Hyperliquid Market Risk Recommendations

Scoring is based on 10 categories worth up to 10 points each, for a base score of 100. Configured stable-pegs and selected wrapped majors are excluded from the scored universe. Examples include USDC, USDT, PYUSD, WBTC, STETH, WSTETH, and XAUT. The full exclusion list is configured manually. The scored universe includes active Hyperliquid futures assets plus non-active assets with at least $50M of market cap.

For each scoring category, values outside the stated scale are capped at the nearest endpoint. For higher-is-better metrics, values above the upper limit get the maximum 10 points and values below the lower limit get 0 points. For lower-is-better metrics such as slippage and HLP OI share, values below the lower limit get 10 points and values above the upper limit get 0 points.

Per-factor grades use the category's 0-10 point score directly. The total Score column is a 0-100 sum and uses the total-score ranges below. The same letter labels are reused for both, but Total Grade and Factor Grade are calculated on different scales.

A dash means no usable data was available for that metric. Missing data may affect factor scoring according to the methodology below.

Warning pills are audit flags and do not directly change the recommendation. Yellow is the first warning level, orange is a stronger warning, and red is the strongest warning.

The HLP orange thresholds are the base thresholds. Yellow starts 10% below those thresholds, and red starts 20% above them. High spot slippage uses one-way slippage from mid price.

Larger markets are treated as more resilient and more economically relevant. Market cap uses the latest cached market-source files and is the median of available market-cap estimates from CMC, CoinGecko, and CoinCap. It is scored on an exponential scale from $1M to $5B.

Credible spot volume means reported spot volume from a conservative venue set, rather than every venue that reports the market. The model focuses on Binance, Bybit, OKX, Hyperliquid, Kraken, and Coinbase because reported volume on some venues can be harder to interpret. The value is the latest cached daily notional volume from each venue's daily candle data, aggregated across the conservative venue set, and scored on an exponential scale from $10K/day to $1B/day.

Spot order book quality is measured as estimated one-way basis-point cost from mid price for a $10K order. Lower cost scores better, using a reverse exponential scale from 500bp down to 0.5bp. The primary venue set is Binance, Bybit, and OKX. If at least two of those venues have usable books, the score uses the best liquidity from those primary venues. If fewer than two primary venues have usable books, the model evaluates the broader spot venue set, including the primary venues, and still uses the best available liquidity. A book is usable when both bid and ask depth are available for the target notional size. The displayed and scored value is half of the best available $10K round-trip spread, so it represents one-way slippage from mid price. This measures best-execution liquidity rather than average venue liquidity. The feature used in scoring is the median of recorded order book depth snapshots.

Oracle robustness uses the same venue-point weights documented by Hyperliquid, applied to current spot-market coverage observed in the latest cached exchange candle data. Binance contributes 3 oracle points; OKX and Bybit contribute 2 oracle points each; Kraken, Kucoin, Gate, and MEXC contribute 1 oracle point each. For example, a coin with Binance and MEXC spot coverage has 4 raw oracle points.

The table's Oracle pts. value is the raw Hyperliquid oracle point total. The factor grade converts that raw value into 0-10 model points on a linear 1-to-11 scale. In effect, the factor score is raw oracle points minus 1, capped to the 0-10 score range. A raw value of 10 receives 9 factor points, which maps to A; a raw value of 11 receives 10 factor points, which maps to A+.

Credible linear futures volume means reported futures volume from a conservative venue set, rather than every venue that reports the market. The model focuses on Binance, Bybit, OKX, and Hyperliquid because reported volume on some venues can be harder to interpret. The value is the latest cached daily notional volume from each venue's daily candle data, aggregated across the conservative venue set, and scored on an exponential scale from $10K/day to $1B/day.

Futures order book quality is measured as estimated one-way basis-point cost from mid price for a $100K order. Lower cost scores better, using a reverse exponential scale from 500bp down to 0.5bp. For each venue, the model computes buy-side and sell-side taker slippage against the live order book relative to mid price, then sums the two sides into a round-trip spread and reports half of that spread. The primary venue set is Binance, Bybit, and OKX. If at least two of those venues have usable books, the score uses the best liquidity from those primary venues. If fewer than two primary venues have usable books, the model evaluates the broader futures venue set, including the primary venues: Binance, Bybit, OKX, Gate, KuCoin Futures, MEXC, Bitget, Bitfinex, BitMEX, HTX, Crypto.com, and Kraken Futures. Hyperliquid futures are included when available. A book is usable when both bid and ask depth are available for the target notional size. The score uses half of the best available $100K round-trip spread across venues, so it represents one-way slippage from mid price. This measures best-execution liquidity rather than average venue liquidity. The feature used in scoring is the median of recorded order book depth snapshots.

Trading volume on Hyperliquid measures whether the market is actively used on the venue. The value is the latest cached daily Hyperliquid notional volume from the daily candle data. It is scored on an exponential scale from $1K/day to $1B/day.

Open interest on Hyperliquid captures active risk and positioning on the venue. The value is the median dollar open interest from recorded Hyperliquid meta/context snapshots joined to mark prices. It is scored on an exponential scale from $1K to $1B.

HLP is the Hyperliquid Liquidity Provider vault. Markets where HLP represents a large share of Hyperliquid open interest are treated as more dependent on platform-provided liquidity, because more market risk is warehoused by HLP. The value is the median HLP share from recorded HLP OI and Hyperliquid meta/context snapshots. Lower HLP share scores better, using a reverse linear scale from 20% down to 0.1%.

Hyperliquid order book quality is measured as estimated basis-point cost for a $10K order from mid price. Lower slippage scores better, using a reverse exponential scale from 500bp down to 0.5bp. The displayed value is half of the computed round-trip spread. The feature used in scoring is the median of recorded Hyperliquid order book depth snapshots.

Non-active Hyperliquid futures assets receive a baseline of 2 points for each Hyperliquid-specific category because they do not have live Hyperliquid volume, open interest, HLP OI share, or Hyperliquid slippage. The baseline keeps missing venue-specific fields from being treated exactly the same as an actively listed market with poor live metrics. Because of this baseline, non-listed assets and actively listed assets are not directly comparable on Hyperliquid-specific factors; compare them primarily through the final recommendation logic and the non-Hyperliquid market quality factors.

The possible recommendations are List, Increase, Reduce, Delist, or blank. List means an unlisted asset scores strongly enough to be a listing candidate. Increase means a listed asset scores strongly enough relative to the next higher leverage tier to be a candidate for higher maximum leverage. Reduce means a listed asset scores weakly enough relative to its current leverage tier to be a candidate for lower maximum leverage. Delist is the Reduce path for assets already in the lowest listed leverage bucket.

Assets are grouped into current maximum leverage buckets: unlisted, up to 3x, up to 5x, up to 10x, and above 10x. For each bucket, the model computes score distribution thresholds from assets currently in that bucket. An asset is plotted against its current leverage bucket and total model score. Green regions are upgrade/listing regions; red regions are downgrade/delisting regions.

Upgrade thresholds are based on the median score of the next higher leverage bucket. A listed asset at or above that threshold receives Increase; an unlisted asset at or above the first listed-bucket threshold receives List. Downgrade thresholds are based on the 75th percentile score of the next lower leverage bucket, except the 3x bucket uses a fixed delist threshold of 30. A listed asset at or below its downgrade threshold receives Reduce, or Delist if it is in the 3x bucket.

Recommendations are based on score and current leverage bucket only. Warning pills are shown separately for auditability and do not directly override the recommendation label. A low absolute score is not automatically a delist signal; it has to be low for the asset's current leverage bucket under the rules above.