Sub-account isolation and the netting problem

A worked example

Leader A opens 5 BTC long. Leader B, a mean-reversion trader, opens 4 BTC short an hour later. Mirrored into one account, your position is 1 BTC long. You have paid the taker fees for 9 BTC of trading and you hold the strategy of neither leader.

When leader B closes for a profit, your account gains nothing recognisable: the exit simply moves the net figure again. The outcome is not a blend of two strategies, it is a third strategy that nobody designed and nobody is monitoring.

What isolation changes

With one sub-account per leader, leader A's 5 BTC long and leader B's 4 BTC short exist side by side. Each position carries its own margin, its own liquidation price and its own PnL. The strategies stay intact and, crucially, stay measurable.

Measurability is the underrated half. Isolated accounts give per-leader attribution: you can see which leader produced which result rather than inferring it from a merged equity curve. Scoring and replacement decisions — including the jump-adjusted drawdown used for soft-issue strikes — all depend on that attribution being clean.

• Opposing positions never cancel — both leaders' intent is preserved.
• Liquidation is contained: one leader's blow-up cannot reach into another's margin.
• Per-leader PnL attribution stays exact, which is what makes replacement decisions defensible.

What netting destroys that you cannot get back

Netting is usually discussed as a performance problem. The more damaging loss is informational. Once two leaders share an account, your equity curve is a single series produced by an unknown blend of their decisions, and there is no procedure that separates it afterwards — the information was never recorded, only the net effect.

That matters because every maintenance decision depends on attribution. Reweighting requires knowing which leader earned what. A soft-issue strike requires distinguishing a genuinely underperforming leader from ordinary variance. Replacement requires being able to defend the removal. In a netted account all three are guesses dressed as decisions, and the guess is systematically biased toward blaming whichever leader traded most visibly.

There is a second-order effect on margin. In a netted account all leaders draw on one margin pool, so the sizing of one leader's position changes another leader's liquidation price. A leader who did nothing wrong can be liquidated because someone else's position consumed the buffer.

• Attribution loss is permanent — it cannot be reconstructed from a merged curve.
• Shared margin couples liquidation prices between unrelated strategies.
• Fees are paid on both legs of a position pair that ends up flat.
• Replacement decisions become discretionary because the measurement is missing.

The cost of isolation

Isolation is not free. Margin is fragmented across sub-accounts, so total capital efficiency is lower than a netted book: capital sitting as margin behind leader A is unavailable to leader B. This is a deliberate trade — capital efficiency in exchange for strategy integrity and containment.

It is also why diversification is gated. Below $100k of mirrored volume, Starter mode runs a single leader in one sub-account, because splitting a small balance across ten margined sub-accounts produces positions too small to track their leaders faithfully.

Risk that isolation does not remove

Isolation contains netting and liquidation contagion. It does not protect against correlated losses: if every leader is long the same market into the same downturn, isolated accounts simply record the loss separately. Nor does it protect against exchange-level, oracle or liquidity events.

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.