Point-in-time inputs
Every input is recorded with the time it was read, in UTC. A computation begins only once the trade archive has caught up with the time of computation, and inputs read at materially different moments are not combined.
Methodology · Version 0.9
Brierz Vault Rating Methodology
A Brierz rating is an independent opinion on the strength of a Hyperliquid vault's trading record and on the amount of evidence behind it. It addresses a single question: does the vault's on-chain record show evidence of trading skill, or can its results be explained by chance, market direction or capital flows?
Methodology version 0.9 · Effective 23 October 2026 · Trading data from 25 May 2025 · Page revised 25 October 2026
2,939Vaults with a positive balance, assessed
265Rated: a score, a 90% range and a tier
2,674Not rated, each with its reason
0–100The scale. 50 means no evidence either way
Overview
Each rating consists of three elements:
- a score from 0 to 100 on a fixed scale, where 50 is the neutral point;
- a 90% range, which expresses the uncertainty of the score given the available evidence;
- a tier, one of five, assigned only when the range is precise enough to support it.
Where the data cannot support a reliable assessment, Brierz does not publish a score. It publishes the reason instead.
This document describes the scope, inputs, framework and controls of the methodology. Detailed parameters (component weights, scale calibrations and thresholds) are fixed for each version and documented internally; they are not published.
Nature of a rating. A rating is an opinion formed under a documented, versioned method. It describes a vault's record up to the date of computation. It is not a forecast of future returns, not investment advice, and not a recommendation to deposit in or withdraw from any vault. It is not an assessment of the vault leader as an individual.
Section 1
1. Scope and coverage
Instruments. The methodology covers Hyperliquid vaults trading perpetual futures. Trader wallets are assessed under the same framework; wallet ratings are not yet published.
Population. Brierz assesses every Hyperliquid vault with a positive balance, not a curated selection. Each vault receives either a rating or a stated reason for not being rated, and both outcomes are published.
Data horizon. Trading data begin on 25 May 2025, the start of Hyperliquid's public trade archive. Activity before that date is outside the scope of the assessment. The review of the leader's own deposits and withdrawals extends to the vault's inception wherever the venue's records allow.
Rated, with a tier265 · 9.0%
Not rated, by reason
Results driven by leader flows1,586 · 54.0%
Insufficient independent decisions408 · 13.9%
Risk-adjusted return not measurable278 · 9.5%
No trading activity264 · 9.0%
Seven other reasons138 · 4.7%
Brierz vault snapshot, 23 Oct 2026 · all Hyperliquid vaults with a positive balance
Section 2
2. Data sources and processing
Brierz uses public Hyperliquid data only. Trade history is read from Brierz's own copy of the venue's public trade archive; account state, transfers and prices are read from the official Hyperliquid interface. No data are supplied by vault leaders.
| Input | Source | Use in the assessment |
|---|---|---|
| Executed trades | Brierz copy of the public trade archive | Positions, realised results, independent decisions |
| Funding paid and received | Official interface | Part of the vault's trading result |
| Leader deposits and withdrawals | Official interface | Flow assessment; separated from trading results |
| Depositor flows | Official interface | Excluded from returns and from the flow assessment |
| Account state and vault details | Official interface | Capital at computation; vault terms |
| Liquidations of the vault's own positions | Brierz copy of the public trade archive | Risk adjustment over a recent period |
| Market prices | Official interface | Valuation of open positions |
Flow-adjusted results
The vault's results are reconstructed period by period from its trades, funding and the valuation of open positions. Capital moving in and out is removed, so a large deposit is not counted as a gain and a large withdrawal is not counted as a loss.
Assessment window
Each rating states the period it covers, with start and end dates. For many older vaults the window covers the most recent months and rolls forward with each recomputation. Results are reported as realised over the window; annualised figures are not published.
Section 3
3. Rating framework
3.1 Components
The score is built from four components. Each is measured over the assessment window and placed on a fixed 0–100 scale before the components are combined.
| Component | What it assesses | Status |
|---|---|---|
| Risk-adjusted return | Return earned relative to the volatility and downside risk taken | Required: without it, no score is published |
| Consistency | Whether results are stable across sub-periods or concentrated in a single episode | Used when the window is long enough |
| Position sizing | Discipline and stability of position sizes across trades | Used when enough trades are available |
| Downside control | Depth of drawdowns and the management of losses | Used when measurable |
When an optional component cannot yet be measured, the score is computed from the remaining components and the rating states how many were used.
3.2 Fixed scales
Component scales are set by the methodology version, not by ranking vaults against one another. A vault's score therefore does not change because other vaults enter, leave or perform differently. Scales are calibrated as part of each methodology version and remain fixed until a new version is released.
