over, for a full year. A vault that had a couple of good weeks gets stamped with 500%, 800%, sometimes 2,000% "APR." It looks like a track record stretching into the future. It's a snapshot of a moment, dressed up as a year.
Here's what happens to that costume when the moment passes.
| Vault | Page showed | Weeks later | |
| Deposit. Forget. Wake up richer | 818% | → | −3% |
| Enjoyooor V3 | 738% | → | −44% |
| Opportunistic Fund 1 | 775% | → | −4% |
| SG trading | 2,330% | → | 433% |
| Tera Liquid | 423% | → | −16% |
These are real vaults, pulled straight from the daily snapshots we've been keeping — the same public metric the official page displays, captured weeks apart. Enjoyooor V3 went from a projected 738% a year to −44%. Opportunistic Fund 1: 775% to −4%. SG trading's headline was 2,330% before settling near 433%. Same vaults. Same number. A few weeks of daylight between the promise and the reality.
Why it's a trap, in layers
- You're extrapolating from almost nothing. A short window is barely more than a single data point. Annualizing it assumes that one lucky month repeats twelve times.
- Compounding amplifies the noise. The math is roughly (1 + r)12. A small, fluky monthly gain gets multiplied into a headline that looks like genius.
- It captures one market mood, not all-weather skill. A three-week window sits inside a single regime — a rally, a chop, a squeeze. The APR describes that mood. When the mood flips, "800% a year" becomes a loss, exactly as the table shows.
- It hides the risk entirely. APR is one number; it says nothing about the path — the leverage, the drawdown, the fat left tail. A return figure with no risk figure beside it is the dangerous half of the story. Being up on paper is not the same as having made money.
- Young vaults have no history to annualize. Much of the leaderboard is weeks old. You're projecting a year from a fortnight — and you can't see a drawdown that hasn't happened yet.
The deeper harm: it's a momentum-chasing machine
Here's the part that goes beyond "short samples are noisy." Sorting a leaderboard by annualized short-window APR doesn't just show you noise — it systematically surfaces whoever just got hot. And in markets, the freshly-hot cohort is precisely the one most likely to cool off. Short-term winners tend to mean-revert. So the default view of the vault page is, functionally, a machine that points you at the trades most likely to disappoint next — it trains you to buy tops. The top of that list is closer to a sell list than a buy list.
This is the same dynamic that made CEX copy-trading leaderboards so toxic: rank by a recency-biased number, and you manufacture stars who are really just the temporarily lucky.
What to look at instead
The fix isn't to distrust every high number — it's to stop reading a single recent-window figure as if it were a verdict. Concretely:
- The longest window you can get, not the last one. Judge a vault on its whole curve, through good and bad regimes — not on the month that flatters it.
- Risk-adjusted, not raw. Return per unit of drawdown (Calmar), consistency (share of positive periods), and the worst window it actually lived through. We explain each of these in plain language.
- Against the graveyard. Two of three vaults are already dead; the survivors' numbers only mean something when you remember what you're not seeing.
- An honest board, not the default sort. This is exactly why we built a leaderboard that doesn't rank by the trap — it computes receipts from the full curve and flags the artifacts. And it's why the metric that matters was never the biggest headline number.
The annualized APR isn't a lie. It's worse than a lie — it's a true number that means something completely different from what it looks like. It tells you what just happened, amplified twelvefold, and dressed as what will happen next. Read it as a short-term signal being shouted at long-term volume, and you'll stop letting it pick your vaults. Read the whole picture instead.