Liquidation risk when you copy trade on Hyperliquid

How liquidation works on Hyperliquid

A perpetual futures position is collateralised. You post margin, the exchange marks your position continuously against a mark price, and while your equity stays above the maintenance requirement for that position, nothing happens.

When equity falls below maintenance margin, the position is liquidated: closed by the exchange rather than by you, at whatever the book offers at that moment. There is no negotiation, no waiting for a bounce, and no partial credit for having been right eventually.

Two details make this sharper than most people expect. The trigger is the mark price, not the last trade you saw, so a wick on a thin venue reference can matter. And the close happens into the same liquidity that just moved against you, which is usually thinner than it was a minute earlier.

Maintenance margin, mark price and the trigger

Maintenance margin scales with position size and with the market's risk tier. Larger notional in a less liquid perp requires proportionally more equity to stay alive.

The practical consequence: the distance between your entry and your liquidation price is not a fixed percentage. It is a function of leverage, of position size relative to your equity, and of the market you are in. The same nominal leverage in BTC and in a small-cap perp does not carry the same survival distance.

Cross margin versus isolated margin

Under cross margin, all positions in an account share one equity pool. A losing position draws down the equity supporting every other position, and a liquidation can cascade through the book.

Under isolated margin, each position is backed by margin assigned to it. A loss is capped at that assignment, and other positions are unaffected. You give up capital efficiency and you gain containment.

This distinction is the single most important lever in copy trading, because copy trading multiplies the number of simultaneous positions in an account and therefore multiplies the ways in which a shared margin pool can be drained by something you did not choose.

How liquidation risk transfers when you copy

The instinctive model is that copying a leader gives you their risk. It does not. It gives you a related but distinct risk, and the differences all run through your own margin state.

Your margin state is not the leader's margin state

A leader running a $2m book may open a position that represents a small fraction of their equity. Mirrored proportionally into a $20k account, the same trade can be a large fraction of yours — or, if leverage is capped, a much smaller one.

The leader also knows their whole book. They may be holding an offsetting position elsewhere, a spot hedge, or dry powder they intend to add. You see the perp leg on-chain. Copying one leg of a hedged structure gives you the risk of the leg without the protection of the hedge.

Entry-price divergence moves your liquidation price

A mirrored order is a new order, placed after the leader's fill is already public. You fill at a different price. That difference is usually small, and it is not symmetric in consequence: a worse entry moves your liquidation price closer to the current market.

At low leverage this is immaterial. At high leverage it is not. If the distance from entry to liquidation is a few percent, a fraction of a percent of adverse entry divergence is a meaningful share of your entire buffer. Illustratively, the same trade can survive a wick in the leader's account and be liquidated in yours, purely because your entry was fractionally worse and your buffer fractionally thinner.

This is the mechanism behind an outcome that surprises people: the leader shows a winning trade for the period, and the follower shows a realised loss on the same trade.

A leader who survives can still liquidate a follower

Leaders manage margin actively. They add collateral, reduce size, or move funds between accounts when a position is under pressure. Those actions are account management, not trades, and a mirroring system that copies trades does not automatically reproduce them.

The follower's account therefore rides out the same drawdown with a static margin base while the leader is topping theirs up. The leader survives. The follower may not. Nothing about that requires the leader to have done anything wrong.

Isolated sub-accounts versus one shared account

Run several mirrored leaders inside one account and their positions net into one book with one margin pool. A liquidation there is not a leader-level event; it is an account-level event that closes exposure belonging to leaders who were doing fine.

Run each leader in one Hyperliquid sub-account per mirrored leader and the failure is contained. The worst case for a sleeve is the loss of the margin assigned to that sleeve. The other sleeves keep their positions, their margin and their liquidation distance.

Containment is not free. Splitting margin across sleeves means each sleeve has a smaller buffer than the combined pool would have provided, so an individual sleeve is easier to liquidate than the same position would be in a cross-margined account. What you buy is the guarantee that one liquidation does not become all of them.

