Transparency

Methodology

Every score on this site is published in full. Nothing here is a prediction, and no ranking position is ever for sale.

Qualified Vault Universe

Public rankings and the default screener only contain qualified vaults. A vault qualifies when it is active, at least 30 days old, holds at least $10,000 of TVL, has enough observations to calculate risk statistics, shows no obvious broken data, is open to new deposits, and is a standalone followable vault — not an internal component of another vault. Child vaults and vaults that have disabled deposits are flagged as not_followable and excluded from rankings and the screener. Everything else is flagged as insufficient_history, low_tvl, inactive or data_quality_issue. These thresholds are configurable and will be tightened as our historical record deepens.

HV Score (0–100)

HV Score summarizes historical performance, risk, consistency, manager alignment and data confidence. It is not a prediction of future returns. Components: Performance 35%, Risk 25% (inverted — lower risk scores higher), Consistency 15%, Manager Alignment 7.5%, Liquidity / Capacity 7.5%, Confidence 10%. Short track records are capped: below 90 tracked days the HV Score cannot exceed 60, rising linearly to 75 at 179 days, with no cap from 180 days. Mostly idle vaults — trading in fewer than half of tracked months (a month is active when the sum of absolute daily returns exceeds 1%) — lose up to 15% of their score. Consistency includes earning months (25%): only months with a return above +1% count, so trading without making money is not rewarded. For track records over 270 days, if more than half of total tracked growth came in the last 90 days, Performance is scaled down linearly to ×0.75 (when all of it did), so a recent surge after a long flat stretch is not over-rewarded. Every component is stored separately so the methodology can evolve without rewriting history.

Performance Score

Weighted across 30D (20%), 90D (35%), 180D (30%) and a time-decayed long-term annualized return (15%, half-life 180 days, so old performance fades gradually). Each window passes through a logistic curve calibrated so strong performers still differentiate — exceptional returns do not all saturate at the same score. Returns are time-weighted: deposits and withdrawals move vault equity but not cumulative PnL, so capital flows cannot inflate measured performance. All scoring inputs are computed from our own continuously tracked daily record — the same series shown on the vault profile — so performance earned before our tracking began does not influence scores. When history is missing, weights are redistributed and the Confidence Score falls.

Risk Score (0–100, higher = riskier)

Built from annualized volatility (35%), maximum drawdown (35%), worst day (15%) and worst week (15%) — purely historical characteristics. Current drawdown is deliberately excluded: a vault is not penalized for happening to be underwater today; the current state lives in Vault Health and Copyability instead. Labels: 0–20 Low, 21–40 Moderate, 41–60 Elevated, 61–80 High, 81–100 Extreme. Leverage, position concentration and directional exposure are being added as our position snapshot history accumulates.

Confidence Score

Answers how much evidence exists that a performance profile is meaningful: vault age, number of observations, length of history, drawdown history, and TVL stability. A brand-new vault with spectacular returns will always score lower here than an established one. Labels: High, Medium, Low.

Consistency Score

Measures whether results repeat: share of positive weeks and months, average drawdown depth, and profit concentration. A vault where most lifetime profit came from a single period is penalized, regardless of headline return.

Manager Alignment

Derived from the leader's own stake as a share of vault equity. Larger manager ownership increases alignment. Once our snapshot history is long enough, significant leader withdrawals will also generate warnings.

Copyability

Copyability is separate from HV Score. It answers one question only: how suitable is this vault for a new follower today? A historically excellent vault can be a poor or limited choice to copy right now if it is in a deep drawdown, very small, or temporarily running hot.

It is built from three inputs:

Vault size / capacity — 35%. TVL is log-scaled. $10,000 scores 0, $10,000,000 scores 100, and a $100,000 vault sits around 50. Smaller vaults are penalised because capacity, slippage and sustainability are harder to judge.
Current drawdown — 35%. Drawdowns are not treated as uniformly bad. For proven vaults (high HV Score, Consistency and Confidence), a normal historical drawdown is a neutral-to-slightly-attractive entry point — entering a quality vault on a dip is not riskier than entering at a high. For young or unstable vaults a drawdown remains a penalty (none scores 100, -30% scores 0). A drawdown near the vault's historical maximum is penalized for everyone, as it may signal a regime change rather than a normal dip.
Current risk level — 30%. The inverse of the Risk Score. Lower risk scores higher here. High volatility or a large recent drawdown pulls this component down.

The weighted composite is then translated into a label:

Excellent — composite ≥72
Good — composite ≥55
Limited — composite ≥38
Poor — composite <38

Copyability never affects HV Score, rankings or any organic metric. It is a forward-looking entry-suitability signal, not a judgement on past performance.

Vault Health

A recent-conditions indicator: Excellent, Healthy, Watch or Warning. It reacts to current drawdown relative to the historical maximum and to recent performance deterioration. Flow, leverage and concentration warnings are added as snapshot history builds.

Data & Limitations

Data is collected from public Hyperliquid endpoints and stored as our own continuous historical record. Vault history from the source is coarser further back, so long-horizon statistics are approximations. We never fabricate metrics: where evidence is insufficient, the site shows Insufficient Data, Not Available or Low Confidence instead of a number.