DogeAdaOneHourLeadLagConvergenceLongShort

DOGE-ADA 1H Lead-Lag Convergence Long-Short: Trade ADA in the Direction of DOGE's Prior Large Move

A long-short, multi-instrument lead-lag convergence strategy on DOGEUSDT.BINANCE and ADAUSDT.BINANCE — both confirmed testable per recent pipeline successes. Structurally distinct from every prior portfolio strategy: not pairs cointegration (statistical equilibrium), not cross-sectional MR (symmetric ranking), not Keltner range trading (single-instrument), not vol-targeting (sizing). This is a DIRECTIONAL CROSS-ASSET INFORMATION TRANSMISSION strategy — when one asset leads a directional move, the OTHER (laggard) tends to converge in the same direction within 1-3 bars. The mechanism (Lo-MacKinlay 1990 'When Are Contrarian Profits Due to Stock Market Overreaction?'; Hou-Moskowitz 2005 'Market Frictions, Price Delay, and the Cross-Section of Expected Returns'; Caporale et al 2018 crypto-specific extension): in correlated-asset universes, attention/flow disparities cause one asset to react faster to common-factor news. Within memecoin/alt-L1 pairs like DOGE-ADA, DOGE (higher attention beta) typically leads ADA on cycle-rotation moves; ADA (slower flow) plays catch-up. By detecting a large 1H DOGE move and trading the SAME direction on ADA, the strategy captures the information-transmission edge. Symmetric: if ADA moves first (rarer but happens during ADA-specific events), the strategy trades DOGE in ADA's direction. Single-bar entries with tight time-based exits prevent the strategy from being trapped if the lead-lag fails to materialize. Uses ONLY DOGE and ADA 1H bars — both confirmed working. Fills critical gaps: (1) NEW mechanism class (lead-lag convergence, not present in portfolio), (2) MULTI-INSTRUMENT scope (extends 19.4% pairs/multi category), (3) LONG-SHORT direction (15.2% → toward 45%), (4) SHORT-HORIZON timeframe (mid_30m_2h at 10.6% — under-quota), (5) confirmed-testable instruments (avoids audit_stale across majors). Position sizing: 12% of equity per trade. Risk per trade: 0.8% of equity capped by tight time-based ATR stop.

Implements a new mechanism class for the portfolio — directional cross-asset information transmission (lead-lag convergence; Lo-MacKinlay 1990, Hou-Moskowitz 2005, Caporale 2018) — structurally distinct from cointegration pairs (equilibrium), cross-sectional ranking (symmetric), single-instrument range trading, and vol-targeting. The edge: in correlated alt pairs, the higher-attention asset (DOGE) reacts faster to common-factor news and the laggard (ADA) catches up in the same direction within 1-3 bars, so the strategy detects a large dominant 1H move in one leg and trades the OTHER leg in that direction. It is symmetric (either asset can lead) and goes both long and short, requiring the futures venue for shorts; at only 12% notional per trade leverage stays 1.0 (no unused-leverage gate, no amplification). Correct multi-instrument handling is the crux: both legs are 1H, so I evaluate on the aligned timestamp (triggered from whichever leg arrives second, gated by matching max-ts across legs and a last-eval dedupe) and read contemporaneous prices from a timestamp-keyed store, avoiding the stale-leg pitfall. Tight exits are the risk control — lead-lag edges decay quickly, so a 3-bar time stop plus a 1.5x-ATR stop prevent being trapped when convergence fails. Layer-2 compliance is preserved: on_extra_bar ingests the DOGE leg directly and calculate_signal returns a close-varying monitoring value while trades route through _evaluate. Both DOGE and ADA 1H are confirmed-testable, avoiding audit_stale.

