Combining strategies in one basket

The netting problem

Hyperliquid nets positions within an account. If a trend follower is long a market while a mean-reversion leader is short it, a single shared account holds nothing at all — while both leaders continue to pay fees and both signals continue to be generated.

Worse, the netted account produces no attributable result for either leader, which destroys the data scoring depends on. Isolation is not a convenience feature; it is the precondition for measuring anything.

• One sub-account per leader keeps opposing positions live and separable.
• Per-leader PnL remains attributable, so scores stay meaningful.
• Margin and liquidation risk stay contained within each sub-account.
• Replacement affects one sub-account, not the whole book.

How weights are set

Capital is allocated in proportion to composite score rather than split evenly, so exposure follows the strength of the evidence. Per-leader notional ceilings prevent a single high score from dominating, and leverage is capped independently of what the leader uses.

The combination target is mechanism spread, not merely count: several archetypes, several markets, several holding periods.

What it does not do

Diversification reduces the chance that everything loses at once. It does not prevent loss, and in a broad market-wide drawdown correlations rise and the benefit shrinks precisely when it is most wanted.

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.