How the research works

The edge: funding / basis capture

On crypto perpetual-futures venues, the funding rate is the periodic payment exchanged between long and short positions to keep the perpetual price tethered to spot. When the market is net-long — which it usually is — funding is paid by longs to shorts. A delta-neutral position (short the perpetual, hold the spot as a hedge) carries little to no directional price exposure and collects that funding stream as its return source.

Why the inefficiency persists

This is not a pricing error that arbitrage erases. It is structural: there is persistent, leverage-hungry long demand in crypto, and someone has to be paid to take the other side. The funding payment is the price of that demand. As long as retail and momentum capital want leveraged long exposure, a hedged short is compensated for providing it. That is why the signal is observable over long windows rather than a fleeting anomaly — though the size of the payment varies with sentiment, and it can turn negative when the market flips net-short.

The costs, modeled in full

A funding edge is small per period, so costs decide whether anything survives. The backtest charges, rather than assumes away:

• — Trading fees on both legs of the hedge, every time the position is opened, closed, or rebalanced.
• — Slippage on execution, so fills are not assumed to land at the mid-price.
• — Funding itself, charged in both directions — the strategy pays funding in the periods it runs against it, not only collects it.
• — Rebalancing drag, because keeping the position neutral has a turnover cost.

The most important finding is a cost-sensitivity one: the edge is fragile to rebalancing frequency. Reweighting daily bleeds the thin funding return away in fees and slippage; holding the position and rebalancing on a weekly or biweekly cadence is what lets the net result survive. That is a finding the data produced, not a knob tuned to produce a nicer curve.

The bias checks

A backtest is only as honest as the traps it guards against. These are the ones that matter most here:

• — Look-ahead: positions use only information available at the time of the decision; no future funding or price leaks into a past trade.
• — Survivorship: the universe is the liquid majors, and the result is not built on assets selected because they happened to do well in hindsight.
• — Overfitting: the parameters are kept deliberately coarse; a random-data smoke test is run to confirm the harness reports roughly zero edge on noise, so a positive result is not an artifact of the machinery.
• — Out-of-sample: the window is split so the result is checked on a segment the design did not see, rather than reported on the whole sample at once.