Lead-lag convergence premise is falsified and the strategy is decisively losing: profit factor 0.77, Sharpe -2.58 with a fully-negative confidence interval (-3.68 to -1.58), expectancy -$23.4/trade, total return -35.1%, and NEGATIVE returns in every single year 2020-2026 across an ample 2,294-trade sample. The directional bet (trade ADA in DOGE's prior-move direction) has no edge at the 1H horizon — the laggard does not reliably continue in the leader's direction — and it is additionally fee-toxic (commission 12.32% of gross plus $17.4k modeled impact from high-frequency 1H churn). This is structural, not a tunable-parameter issue: optimization cannot rescue a mechanism that loses gross-of-fees in every regime, so it is not iterate. FAILURE PATTERN: cross-asset directional lead-lag 'convergence' on 1H crypto pairs has no exploitable edge after costs and loses consistently across all regimes.

Multi-instrument long-short lead-lag convergence on DOGEUSDT + ADAUSDT Binance USD-M futures, 1H bars. Each aligned hour it computes both assets' 1-bar returns: if DOGE prints a large move (|ret| >= 2%) that dominates ADA's (>= 1.5x), it enters ADA in DOGE's direction (the laggard converges to the leader); symmetrically, if ADA leads, it enters DOGE in ADA's direction. Single directional leg on the laggard, capped at 12% of equity notional, with a time stop (3 bars) and a 1.5x-ATR(14) stop. Pure OHLCV, two instruments, no supplementary data. Leverage 1.0 (margin venue only needed for the short side).

Correct, clean implementation of the intended lead-lag mechanism (timestamp-aligned dual-leg evaluation, both directions, ATR/time stops); ample sample (2,294 trades).

Decisively losing: profit factor 0.77, Sharpe -2.58 with the entire CI negative (-3.68 to -1.58), expectancy -$23.4/trade, total return -35.1%.

Negative in EVERY year 2020-2026 — no regime where the edge exists.

Fee-toxic: commission 12.32% of gross (2,294 1H trades) plus $17.4k modeled impact; high-frequency directional churn.

Premise falsified: the laggard (ADA) does not reliably continue in the leader's (DOGE) direction at the 1H horizon.

DogeAdaOneHourLeadLagConvergenceLongShort tested a directional cross-asset thesis — trade ADA in the direction of DOGE's prior large 1H move (and symmetrically) — drawing on lead-lag literature and aiming to fill several pipeline category gaps at once. The mechanism was implemented cleanly and generated a large 2,294-trade sample, but the results were unambiguously bad: profit factor 0.77, Sharpe -2.58 with an entirely negative CI, -35.1% total return, and losses in every year 2020-2026, compounded by commissions eating 12.32% of gross. The analyst abandoned it at backtest review after one iteration, concluding the convergence premise was simply falsified at the 1H horizon and could not be rescued by optimization. The logged failure pattern is that 1H crypto cross-asset directional lead-lag has no edge after costs and loses across all regimes.

Cross-asset directional lead-lag 'convergence' on 1H crypto pairs has no exploitable edge after costs; an academically-motivated information-transmission thesis must still be validated against fees and across regimes before assuming the laggard follows the leader.

The analyst issued an 'abandon' verdict at the pre-optimization backtest-review gate, judging the core premise falsified — the laggard does not reliably continue in the leader's direction at the 1H horizon — and the mechanism loses gross-of-fees in every regime, so it was not a tunable-parameter issue and optimization was never attempted.

A long-short, multi-instrument lead-lag convergence strategy on DOGEUSDT and ADAUSDT 1H futures bars that, after detecting a large dominant 1H move in one asset (usually DOGE as the leader), entered the laggard (usually ADA) in the same direction with tight time and ATR stops, betting the slower asset would converge within 1-3 bars.

Across an ample 2,294-trade sample over ~1,025 days it lost decisively: profit factor 0.77, Sharpe -2.58 with a fully-negative confidence interval (-3.68 to -1.58), expectancy -$23.4/trade, total return -35.1%, and 37.0% max drawdown — with negative returns in every year from 2020 to 2026. It was also fee-toxic, with commissions at 12.32% of gross plus ~$17.4k of modeled impact from high-frequency 1H churn.