3.3 Independent decisions
A single trading idea may be executed through many orders. Orders that belong to the same idea are counted as one independent decision. Evidence accumulates only with new decisions, so splitting a position into many fills does not strengthen a record.
3.4 Weight of evidence
The score reflects both the strength of the record and the amount of evidence behind it. Records with few independent decisions are held close to the neutral point of 50, however strong their returns, and move away from it only as evidence accumulates. A record that shows no demonstrated edge cannot score above 50, regardless of its length. A smooth value curve alone does not earn a high score; it must be supported by the trades themselves.
3.5 Risk adjustments
Two specific risks reduce the score, and the rating notes when they apply:
- Concentration: a large share of trading in a single asset.
- Own liquidation: liquidation of the vault's own positions within a recent period.
Section 4
4. Uncertainty and tier assignment
4.1 The 90% range
Every published score is accompanied by a 90% range. The range is estimated by resampling the vault's own history in blocks, which preserves the sequence and clustering of its trades, and recomputing the score on each resample. A short or sparse record produces a wide range. Where the ranges of two vaults overlap, the difference between their scores may not be meaningful, and the ranges should be compared rather than the point scores alone.
4.2 Tier grid
The tier grid is fixed for the methodology version and is not refitted to each new population, so a tier carries the same meaning from one publication to the next.
| Tier | Score | Rated vaults, 23 October 2026 |
|---|---|---|
| High | 72.5 and above | 1 |
| Above average | 57.5 to 72.5 | 20 |
| Average | 42.5 to 57.5 | 173 |
| Below average | 27.5 to 42.5 | 70 |
| Low | Below 27.5 | 1 |
The three central tiers are 15 points wide and centred on the neutral point. An Average tier indicates that the record does not yet distinguish skill from chance; it is not a judgement of mediocre performance. On 23 October 2026, 11 of the 17 rated vaults that had doubled their capital were rated Average.
Swipe sideways to see the whole chart.
Illustration · Cold Process computed 21 Oct 2026, Topiary Capital 19 Oct 2026
Both vaults more than doubled their capital over roughly seven months. The range of Cold Process lies wholly above 50; the range of Topiary Capital crosses 50, so its record does not yet distinguish skill from chance.
4.3 Conditions for a tier
A tier only when the range supports it.
A tier is assigned from the score only when the 90% range is sufficiently narrow and stable to support it. Otherwise the vault is not given a tier, and the reason is stated (section 5).
4.4 Rating notes
| Note | Meaning | Affects the score |
|---|---|---|
| Near tier boundary | A substantial part of the range lies in an adjacent tier; limited new data could change the tier. | No |
| Provisional range | The range is less certain than usual and should be read with additional caution. | No |
| Liquidations partly verified | The liquidation review could not cover the full recent period. The rating also shows the outcome assuming a liquidation in the unverified part. | No |
| Concentration | A large share of trading was in a single asset. | Yes, reduced |
| Own liquidation | The vault's own positions were liquidated within a recent period. | Yes, reduced |
Section 5
5. Rating outcomes
Each vault assessed ends in one of three outcomes:
- Rated: score, 90% range and tier are published.
- Scored without a tier: a score was computed, but the range or a subsequent check does not support a tier.
- Not rated: the record cannot support a reliable score, and the reason is published.
Swipe sideways to see the whole diagram.
Assessment sequence · three outcomes, each published
A refusal to rate is a result in its own right. A figure the data cannot support would appear precise while carrying no information. Where the evidence is borderline, the methodology favours refusal: an overstated score imposes a greater cost on those who rely on it than a withheld score imposes on the vault. Every refusal states its reason together with the vault's own figures behind it. Checks are applied in a fixed order: first the integrity of the record itself, then computational limits, then the more demanding inputs.