Leverage and position sizing effects

Leverage is the variable that converts an uncomfortable drawdown into a terminal one. It compresses the distance to liquidation linearly while amplifying PnL, which means it degrades survivability faster than it improves returns.

Copying a leader's nominal leverage is rarely sensible. Their leverage choice is calibrated to their equity, their hedges and their intent to intervene. A mirrored account has none of those. Capping mirrored leverage independently of the leader deliberately breaks the ratio, and it is the right thing to break.

Per-leader notional ceilings do related work from the other direction: they bound how large any one sleeve can become relative to the account, so a single market's move cannot dominate the outcome even if the leader is willing to let it dominate theirs.

• Higher leverage shortens the distance to liquidation proportionally.
• Entry divergence costs more buffer at higher leverage.
• Notional ceilings bound the size of any one sleeve's failure.
• Leverage caps break the leader-follower ratio on purpose.

Cascading liquidations and crowded trades

The dangerous scenario is not one leader being wrong. It is several leaders being wrong in the same direction at the same time, which is more likely than independence would suggest.

Correlated leaders, one direction, one market

Successful Hyperliquid traders read the same order flow, trade the same liquid perps and often express the same macro view. A basket of high scorers can quietly become several sleeves of the same trade.

When that trade unwinds, sleeve-level isolation still prevents one liquidation from closing another sleeve's positions, but it does not prevent all the sleeves from losing simultaneously. Isolation is a containment mechanism, not a correlation hedge.

This is why breadth of markets traded belongs in leader scoring at all: it is a proxy for whether adding a leader adds a distinct bet or another copy of the existing one.

Thin books, funding resets and gap risk

Liquidations feed on themselves. Forced closes push price further in the direction that triggered them, which triggers more liquidations. In thin weekend books and in smaller perps, that loop runs faster and further than the depth chart suggests.

Funding resets add a scheduled point of pressure: crowded positioning means a large funding payment falls due at a specific timestamp, and accounts already close to maintenance can be pushed through it by the payment itself.

None of this is exotic. It is the normal behaviour of leveraged markets, and it is the reason survivability — not return — is the metric that decides whether a strategy is copyable.

Controls that reduce, but do not eliminate, the risk

There is a short list of controls that measurably change liquidation outcomes, and a longer list of things that sound protective and are not.

The ones that work are structural: isolation so failures cannot spread, leverage caps so the buffer stays wide, per-leader notional ceilings so no sleeve dominates, and score floors that remove leaders whose position discipline or survivability has deteriorated before the market tests it.

The ones that do not work reliably are reactive: stop-losses that assume a fill at your price in a gapping market, and manual intervention that assumes you are awake.

• Isolated margin per leader — contains a failure to one sleeve.
• Leverage caps applied independently of the leader.
• Per-leader notional ceilings relative to account equity.
• Score floors on drawdown depth and liquidation history.
• Keeping unused margin in the account rather than deployed.

Honest limitations

No control set makes liquidation impossible. A sufficiently violent move liquidates conservatively sized positions too, and in a genuine cascade the fill you receive can be far worse than the trigger price implies.

Isolation guarantees containment of one sleeve, not preservation of the account. Simultaneous adverse moves across correlated sleeves can produce a large aggregate loss with no individual sleeve behaving unusually.

And every mirrored account inherits an execution layer it does not control: latency, book depth and rounding all affect the price at which risk is actually taken on.

Past performance is not indicative of future results. Perpetual futures are leveraged instruments and carry a substantial risk of loss, including the loss of your entire position.

Conclusion

Evaluating a copy-trading system on the leaders' returns answers the least important question. The question that decides your outcome is what a bad day does to your margin: whether one leader's failure can close another's positions, whether mirrored leverage is bounded, and whether a liquidation is a contained event or an account-level one.

The controls in place here — isolation, leverage caps, notional ceilings and survivability floors — are described in full on the risk page, alongside the risks they do not address.