Modelled versus observed liquidation levels
Of $10.1B in 88,468 tracked Hyperliquid positions (as of 2026-11-22 01:11 UTC), 71.1% is cross margin, whose liquidation price depends on the whole account, so it has no fixed liquidation distance, and a further 11.6% has no liquidation price at all. The tracked positions equal 96.6% of Hyperliquid open interest.
Two ways to draw liquidation levels
Modelled levels start from open interest and assume how it is spread over a set of leverage values. For each assumed leverage they place a liquidation price at a fixed distance from an entry price, about 1 divided by the leverage less the maintenance margin. The result depends on the tiers and weights chosen; the positions themselves are not known.
Observed levels start from positions. On Hyperliquid every account is public, so the size, entry price, margin mode and the liquidation price the exchange computes for each position can be read. Cross margin, added margin and partial closes are already in that price. The limit is coverage: positions of wallets that are not tracked are not counted.
Our heatmap uses observed levels. The numbers below measure, on our own stored positions, how far observed levels are from what a tier assumption would give. They say nothing about any other product.
Leverage implied by the real liquidation price
updated 8m ago (2026-11-22 01:11 UTC) · source: Blockmetric ResearchImplied leverage = 1 / (distance from the entry price to the liquidation price, as a share of the entry price). The maintenance margin brings the liquidation price closer than 1 / leverage, so an isolated position shows a somewhat higher value than its setting (25x on a market with 40x maximum leverage shows about 36x); extra margin moves the price further away and the implied leverage down. The level comparison below includes the maintenance margin.
Isolated positions at round leverage levels
With levels 3x, 5x, 10x, 20x, 25x, 40x, 50x and 100x and a tolerance of ±10%, 38.9% of comparable isolated notional with two or more usable levels sits at a level, 32.2% between levels and 28.9% outside the range of the levels; the bands cover 37.6% of that range. 1,254 comparable positions ($21.6M) can use only one level, so they have no range and are counted apart: 12.8% of that notional sits at its level.
| Level | Positions | Notional | Notional share | Range the bands cover |
|---|---|---|---|---|
| At 3x | 3,129 | $280M | 17.6% | 20.3% |
| At 5x | 2,070 | $115M | 7.3% | 11.2% |
| At 10x | 1,924 | $139M | 8.7% | 4.3% |
| At 20x | 498 | $29.9M | 1.9% | 1.0% |
| At 25x | 163 | $20.7M | 1.3% | 0.6% |
| At 40x | 306 | $29.6M | 1.9% | 0.2% |
| At 50x | 37 | $4.3M | 0.3% | 0.1% |
| At 100x | 0 | $0 | 0.0% | 0.0% |
| At a listed level | 8,127 | $619M | 38.9% | 37.6% |
| Between levels | 5,310 | $511M | 32.2% | 62.4% |
| Outside the range of the levels | 6,456 | $459M | 28.9% | - |
Notional share: of the $1.6B of comparable isolated notional with two or more usable levels (19,893 positions), the same base as the band column. At a level = the observed distance from entry to the liquidation price is within ±10% of the distance Hyperliquid's isolated-margin rule gives at that leverage, maintenance margin included; a level counts where the position's margin tier allows it. Range the bands cover = the part of the range of each position's levels (from ±10% below the smallest to ±10% above the largest expected distance, since at a level counts the whole band) that the bands cover, uniform over the distance: the share that would sit at a level if positions were spread evenly over that range. Not in the table: $21.6M in 1,254 positions with one usable level, which have no range (counted apart in the sentence above). Isolated positions of $1M or more: 84 of 237 sit at a level.
How the tracked notional splits
| Group | Positions | Notional | Share |
|---|---|---|---|
| Isolated, comparable with the levels | 21,147 | $1.6B | 16.0% |
| Cross margin (no fixed tier distance) | 40,363 | $7.2B | 71.1% |
| Not comparable (data not stored, or no listed level allowed) | 456 | $128M | 1.3% |
| No liquidation price | 26,502 | $1.2B | 11.6% |
Cross and isolated margin
| Group | Positions | Notional | Share |
|---|---|---|---|
| Cross margin | 66,460 | $8.3B | 82.6% |
| Isolated margin | 22,008 | $1.8B | 17.4% |
Cross-margin positions whose liquidation price is more than 25% away from where an isolated position at their own leverage setting would be liquidated (maintenance margin included): 96.7% of their notional (38,919 of 39,883). The rest of the account backs them, so the setting alone does not tell where they liquidate.
