Applications open · Grid-only SMA

Hyperliquid SMAs — grid-only and unhedged

We are accepting Separately Managed Accounts on Hyperliquid again. The condition is the product: these accounts run screened, long-only oscillation grids only. There are no hedging bots on this venue.

Request SMA access View Bot Research

The risk profile, stated first

Hyperliquid grids are unhedged. They accumulate long inventory as price falls and sell it back as price recovers. In a sustained decline, the account carries that inventory and the drawdown that comes with it.

This is not a hedged strategy with a missing feature; it is a deliberately different product. If that exposure does not suit you, our hedged systems run on OKX and Binance.

What the grid is built to collect

Repeated movement through a range, not a directional forecast.

Oscillation

Orders are distributed through a defined range. Downward fills add inventory and a recovery releases it rung by rung, converting back-and-forth movement into realised cycles.

No hidden hedge

Hyperliquid uses one-way positions for this mandate. There is no offsetting short leg, and the grid must be judged on its own inventory, margin and sustained-decline risk.

How a market earns a seat

Every active Hyperliquid perpetual is screened; passing is not assumed.

Six independent mean-reversion checks across the active perpetual universe

An out-of-sample holdout to test whether the behaviour survives unseen data

Estimated half-life to measure how quickly price displacements decay

Hurst exponent and variance ratio to reject persistent or random-walk regimes

Measured order-book impact at size, plus current funding and trading costs

Estimated net cycle edge after impact, fees and funding

Hurst exponent

Hurst asks whether moves tend to continue or reverse. A reading around 0.5 is consistent with a random walk, above 0.5 points towards persistence and below 0.5 towards mean reversion. For a grid, we want a convincingly low reading rather than trend persistence.

Variance ratio

Variance ratio asks how much short-horizon movement survives over longer horizons. A reading near 1.0 resembles a random walk; around 0.5 means roughly half the movement cancels itself out—the kind of oscillation a grid is designed to collect.

Statistics are not enough

A market can pass every reversion test and still be rejected. If measured impact, fees and funding consume more than the estimated cycle earns, net cycle edge is negative and the market does not trade.

Grid-only. Unhedged. Screened.