Funding rates: the cost of holding a perpetual that nobody budgets for

A futures contract with an expiry converges to spot because it has to — at settlement there is nowhere else to go. A perpetual never expires, so it needs a different mechanism. That mechanism is funding: a periodic payment between the two sides of the market that pulls the contract price back toward the index.

The mechanics

When the perpetual trades above the index, longs pay shorts. When it trades below, shorts pay longs. The payment happens at fixed intervals — commonly every eight hours — and it is proportional to your position's notional, not to your margin.

That last detail is the one that surprises people. Funding is charged on the notional. At 10× leverage, a funding rate that looks like a rounding error against the contract value is ten times larger against the capital you actually put up.

# Funding is on notional, so leverage multiplies its effect on your equity.
notional      = qty * mark_price
payment       = notional * funding_rate          # positive rate: longs pay
periods_held  = hours_held / 8
total         = payment * periods_held

# Against margin, not notional:
cost_on_capital = total / (notional / leverage)  # = total * leverage / notional

Our funding calculator does this arithmetic including the annualised figure, which is usually the number that makes the point.

Small rates, large annual numbers

The baseline rate is set so that a balanced market drifts gently. The interesting cases are the others. A position held through a period of elevated funding can lose more to carry than to price movement, and the loss is invisible in the entry-and-exit arithmetic most bots log.

Where bots get it wrong

• Funding as a reporting line. Recorded after the fact, in the P&L, where it explains a discrepancy rather than preventing one. It belongs in the decision.
• Holding a hedge indefinitely. A long and a short of equal size are directionally flat and pay funding on both legs continuously. See hedge mode — this is the most expensive way to avoid taking a stop.
• Ignoring it in short-horizon strategies. Fair, up to a point: a position held twenty minutes rarely crosses a funding boundary. But rarely is not never, and a strategy that occasionally holds through settlement needs the cost modelled or it will be surprised.
• Ignoring the boundary itself. Price and open interest often behave differently in the minutes around settlement, as positions are adjusted to avoid or capture the payment.

Putting it in the decision

Two concrete integrations, both cheap:

1. Expected cost in the entry filter. Estimate how many funding periods the position is likely to span and add the expected payment to the cost side of the edge comparison, alongside fees and slippage.
2. Funding in the risk engine. Accumulated funding on an open position counts toward the daily loss limit. A position bleeding carry is losing money whether or not the price has moved.

Funding as a signal, carefully

Persistently high positive funding says leveraged longs are crowded and paying for the privilege. That is genuine information about positioning, and it appears in plenty of strategies as a feature — ours included, as part of the episode context in the pump-fade model.

The caution is that it is a crowded, well-known signal, and the obvious trade — fade extreme funding — has the payoff shape of picking up small amounts repeatedly and occasionally being run over. If you use it, size it as what it is.

The short version

• Funding is charged on notional; leverage multiplies its effect on your capital.
• Add expected funding to the cost side of the entry decision, not just to the report.
• Count accumulated funding toward the daily loss limit.
• Show it per position and per leg in monitoring.
• A hedge is a position with a continuous carry cost, not a pause button.