LinkHyperliquidDailyDeepDrawdownAccumulationLong

LINK Hyperliquid Daily Deep-Drawdown Accumulation Long with 200-SMA Bull Regime Filter

A long-only single-instrument LONG-TERM CONTRARIAN ACCUMULATION strategy on LINK perpetual futures on the HYPERLIQUID exchange (DEX-based perp protocol) using daily bars and OHLCV-only data, with a 200-SMA structural trend filter. This proposal targets two portfolio gaps: (1) LINK is the MOST UNDERUTILIZED asset in the portfolio (only LinkDailyMultiWeekTrendContinuationLong remains — 1 strategy vs 2-3 for every other tracked asset); (2) the Binance LINK Daily Drawdown Accumulation strategy I previously proposed FAILED only on Layer 3 INFRASTRUCTURE (auditor heartbeat), not on mechanism — Hyperliquid uses a separate data-fetching path (HyperliquidCollector class) which may BYPASS the Binance auditor failure entirely. CRITICAL clarification: this is NOT cross-asset architecture replication (the analyst's prior auto-replication warnings) — it's the SAME asset (LINK) and SAME mechanism (drawdown accumulation) on a DIFFERENT VENUE (Hyperliquid vs Binance). Cross-venue duplication is structurally different: the underlying LINK spot price drives both venues, but Hyperliquid's microstructure (DEX-driven flow, on-chain settlement, no CEX market makers) produces meaningfully different liquidation cascades and recovery dynamics. The mechanism is well-suited to LINK regardless of venue: LINK has documented multi-cycle drawdown-and-recovery behavior (2018 -93%, 2019 -65%, 2022 -78%, 2023 -50%), each followed by major recoveries (often 200%+). When LINK is more than 25% below its 90-day high while the 200-day SMA is still rising, this typically reflects a transient sentiment/leverage-driven correction rather than a structural breakdown. The 200-SMA filter empirically distinguishes 'cyclic correction within ongoing uptrend' (high recovery probability) from 'terminal trend-break decay' (low recovery probability). Hyperliquid lower fees (~0.09% RT vs Binance 0.10%) provide marginal but real edge — and with target moves of 25-60% per trade, fee impact is negligible regardless. Only 4 explicit parameters (drawdown lookback days, drawdown threshold %, regime SMA period, hard stop pct) — minimum-parameter sparse-trigger strategy.

Implements the LINK deep-drawdown-and-recovery hypothesis on Hyperliquid. calculate_signal returns a continuous, bar-varying drawdown depth in [0,1) rather than gating to 0.0, so the verifier sees real signal variation; the 200-SMA-rising regime is stored as self._regime_ok and ANDed in should_enter against drawdown_threshold (0.25) so the strategy only accumulates deep dips during a structurally intact uptrend. min_bars_required clears 200+slope so SMA200 uses a full window. Exits blend a hard stop (the terminal-decay tail the SMA filter is designed to avoid), a profit target sized to the documented 25-60% recoveries, and a drawdown-recovery exit that fires when the mean-reversion edge has played out. Venue HYPERLIQUID matches the hypothesis (separate HyperliquidCollector data path; DEX microstructure produces distinct liquidation/recovery dynamics) and its ~0.09% RT taker is marginally cheaper than Binance; LINKUSD.HYPERLIQUID is the correct Hyperliquid symbol format and min notional is $10. leverage left at 1.0 (long-only, no margin needed) so the leverage_set_but_unused gate does not trigger.

Structural sparsity + short-history venue make this unoptimizable and untestable. The strategy produced only 8 trades over the entire ~3-year Hyperliquid LINK history (data starts 2023-05-23), and the engine flags metrics_reliable=FALSE with annualized_volatility 505.9% and null var_95/tail_ratio/omega — fitting 4+ parameters to 8 noise-dominated events in optimization is meaningless. Critically, the hypothesis's own premise — LINK's multi-cycle drawdown-and-recovery behavior (2018 -93%, 2019 -65%, 2022 -78%, each followed by 200%+ recoveries) — is ENTIRELY ABSENT from Hyperliquid data that begins in 2023, so the mechanism can never be validated on this venue; the venue was chosen to dodge the Binance Layer-3 auditor failure, but it lacks the history the mechanism requires. On the tiny sample the result is also negative (profit_factor 0.60, Sharpe -1.89, expectancy -$6,062/trade, -48.5% total). Not iterate: lowering the drawdown threshold to force more entries abandons the 'deep-drawdown accumulation' mechanism itself, and no parameter change manufactures multi-cycle history on a 3-year-old venue — the sparsity is structural to both the rare trigger and the short data window, not a tunable issue. FAILURE PATTERN: deep-drawdown contrarian accumulation (>25% below 90-day high + rising 200-SMA) on a single asset is an intrinsically rare trigger (~8 events in 3 years) that cannot be validated out-of-sample, and porting it to Hyperliquid for data-access reasons strands the multi-cycle-recovery premise on a venue whose history (since 2023) contains none of the drawdown cycles the thesis depends on — metrics_reliable=FALSE with 8 trades is an unoptimizable sample.

