Hyperliquid Market Making: Capturing Maker Rebates at Scale

Hyperliquid's fee schedule rewards passive liquidity with maker rebates that compound meaningfully at volume — but only if your bot is positioned to collect them consistently. The difference between a market maker that breaks even and one that runs profitably often comes down to three things: how tightly quotes are placed relative to the mid, how inventory risk is managed under skew, and whether the bot can sustain the volume thresholds that unlock higher rebate tiers. This post walks through each layer.

Understanding Hyperliquid's Maker Rebate Tiers

Hyperliquid's CLOB grants maker rebates that increase with 14-day rolling volume. Crossing a tier boundary does not just improve your margin per trade — it compresses the break-even spread at which passive quoting becomes viable, which in turn lets you tighten quotes, win more fills, and compound into the next tier faster.

The practical implication: your bot should track its trailing volume continuously and apply a different quoting aggression at each band. Near a tier ceiling, pulling back spread width slightly to capture extra volume makes sense. Once comfortably inside a tier, spread can widen to harvest rebate with lower adverse-selection risk.

A few mechanics that matter here:

• Order placement latency. Hyperliquid's matching engine is fast, and stale quotes at the top of the book get lifted immediately on any directional move. WebSocket order-book feeds with sub-100 ms round-trip are the floor; anything slower and you are effectively providing free options to faster participants.
• Tick size and queue position. Because Hyperliquid is a price-time priority CLOB, queue position inside a price level matters. The bot must decide when to cancel-and-replace versus rest, balancing queue priority against quote staleness.
• Cancel fees. Excessive cancel-replace cycles eat into rebate capture. Good bots model expected fill probability before deciding to replace, rather than repricing on every tick.

Inventory Skew and Risk Controls

A maker bot that ignores inventory will eventually hold a concentrated directional position at the worst possible time. The standard remedy is inventory skew: as the bot accumulates long exposure, bid prices are shaded down and ask prices shaded up, making it cheaper to attract the offsetting flow organically rather than crossing the spread to flatten.

The skew function should be calibrated to the asset's realized volatility and the funding rate. On Hyperliquid perpetuals, a persistently positive funding rate signals crowded longs — resting bids aggressively into that environment increases adverse-selection risk, so the skew should be more aggressive. Reading the funding rate every epoch and folding it into the fair-value estimate prevents the bot from systematically leaning the wrong way.

Hard inventory limits act as a backstop. When gross exposure exceeds a configurable threshold — expressed in notional or as a fraction of available margin — the bot enters a one-sided quoting mode, only posting orders that reduce the position. A kill-switch layer sits above this: if unrealized loss crosses a daily drawdown ceiling, all open orders are cancelled and the bot halts until a human reviews the state. These are not optional features; they are the difference between a controlled loss and an account wipe.

Our hyperliquid market maker case study shows how these layers interact in a production system, including how skew coefficients were tuned across different volatility regimes.

Execution Infrastructure

Rebate capture at scale demands low-latency execution infrastructure. For a Hyperliquid bot specifically:

• Direct API connectivity to Hyperliquid's order entry endpoint with connection pooling and keepalive, avoiding TLS handshake overhead on every order.
• Local order state management. Rather than polling REST for order status, maintain an in-memory order book of resting orders, updated via WebSocket fills and cancels. This avoids rate limits and keeps the feedback loop tight.
• Quote simulation before submission. Before placing or replacing a quote, the bot should verify that the intended price is still inside a configurable distance from the current mid. A stale signal causing the bot to post a bid well above mid is effectively a market buy — a simulation check catches this before it reaches the engine.
• Circuit breakers on data staleness. If the WebSocket feed lags or disconnects, all resting quotes should be cancelled immediately. Quoting blind is categorically worse than not quoting at all.

For teams considering a Solana comparison point, the architecture differs sharply: Solana bots rely on Jito bundles, priority fees, and geyser streams to compete in a mempool-driven environment. Hyperliquid's deterministic CLOB is a cleaner surface for passive strategies, though it demands its own set of operational disciplines.

If you want to explore the broader landscape of trading-bot services we build — including cross-venue and prediction-market strategies — the services page covers the full stack.

---

Building a robust hyperliquid maker rebate bot requires getting all three layers right simultaneously: tier-aware quoting logic, inventory-skew risk controls, and low-latency infrastructure. If you are ready to move from prototype to production, start a project with us and we can scope the right architecture for your target markets and volume.