Leverage settings
| Group | Positions | Notional | Share |
|---|---|---|---|
| 10x | 24,242 | $2.5B | 25.2% |
| 20x | 8,021 | $2.2B | 21.5% |
| 5x | 19,253 | $1.4B | 13.5% |
| 3x | 19,229 | $913M | 9.1% |
| 40x | 1,929 | $608M | 6.0% |
| 25x | 1,168 | $456M | 4.5% |
| 2x | 4,152 | $327M | 3.2% |
| 4x | 2,493 | $315M | 3.1% |
| Other settings | 7,981 | $1.4B | 13.8% |
Coverage by coin
| Coin | Tracked | Open interest | Coverage |
|---|---|---|---|
| BTC | $2.2B | $2.2B | 98% |
| HYPE | $1.1B | $1.1B | 98% |
| ETH | $1B | $1B | 99% |
| NVDA · xyz | $485M | $503M | 96% |
| SP500 · xyz | $479M | $495M | 97% |
| XYZ100 · xyz | $355M | $367M | 97% |
| BRENTOIL · xyz | $300M | $309M | 97% |
| SOL | $280M | $290M | 97% |
| GOLD · xyz | $245M | $260M | 94% |
| SKHX · xyz | $242M | $255M | 95% |
| ANT · xyz | $232M | $243M | 96% |
| ZEC | $218M | $223M | 98% |
| SILVER · xyz | $165M | $172M | 96% |
| MU · xyz | $142M | $146M | 97% |
| CL · xyz | $124M | $138M | 90% |
Method and limits
- Positions: every open position of the tracked Hyperliquid wallets as last polled (newest poll 2026-11-22 01:11 UTC), with size, entry price, notional at the mark, leverage setting, margin mode and the liquidation price the exchange reports. Notional shares are weighted by notional at the mark.
- Coverage: tracked notional per coin divided by Hyperliquid open interest, which counts long and short positions, like the sum of position notionals. Below 100% means some positions are not tracked; their levels are missing from the heatmap, never estimated.
- Cross margin: the liquidation price of a cross-margin position depends on the equity of the whole account, including the profit and loss of its other positions and the free margin, and moves with them. There is no fixed distance from entry to compare, so cross positions are shown as their own group and left out of the level comparison.
- Level comparison, isolated positions: for each of 3x, 5x, 10x, 20x, 25x, 40x, 50x and 100x that the position's margin tier allows (up to its maximum leverage), the distance from entry at which Hyperliquid liquidates an isolated position is computed with the same function as our liquidation price calculator, using the maintenance margin of the coin's margin tier for the position's size (rate = 1 / (2 x the maximum leverage of the margin tier); on one tier (1/L - rate) / (1 - rate) for a long and (1/L - rate) / (1 + rate) for a short). At a level = the observed distance is within ±10% of one of those distances (the closest one counts). Between levels = inside the range from the smallest to the largest of those distances, in no band. Outside the range = beyond the lowest or the highest usable level, for example more margin than a 3x position holds.
- Range the bands cover: for each comparable position with two or more usable levels, the part of the range of its levels (from ±10% below its smallest to ±10% above its largest expected distance: at a level counts the whole band) that the bands cover, uniform over the distance and weighted by nothing; the table averages it over the notional of these positions, the same base as the shares next to it. It is the share that would sit at a level if positions were spread evenly over the range, the figure the share at a level reads against. Positions with a single usable level have no range and are reported apart.
- The level list and the tolerance are choices. More levels or a wider tolerance widen the bands: more notional counts as at a level, and the range the bands cover grows with it. Fewer levels or a tighter tolerance do the opposite. With ±10% neighbouring bands do not overlap, because neighbouring levels lie at least a factor of 1.25 apart.
- Isolated positions move off round levels: added margin moves the liquidation price away from entry and removed margin moves it closer, and a partial close or an increase at another price can change size, entry price and margin in different proportions. A position between levels went through one of these, or was opened at a leverage that is not on the list; the stored data does not tell which.
- Not comparable: positions with a liquidation price whose margin mode or entry price is not stored, whose market has no stored maximum leverage, or whose market allows none of the listed levels, are counted on their own and left out of the comparison.
- No liquidation price: the account holds enough margin that no price move liquidates the position now. A tier model still draws a level for such positions.
- The numbers change with every poll. The study was first published 20 Nov 2026; the date above is the data time.
Cite this data
Source: Blockmetric (blockmetric.io), Modelled versus observed liquidation levels, retrieved 22 Nov 2026, 01:11 UTC, https://blockmetric.io/research/modelled-vs-observed-liquidation-levels
Free to use with attribution and a link to the page. How to cite
Data as of 2026-11-22 01:11 UTC. Source: Blockmetric Research. · Report incorrect data