Not rated: reasons applied before scoring
| Reason | Description |
|---|---|
| Results driven by leader flows | The leader's own deposits and withdrawals are large relative to the vault's capital; reported returns would primarily reflect capital movements rather than trading. Depositor flows are not counted. |
| Leader flows not fully readable | The leader's deposits and withdrawals could not be read in full, so flows cannot be excluded as the driver of results. |
| Insufficient independent decisions | The record contains too few independent decisions to distinguish skill from chance. |
| Risk-adjusted return not measurable | The history is too short or too flat to measure return relative to risk. |
| Assessment window too short | Insufficient time has elapsed to support an assessment. |
| No trading activity | No trades by the vault are recorded since the start of the data. |
| Incomplete or inconsistent source data | A required input was missing, partly retrieved or read at a different time from the others. The vault is reassessed in the next cycle. |
| Value curve inconsistent with trades | The vault's value curve and its trade-level results point in opposite directions, indicating that one of them is unreliable. |
| Trading history exceeds current capacity | The history is too large for the current calculation. Partial records are not rated. |
Scored without a tier: reasons applied after scoring
| Reason | Description |
|---|---|
| Insufficient precision | The 90% range is too wide or too unstable to support a tier. |
| Range not reliably estimable | The range could not be estimated reliably, or the history is insufficient to estimate it. |
| Tier sensitive to unverified liquidations | The tier would change if a liquidation had occurred in the part of the recent period that could not be verified. |
| Recomputation check not passed | Rescoring the stored inputs did not reproduce the score. Results that fail this check are not published. |
Section 6
6. Quality control and reproducibility
Audit trail
Each vault page closes with an audit trail listing every input used: its type (trade history, funding, deposits and withdrawals, account state, liquidations), its source and the time it was read, in UTC. The audit trail also states the input verification code, the methodology version and the time of computation.
Input verification code
The code is a SHA-256 fingerprint of all inputs a rating used. Any change to any input changes the code. It demonstrates that the inputs behind a published rating have not been altered since computation. It does not certify the accuracy of Hyperliquid's own records, which Brierz reads as published by the venue.
Recomputation check
Before publication, every score is recomputed from its stored inputs and must be reproduced exactly. A derived figure (range or tier) is published only if it is consistent with the published score.
Regression testing
Each methodology version is tested against a fixed set of reference cases covering known edge conditions, each with its expected outcome recorded in advance. A version that fails any reference case is not released.
Version control
Within a version, the scales, the tier grid and the rules are fixed. Any change is released as a new version. Before release, the stored inputs are recomputed under both the outgoing and the incoming version and every difference is reviewed. Each release states what changed and how many vaults changed tier.
Section 7
7. Rating maintenance
Recomputation cycle
Vaults are recomputed approximately weekly. Each rating shows the date of its own computation. A score may change between cycles without new trades, because the window rolls forward, open positions are revalued and funding accrues.
Replies from vault leaders
A vault leader may respond to a rating from the vault's page. A response does not by itself change a rating. Where a response identifies an error in the data or in the calculation, Brierz corrects it, recomputes all affected vaults and records the correction in the change log with its date, cause and effect.
8. Independence and conflicts of interest
- Vault leaders do not pay to be rated, and no payment can alter a rating.
- Brierz does not manage capital, does not charge fees on deposits or copy trading, and does not receive referral payments from vaults.
- Brierz has not issued a token.
- No position in any list is sold or promoted.
Section 9
9. Interpretation and limitations
Historical scope. A rating describes the record up to its computation date. It is not a forecast of future returns.
Data horizon. Activity before 25 May 2025 is not assessed.
Rolling window. For many older vaults the assessment covers the most recent months; each recomputation drops the oldest days and adds the most recent.
Reading the tier. The tier should be read together with the vault's return and drawdown over the same window. An Average tier can apply both to a vault that has doubled its capital and to one with significant losses when the evidence behind either is limited.
Flow attribution. Hyperliquid's records do not always distinguish the leader's capital from depositors' capital. The flow assessment counts only what the records distinguish; where flows cannot be read in full, the vault is not rated.
Differences from venue figures. Because Brierz removes capital flows and reports over a stated window, a vault's return on Brierz may differ from the figure displayed by the venue.
Coverage of large histories. Vaults with very large trading histories, including some of the most active, are not yet rated.
10. Definitions
| Term | Definition |
|---|---|
| Score | A value from 0 to 100 on a fixed scale; 50 indicates no evidence either way. |
| 90% range | The interval within which the score could plausibly lie given the available evidence. |
| Tier | One of five categories derived from the score when the range supports it. |
| Independent decision | A single trading idea, regardless of the number of orders used to execute it. |
| Assessment window | The period covered by a rating, with stated start and end dates. |
| Leader flows | Deposits and withdrawals of the vault leader's own capital. |
| Input verification code | A SHA-256 fingerprint of all inputs used in a rating. |
| Methodology version | The fixed set of scales, tier grid and rules under which a rating was computed. |
Change log
| Date | Change | Effect on ratings |
|---|---|---|
| 25 Oct 2026 | Methodology page restructured and expanded. | None |
| 23 Oct 2026 | Methodology page published; methodology version 0.9 in effect. | None |
| 20 Oct 2026 | Computations begin only once trade data have caught up with the time of computation. | Affected vaults recomputed |
Use the method in context
Before you trust a return, inspect the evidence.
Read the score, its 90% range, the risk record and the reason when Brierz cannot rate a vault. No single number should do all the work.