AVAX Linear-Payoff Replication: Spot-Hedged Hyperliquid Funding Carry

Hypothesis

Derived by the Research Lead from Angeris, Chitra, Evans and Lorig's A primer on perpetuals: operationalize the linear-payoff replication argument as long AVAX spot on Binance and short an equal AVAX quantity on Hyperliquid when historically observed funding offers sufficient compensation for trading costs. For phi(S)=S, Section 3's replication holds one underlying unit per contract unit and its idealized funding expression is zero. Actual exchange perpetuals have different contractual mechanics; their positive funding is a candidate carry opportunity, not an arbitrage guaranteed by the theorem. The authors specify no exchange strategy, entry threshold or holding period. The market, 4-hour decision interval, persistence filter, cost hurdle and exits below are Research Lead choices. Use signal-close market execution on these continuous venues, the registry's equivalent to next-bar-open execution. This is one applicable operationalization; it does not pretend that listed linear perpetuals reproduce the paper's logarithmic or power contracts.

This tests a funding cash-flow mechanism with an explicit underlying hedge, rather than the failed ATOM directional funding reversal or another momentum indicator. It adds Hyperliquid exposure, two instruments, cross-venue execution and a short leg, addressing portfolio imbalances. The supplied corpus_concentration is empty, so AVAX's recent experiment share cannot be established. The known AVAX cross-venue executable-spread strategy targets price dislocations; this hypothesis targets persistent settled funding and permits near-zero entry basis. Options are more underrepresented, but reproducing the paper's nonlinear perpetual contracts with arbitrary vanilla options would change the mechanism and does not justify inventing another hypothesis.

Implementation

Equal-quantity long AVAX Binance spot and short Hyperliquid perpetual, entered on persistent positive settled funding that exceeds the planned cost hurdle. Both legs exit on nonpositive trailing daily funding, stale funding, a 3 percentage-point adverse basis widening, or the holding deadline.

Implements the Research Lead linear underlying hedge with two real legs, 4-hour timestamp-aligned decisions and no cycle rebalancing. Hourly funding is summed into trailing 24-hour blocks; the lookback mean projects holding-period carry and the positive-day fraction controls persistence. Risk uses combined prefunded account capital and a spread-loss budget, with each leg capped at 20% of combined capital and its own available balance. Funding cash flows are booked by the engine, never fabricated by the strategy. Defaults within the stated tunable domains are Research Lead operationalization choices; the full research plan is preserved. The spread stop is evaluated on aligned closes because a cross-venue spread is not a single-instrument stop trigger. Listed perpetual carry is not a theorem-guaranteed arbitrage.

Verification Results

Projected carry assumes historical funding persists throughout the holding period; early funding exits can realize less carry while still paying both legs' costs. Assess paired-cycle returns including credited funding in the full backtest.

Backtest Review

Sharpe

0.18

Total return

2.39%

Max drawdown

4.98%

Trades

28

Win rate

46.4%

Profit factor

1.07

Detail records confirm 14 hedged cycles: equal AVAX quantities, matched entry/exit times, long spot and short perpetual. Entry basis ranges from -0.2002% to +0.1053%, with no implausible spread capture.

Funding is credited: $7,452 funding supports $4,774 net paired PnL after modeled costs. Average leg return is 0.515%, above the Hyperliquid 0.15% and spot 0.25% review floors.

The mechanism functions, positions are fully closed, and the full-history design can accommodate the slow cycle rate.

Only 14 independent baskets support the result; 28 closed legs are not 28 independent carry observations. Sharpe 0.18 is uncertain.

QA's concern is partly borne out: six consecutive eight-hour cycles in July–August 2025 collected only about $42 funding and lost approximately $1,235 combined. Entry uses trailing-average carry without requiring the latest daily funding to exceed the immediate exit threshold.

Profit is concentrated in 2024; optimization must assess paired-cycle economics and early-exit churn rather than directional leg profit factor. The single-leg benchmark is explicitly not meaningful.

The required outbox file could not be saved because this session permits filesystem reads only.

Analysis

The implementation does what the hypothesis says. The trades are equal-quantity long Binance spot / short Hyperliquid perp cycles, entry basis is within ±0.2%, funding is credited, and accounting is valid.

Risk is contained: max drawdown 4.98% against the pre-registered 10%, and peak margin utilisation 0.45.

Average trade return of 0.362% clears the 0.18% economics floor (2x the Hyperliquid round trip), so per-trade cost is not the failure.

The sensitivity sweep is flat within one Sharpe standard error (0.85), with no cliffs.

All four pre-registered predictions that could be scored were met: trades, return, drawdown and average trade return.

HARD gate failure from numeric_gates: only 1 of 3 calendar years with exposure was positive (33% < 60%), so promotion is impossible. 2024 carries 108% of the profit.

The year-by-year split in the detail file shows the carry decaying. In 2024, 8 legs earned $5,996 of credited funding, paid $404 of commission and made a net +$5,164. In 2025, 20 legs earned only $1,457 of funding, paid $948 of commission plus modelled impact (21-29% of gross), and lost a net $390. In 2026 the persistence gate never opened, so there were no trades at default parameters.

The holdout (2025-11-03 to 2026-10-10, 6 trades, underpowered) has Sharpe -0.107 and t = -0.007. It gives no forward support, though it is also not statistically adverse.

Full-history economics are thin: PF 1.058, Sharpe 0.179, Calmar 0.19, recovery factor 0.48, and 2.55% total return over 2.6 years.

Signal research: 21 observations is fewer than the 30 needed, so the forecast diagnostic carries no information.

significant positive

hard

advisory only

Outcome Summary

Before committing to a funding-carry hedge, check that funding pays enough year by year, because a full-history average can hide a carry confined to one regime that cost-clearing per-trade returns will not reveal.

After optimization the analyst abandoned it as premise_falsified. It failed the HARD positive-years gate: only 1 of 3 years was positive, and 2024 produced 108% of the profit. Funding fell from $5,996 in 2024 (net +$5,164) to $1,457 in 2025 (net -$390), and the strategy made no entries in 2026 because the persistence gate never opened.

Collect positive AVAX funding on Hyperliquid with a hedged position: long AVAX spot on Binance and short an equal quantity of the Hyperliquid perpetual, opened on 4-hour decisions only when trailing funding was persistent and large enough to cover the costs of trading on two venues (from the linear-payoff replication argument in Angeris et al.).

The initial backtest (2024-02-22 to 2026-10-09) ran 14 hedged cycles (28 closed legs) for a 2.39% total return, Sharpe 0.18, profit factor 1.07, max drawdown 4.98%, and an average return of 0.515% per trade. In the post-optimization review the average trade return was 0.362%, the profit factor 1.058, and the underpowered holdout (6 trades) had a Sharpe of -0.107.

Analysis

A primer on perpetuals