AaveCompDefiLendingPairsSpreadMeanReversionLS

AAVE-COMP DeFi Lending Pairs Spread Mean Reversion

A market-neutral (dollar-neutral) pairs trading strategy on two highly correlated DeFi lending blue-chip tokens: AAVEUSDT and COMPUSDT on Binance USD-M futures. AAVE (Aave) and COMP (Compound) are the two largest decentralized lending protocols, sharing nearly identical business models (over-collateralized lending, governance tokens, similar TVL dynamics) and the same macro narrative (DeFi sector flow, ETH gas regime, regulatory headlines around lending protocols). Their price ratio has historically oscillated within a stationary range as idiosyncratic news and TVL flows cause one to outperform the other temporarily before reverting. The strategy holds dollar-neutral positions: when the spread (z-score of log price ratio versus rolling mean) extends beyond a threshold, it shorts the outperformer and longs the underperformer, profiting as the spread reverts to its mean. This intentionally mirrors the successful ADA-DOT pairs strategy (Sharpe 6.4, paper_stage) and ETH-BTC ratio strategy (Sharpe 6.8, paper_stage), applied to a different sector (DeFi lending vs L1 infrastructure) for portfolio diversification. Kept deliberately simple — small parameter set (lookback, entry_z, exit_z, stop_z, risk_pct) to avoid overfitting and frozen-signal verification failures seen in over-parameterized strategies.

Implements the hypothesis with the repo's SyncedBuffer pairs helper (used by the cited ADA-DOT / ETH-BTC pairs) and the centralized multi-leg pattern. SyncedBuffer guarantees the AAVE/COMP ratio is computed on contemporaneous closes (it commits an (a,b) sample only when both legs report the same ts_ns), eliminating the independent-array-trim alignment bug; the base cross-leg barrier (both legs 4H) ensures calculate_signal fires only when COMP's bar for the current AAVE timestamp is present, so the buffer commits one aligned sample per bar. It uses log_ratio (more stationary and sign-symmetric than the raw ratio) z-scored over the lookback, and is two-sided per the hypothesis: short the outperformer / long the underperformer at |z| >= entry_z. Exits implement all three hypothesis conditions: reversion (|z| <= exit_z), a divergence stop (|z| >= stop_z) that caps the structural-regime-break tail risk inherent to pairs MR, and a time stop; entry_sign is re-inferred from the live AAVE position side if state is lost. Entry is dollar-neutral (each leg sized to equity*leg_frac/price, not equal qty) so P&L tracks the ratio reverting rather than crypto beta, and it is atomic -- if either leg fails its qty/min-notional check neither opens, preventing a naked directional bet. The sizing knob is exposed as risk_pct (per the hypothesis param list) and used as the per-leg notional fraction (0.45), keeping gross ~90% within the leverage-1.0 futures margin ceiling so leverage is not set >1 and the unused-leverage gate does not apply. min_bars_required=2 lets the buffer fill while zscore_last returns None until 180 aligned samples accrue, gating warmup without dead bars; the returned z-score is a continuous per-bar signal so Layer-2 frozen-signal detection passes while entries stay threshold-gated. AAVE and COMP are liquid Binance USD-M majors with multi-year 4H history (no data_unavailable risk), and the DeFi-lending sector diversifies the portfolio's existing L1 pairs. should_enter/should_exit/position_size are inert because the base entry path manages only the primary instrument.

AAVE-COMP DeFi-lending pairs MR is a market-neutral strategy with a thin, statistically-insignificant edge (Sharpe 0.27, sharpe_ci_low -0.70 with CI straddling 0, PF 1.06) and two structural problems optimization cannot fix. (1) CAPACITY WALL: impact_cost_pct 32.4% (market impact consumes a third of gross PnL) and capacity_usd only $954K — AAVE and COMP futures are too illiquid for the 45%-per-leg sizing, so the edge exists only at toy scale; reducing size doesn't materially lower the impact ratio and only shrinks the already-tiny capacity. (2) NON-STATIONARY SPREAD / DE-COINTEGRATION TAIL RISK: unlike the proven cointegrated siblings (ADA-DOT Sharpe 6.4, ETH-BTC 6.8), the AAVE/COMP ratio is not cleanly stationary (AAVE structurally outgrew COMP), so the spread trends rather than reverts — the modest +30% total is carried by a few tail-win days against a steady bleed plus violent blowups (2025-05-30 -26% in a single day on a 'dollar-neutral' pair, return_skew -3.76, kurtosis 68.4, max_DD 35.8% with ci_high 78%). Not optimize: a Sharpe-0.27/CI-through-0 base with a 32% impact wall optimizes into a still-undeployable, blowup-exposed strategy. Not iterate: neither the illiquidity nor the AAVE/COMP non-stationarity is addressable by the 5 parameters. FAILURE PATTERN: transplanting the proven pairs-MR mechanism onto a low-liquidity, weakly-cointegrated alt pair (AAVE-COMP) yields a 20x-weaker Sharpe than the cointegrated siblings, a capacity wall (32% impact, ~$1M capacity), and fat-left-tail de-cointegration blowups (-26% single day) — the same de-cointegrated/illiquid-pair signature as the abandoned ETH/BTC-Bollinger, LINK/ETH, and SOL-AVAX pairs. The mechanism is sound on stationary, liquid ratios; AAVE-COMP is the wrong target. The Research Lead should redirect the proven pairs mechanism to a more liquid, genuinely cointegrated pair rather than this DeFi-lending pair.

