@CleansedTweets As an American, I find that there are two types of British accents: (1) those I understand perfectly, and (2) those that are hard to follow, as if they're speaking a foreign language. You're definitely in category 1, like the Queen's English.
15/15
Caveats:
5-month regime. Correlations rise sharply in stress; diversification often fails when you need it most.
Hyperliquid API snapshots understate true MDD.
Calmar is window-sensitive; don't rank vaults on it.
I plan to refresh as data accrues.
14/15
Reality check:
21 weeks is enough for point estimates. It is NOT enough for sizing decisions.
The 95% confidence interval on a sample r of 0.40 at n=21 spans (-0.05, 0.71). The negative correlations could plausibly be zero once more data comes in.
Treat as directional only.
7/15
Sharpe = return per unit of total volatility. Sortino is the same idea, but only penalizes downside moves. Both annualized. Note: on a same-window basis, the spread is much tighter than Growi's full-history table showed.
1/15
There's almost no neutral, third-party performance analysis of @HyperliquidX vaults.
@GrowiFinance just published a vault analysis: https://t.co/lD78uCAkSb
The stats are great, but the ranking has a methodology issue. I re-ran the numbers on a same-window basis. Thread 👇
We just published a deep dive into Hyperliquid vaults after two major market shocks.👇
Key takeaways:
• 85% of serious vaults are now dead or abandoned
• User vaults are net -$82M overall
• Survivors took the hit, recovered, and kept compounding
• Growi HF ranks #1 by Sharpe, with a 2.55 Sharpe ratio.
The lesson: Don’t chase APR. Look at survival, volatility and risk-adjusted returns.
Read the full article ↓
5/15
Methodology note: I compute per-bar return as Δ PnL / prior equity. This isolates strategy performance from depositor flows.
Same idea as Time-Weighted Return that Growi uses. Account value alone is contaminated by deposits/ withdrawals; this strips that out.
3/15
This matters because the Oct 10 '25 and Feb 5 '26 stress events fall inside some windows and outside others.
Growi's 670-day window contains both shocks. Orbit's 145-day window contains only Feb 5. Several short-history vaults (drkmttr, Ultron) avoid Oct 10 by not existing.
2/15
Growi's Sharpe ranking compares vaults with wildly different track records.
Growi: 670 days. Orbit: 145. Systemic HyperGrowth: 250. HLP: 1097.
Each vault's Sharpe is computed over its own full history. Different windows = different market regimes = not directly comparable.
@nayibbukele@NiallHarbison He's the perfect go-to guy on this issue. We need to do better at treating stray dogs around the world. Thank you for your leadership, President Bukele!