SolAvaxHyperliquidRatioMeanReversionPairsNeutralDaily

SOL/AVAX Hyperliquid Cross-Asset Ratio Mean-Reversion Pairs (Same-Venue, Market-Neutral, Daily)

A market-neutral statistical-arbitrage pairs trade between TWO DIFFERENT but strongly correlated Layer-1 perpetuals on Hyperliquid: SOLUSD.HYPERLIQUID and AVAXUSD.HYPERLIQUID. Both are smart-contract L1s that co-move with the same beta to the broad alt market, so their price RATIO is a stationary, mean-reverting spread — when it stretches, idiosyncratic flow (a SOL-specific pump, an AVAX-specific dump) has temporarily decoupled them, and it reverts as relative value reasserts. This is fundamentally distinct from every abandoned spread strategy this session: those were all SAME-ASSET (one coin across two venues, or spot-vs-perp of one coin), which are arbitrage-pinned to within a few bps and therefore sub-fee and untradeable. A cross-asset ratio between two SEPARATE coins is NOT arbitrage-bounded — it routinely swings several percent — so the captured reversion (whole percent) dwarfs the ~0.18% two-leg Hyperliquid round trip. Crucially, BOTH legs trade on the SAME venue at the SAME bar timestamps, so there is no cross-exchange fill-timing window and no two-venue neutrality-leak artifact. Simple, few-parameter z-score construction to resist overfitting. Fills the under-represented Hyperliquid venue, pairs scope, and market-neutral direction buckets in one strategy.

Implements a cross-ASSET ratio pairs trade, which is fundamentally distinct from the abandoned same-asset spreads (one coin across two venues, or spot-vs-perp of one coin). Those are arbitrage-pinned to a few bps and sub-fee; the ratio of two SEPARATE but co-moving L1s (SOL, AVAX) is NOT arbitrage-bounded and routinely swings whole percent, so captured reversion dwarfs the ~0.18% two-leg Hyperliquid round trip. Both legs on the same venue at the same bar timestamps removes the two-venue neutrality-leak / fill-timing artifact that broke prior cross-venue spreads. Stationarity premise: SOL and AVAX are smart-contract L1s with the same broad-alt beta, so idiosyncratic flow (a SOL pump, an AVAX dump) decouples the ratio and relative value reasserts. Few-parameter z-score (4 edge tunables) to resist the walk-forward overfitting seen this session. Fills the under-represented Hyperliquid venue, pairs scope, and market-neutral direction buckets at once. leverage=1.0 with sizing via per_leg_pct (not leverage), so no unused-leverage gate triggers; both legs are HL perps so shorting is permitted.

Drift-dominant non-stationary spread — the analyst-deprecated cross-asset crypto-ratio-pairs failure class, confirmed by the backtest. The hypothesis's premise that the SOL/AVAX ratio is stationary and mean-reverting is empirically false: over 2023-2026 SOL structurally outperformed AVAX (divergent narratives/beta), so the ratio TRENDED upward rather than oscillating around a mean. A z-score mean-reversion that shorts SOL / longs AVAX when the ratio is 'high' (z>=2) therefore fades a persistent trend and gets repeatedly run over as the ratio keeps climbing — producing the unmistakable drift-dominant signature: 2024 annual -20.5%, largest_loss -$18,314, avg_loss ($4,443) > avg_win ($3,779), return_skew -4.17, kurtosis 57. Net result is a loss (total_return -4.3%, Sharpe -0.21, profit_factor 0.96, expectancy -$82.77/trade) that is regime-dependent (2025 +16% does not offset 2024 -20%). The hypothesis correctly distinguishes this from the same-asset sub-fee spread class (the ratio does swing whole percent and clears fees), but the fee economics are moot because the ratio does not reliably revert — it trends; no parameter tuning (z-window, entry/exit thresholds) fixes a non-stationary spread, since a z-score MR on a trending series will always get run over. Not iterate (no code bug; the premise itself is wrong). Not revise_hypothesis onto a different pair: crypto L1 pairs are notoriously non-cointegrated and de-cointegrate as narratives rotate, which is precisely why this mechanism class was deprecated. FAILURE PATTERN: market-neutral cross-asset z-score ratio mean-reversion on two correlated crypto L1 perps (SOL/AVAX on Hyperliquid) fails because the ratio is drift-dominant, not stationary — the dominant coin's structural outperformance makes the MR fade a trend and absorb catastrophic losses (2024 -20.5%, -$18k single loss, skew -4.17); cross-asset crypto-ratio pairs are not a tradeable stationary spread regardless of fee headroom.