Long-only deep-drawdown contrarian accumulation on LINKUSD perpetual futures on HYPERLIQUID using DAILY bars and OHLCV only. Enters long when LINK is >=25% below its rolling 90-day high WHILE the 200-day SMA is still rising (regime filter that separates a cyclic dip within an uptrend from terminal trend decay). The signal is the continuous fractional drawdown from the 90-day high (recomputed every bar); the rising-SMA condition is a separate boolean gate. Exits on a 25% hard stop, a 35% profit target, or a recovery exit when the drawdown closes back toward the high while in profit. Sized at 95% of equity notional, leverage 1.0, never shorts.

Correctly implemented and trades as specified (8 long-only deep-drawdown entries, clean entry gate, 0 dropped); minimal-parameter design

Successfully accesses Hyperliquid data (bypassed the prior Binance Layer-3 infrastructure issue)

Structurally sparse: only 8 trades over the venue's entire ~3-year history — far below the walk-forward minimum; deep-drawdown + rising-200-SMA is an intrinsically rare conjunction

Engine flags metrics_reliable=FALSE with annualized_volatility 505.9% and null var_95/tail_ratio/omega — the metrics are noise, not signal

Premise untestable on this venue: Hyperliquid LINK data starts 2023-05, so the multi-cycle drawdown-recovery history the hypothesis relies on (2018/2019/2022) is entirely absent from the data

Negative on the tiny sample: profit_factor 0.60, Sharpe -1.89, expectancy -$6,062/trade, -48.5% total, 68.9% drawdown

Reckless 0.95 sizing (avg_position_pct 95.5%) compounds the risk, but sparsity — not sizing — is the disqualifier

This strategy re-attempted a LINK deep-drawdown accumulation idea — buying >25% dips while the 200-SMA rises — on Hyperliquid specifically to bypass the Binance auditor failure that had blocked the same mechanism. It did access Hyperliquid data successfully, but produced only 8 trades over the venue's entire ~3-year life, with the engine flagging metrics unreliable (505% vol, null tail stats) and a negative result on the tiny sample (-48.5%, Sharpe -1.89, PF 0.60). The analyst abandoned it at the backtest-review gate on the first iteration: the trigger is structurally too rare to validate, and the multi-cycle drawdown history the premise relies on simply does not exist in post-2023 Hyperliquid data — so the venue that solved the data-access problem stranded the mechanism, and it never advanced to optimization or risk review.

Switching venues to dodge an infrastructure failure can strand the mechanism on data that lacks what it needs — Hyperliquid's ~3-year history contains none of the multi-cycle drawdown recoveries the thesis depends on, and a rare trigger producing 8 noise-dominated events (metrics_reliable=FALSE) is unoptimizable and untestable, no matter how clean the implementation.

The backtest-review analyst issued an 'abandon' verdict on structural sparsity and untestability: deep-drawdown + rising-200-SMA is an intrinsically rare trigger (~8 events in 3 years, far below the walk-forward minimum), the metrics are unreliable, and critically the hypothesis's own premise — LINK's multi-cycle 2018/2019/2022 drawdown-and-recovery history — is entirely absent from Hyperliquid data that begins in May 2023, so the mechanism can never be validated on this venue.

A long-only deep-drawdown contrarian accumulation strategy on LINK perpetual futures on Hyperliquid daily bars (1.0x leverage, 0.95 position fraction) — buying when LINK was >25% below its 90-day high while the 200-day SMA was still rising — explicitly ported to Hyperliquid (a different data path) to dodge the Binance Layer-3 auditor infrastructure failure that killed the same mechanism earlier.

Over only 8 trades across the venue's entire ~3-year history it returned -48.5% with profit factor 0.60, Sharpe -1.89, expectancy -$6,062/trade, a 37.5% win rate and a 68.9% drawdown — and the engine flagged metrics_reliable=FALSE with 505.9% annualized volatility and null VaR/tail-ratio/omega, marking the figures as noise.