Dollar-neutral DeFi-lending pairs mean-reversion on AAVEUSDT (leg A) and COMPUSDT (leg B) Binance USD-M futures, 4H bars. On each timestamp-aligned bar it commits the (AAVE, COMP) close pair to a SyncedBuffer, forms the log(AAVE/COMP) ratio, and z-scores it over a 180-bar rolling window. When z >= +2.0 it shorts AAVE and longs COMP; when z <= -2.0 it longs AAVE and shorts COMP, each leg sized to equal USD notional (~45% of equity per leg). It closes both legs when |z| reverts to <= 0.5 (mean reversion captured), when |z| >= 4.0 (divergence stop / regime break), or after a 180-bar time stop. Pure OHLCV, no supplementary data, leverage 1.0 (~90% gross). Mirrors the ADA-DOT and ETH-BTC pairs applied to the DeFi-lending sector.

Genuinely market-neutral (beta -0.02, benchmark_correlation -0.07); positive in all 6 years; healthy 238-trade sample

Clean atomic two-leg pairs implementation; mechanism is proven on cointegrated siblings (ADA-DOT 6.4, ETH-BTC 6.8 paper)

Thin/insignificant edge: Sharpe 0.27, sharpe_ci_low -0.70 (CI straddles 0), PF 1.06 — vs 6+ on the working sibling pairs, indicating AAVE/COMP is far less cointegrated

Capacity wall: impact_cost_pct 32.4% (impact eats a third of gross), capacity_usd only $954K — AAVE/COMP futures too illiquid for this sizing; undeployable at scale

Violent de-cointegration tail risk: 2025-05-30 -26% in one day, return_skew -3.76, kurtosis 68.4, max_DD 35.8% (ci_high 78%), DD duration 672 days

Edge mostly carried by a few tail-win days against a steady bleed — not a robust reverting spread

AaveCompDefiLendingPairsSpreadMeanReversionLS extended the proven pairs-MR mechanism — paper-stage at Sharpe 6+ on ADA-DOT and ETH-BTC — to the DeFi-lending sector, betting the near-identical AAVE and COMP business models would keep their ratio stationary. The build was clean and genuinely dollar-neutral, but the edge came in ~20x weaker (Sharpe 0.27, CI straddling zero, PF 1.06) because AAVE structurally outgrew COMP, so the spread trended rather than reverting, producing a -26% single-day blowup and a 35.8% drawdown. On top of that, a 32% impact cost and ~$954K capacity made it undeployable at scale. The analyst abandoned it at the backtest-review gate, calling AAVE-COMP the wrong target — the same de-cointegrated/illiquid-pair signature as the abandoned ETH/BTC-Bollinger, LINK/ETH, and SOL-AVAX pairs — and recommending the mechanism be redirected to a more liquid, genuinely cointegrated pair.

The pairs mean-reversion mechanism is sound only on genuinely cointegrated, liquid ratios — transplanting it onto a weakly-cointegrated, low-liquidity alt pair yields a Sharpe ~20x weaker than the working siblings, a capacity wall, and fat-left-tail de-cointegration blowups, so the asset-pair selection (stationarity and liquidity) is what determines success, not the mechanism.

The analyst abandoned it at the backtest-review gate before optimization for two structural reasons optimization cannot fix: a capacity wall (market impact consuming a third of gross PnL on ~$1M capacity, undeployable at the 45%-per-leg sizing) and a non-stationary spread — AAVE structurally outgrew COMP, so the ratio trends rather than reverts, leaving a thin Sharpe carried by a few tail-win days against a steady bleed plus blowups.

A dollar-neutral, OHLCV-only pairs mean-reversion on AAVEUSDT and COMPUSDT 4H futures — the two largest DeFi lending tokens — z-scoring the log price ratio over a 180-bar window and, at |z| ≥ 2, shorting the outperformer while longing the underperformer in equal dollar notional, exiting on reversion to the mean, a divergence stop, or a time stop, mirroring the paper-stage ADA-DOT (Sharpe 6.4) and ETH-BTC (Sharpe 6.8) pairs in a new DeFi-lending sector.

The build was genuinely market-neutral (beta -0.02) and positive in all 6 years over 238 trades, but the edge was thin and insignificant: 30.1% total return (CAGR 5.7%), Sharpe 0.27 with a CI low of -0.70, profit factor 1.06 — versus 6+ on the working sibling pairs. It also showed a 35.8% max drawdown (CI high 78%), violent de-cointegration tails (a -26% single day, skew -3.76, kurtosis 68.4), 32.4% impact cost, and only $954K capacity.