Market-neutral cross-asset ratio mean-reversion pairs trade between two correlated Hyperliquid L1 perpetuals: SOLUSD.HYPERLIQUID (primary) and AVAXUSD.HYPERLIQUID (extra), on 1-DAY bars. ratio = sol_close/avax_close; the strategy trades the rolling z-score of that ratio over zscore_window bars. Both legs are perps on the SAME venue at the SAME daily timestamps, so there is no cross-exchange fill-timing window. Ratios are appended to a SINGLE synchronized buffer (one value per primary bar, only after the same-timeframe AVAX bar for that timestamp has arrived via the base sync barrier), so the z-score is never computed on misaligned per-leg arrays. ENTRY when flat on both legs: z>=+entry_z (SOL rich) -> SHORT SOL + LONG AVAX (primary SELL); z<=-entry_z (SOL cheap) -> LONG SOL + SHORT AVAX (primary BUY). EXIT (flatten both): |z|<=exit_z after min_hold_bars, or frozen max-hold / combined-PnL kill switch. Equal USD notional per leg => dollar-neutral; PnL is ratio reversion. 4 edge parameters (zscore_window, entry_z, exit_z, per_leg_pct); hold/sync/kill backstops frozen.

Clean implementation: same-venue same-timestamp legs (no cross-exchange fill-timing artifact), atomic two-leg neutral hedge, sensible z-score construction, fills under-represented HYPERLIQUID/pairs/market-neutral buckets.

Correctly avoids the same-asset sub-fee spread trap — the SOL/AVAX ratio does swing whole percent, so per-trade economics clear the ~0.18% HL round-trip fee.

Core premise (SOL/AVAX ratio is stationary/mean-reverting) is empirically false: SOL structurally outperformed AVAX 2023-2026, so the ratio TRENDED — the z-score MR fades a persistent trend and gets run over.

Drift-dominant failure signature: 2024 annual -20.5%, largest_loss -$18,314, avg_loss > avg_win, return_skew -4.17, kurtosis 57 — the 'fade a trend, get crushed when it continues' profile, not a reverting spread.

Net negative and regime-dependent: total_return -4.3%, Sharpe -0.21, profit_factor 0.96, expectancy -$82.77/trade; 2025 +16% does not offset 2024 -20%.

Matches the analyst-deprecated cross-asset crypto-ratio-pairs class (drift-dominant, not a stationary process); only 33 entries over ~3 years of HL history — not optimizable and not fixable by parameter tuning.

ratio trended (2024 -20.5%, skew -4.17)

stationary/mean-reverting

positive

SolAvaxHyperliquidRatioMeanReversionPairsNeutralDaily traded the SOL/AVAX price-ratio z-score as a same-venue, market-neutral pairs reversion, arguing a two-coin ratio swings enough to clear fees unlike the session's sub-fee same-asset spreads. The implementation was clean — synchronized same-timestamp legs, atomic neutral hedge, no fill-timing artifact — and fees indeed cleared, but the premise was wrong: SOL structurally outran AVAX so the ratio trended, and fading it produced a -4.3% return, -0.21 Sharpe, a 2024 -20.5% year, and a -$18,314 single loss with skew -4.17. The analyst ruled it the deprecated drift-dominant cross-asset ratio-pairs class — non-stationary, unfixable by tuning, and not rescued by swapping pairs since crypto L1s de-cointegrate as narratives rotate. It ended after one iteration as abandoned, never advancing to optimization or risk review.

Cross-asset crypto L1 ratio pairs are not a tradeable stationary spread regardless of fee headroom: SOL/AVAX is drift-dominant, not cointegrated — the dominant coin's structural outperformance makes a z-score MR fade a trend and get run over (2024 -20.5%, -$18k single loss, skew -4.17) — and rotating to another L1 pair only re-encounters the same de-cointegration, which is why the class is deprecated.

It was abandoned at the pre-optimization backtest-review gate (verdict: abandon) because the core premise is empirically false — SOL structurally outperformed AVAX 2023-2026 so the ratio trended rather than mean-reverted, making the z-score MR fade a persistent trend and absorb catastrophic losses — a non-stationary spread no parameter tuning can fix, so optimization and all later stages were never reached.

A same-venue, market-neutral cross-asset ratio mean-reversion pairs trade between two correlated Hyperliquid L1 perps — trading the rolling z-score of the SOLUSD/AVAXUSD price ratio on daily bars, shorting SOL + longing AVAX when the ratio stretched rich (z≥entry) and the reverse when cheap, flattening on reversion in equal USD notional — pitched as distinct from the session's sub-fee same-asset spreads because a two-coin ratio swings whole percent and clears the ~0.18% HL round-trip.

Over ~3 years (301 days of HL history) and 33 pairs (66 leg-trades) it was net negative: total return -4.3%, Sharpe -0.21 (CI [-1.70, 2.14]), profit factor 0.96, expectancy -$82.77/trade, win rate 53% but avg_loss ($4,443) > avg_win ($3,779), with a drift-dominant signature — 2024 annual -20.5%, largest loss -$18,314, return skew -4.17, kurtosis 57 — only partly offset by 2025 +16%; fees were a non-issue (commission 1.24% of gross).