SolHyperliquidOneHourTimeSeriesMomentumVolTargetedLS

SOL Hyperliquid Perp 1H Time-Series Momentum, Volatility-Targeted (Single-Instrument, Long-Short, Retail-Trend Harvest)

A long-SHORT, SINGLE-INSTRUMENT, single-venue time-series momentum (trend) strategy on SOLUSD.HYPERLIQUID perpetual futures using 1-HOUR OHLCV bars only (NO funding/OI/liquidation/IV supplementary data, NO second instrument, NO CASH-spot path, MARGIN account ~0.09% RT). Direction is set by the sign of a single lookback return; position SIZE is set by volatility targeting (scale notional inversely to recent realized volatility to hit a constant risk target), which is the key differentiator from a fixed-size or ATR-stop trend system and directly prevents the un-normalized-sizing blowups (e.g. the 81% drawdown that sank the AVAX funding-momentum strategy). Only 2 core parameters (momentum lookback, vol-target) plus a vol estimator window — deliberately minimal to dodge the overfit signature that killed the multi-parameter breakout/veto strategies this session. Risk profile: target ~25% annualized volatility per position, notional = (target_vol / realized_vol) * equity, hard-capped at 2x leverage and 30% of equity max single-position notional; risk realized through the vol scaling itself plus a wide 4.0x ATR(24) catastrophe stop; single position at a time; flip on momentum sign change. Low-to-moderate turnover (1H TSM holds many hours to days) keeps fee drag modest vs the 0.09% RT cost. Fills the largest venue gap (HYPERLIQUID 6.1%->20% target), the direction gap (long_short 14.3%), and nudges the under-weight mid/short horizon buckets, while staying on the only structurally-clean, testable HL path (single-instrument OHLCV on a major with dense ~2.5yr 1H history).

Single-instrument long-short TSM using the standard FactoryStrategy hooks. Direction is the pure sign of the lookback return (calculate_signal returns the continuous momentum, which varies every bar so Layer-2 frozen-signal detection passes; the sign decision is in should_enter). Sizing is the hypothesis's core differentiator -- volatility targeting: scale = target_vol/realized_vol where realized_vol is the annualized std of hourly log returns (x sqrt(8760)); notional = scale x equity, which is the actual risk control that prevents un-normalized-sizing blowups. On SOL 1H, realized vol annualizes to ~60-150%, so scale ~ 0.25/0.9 ~ 0.28 and the position sits ~25-30% of equity, essentially never reaching 100%; I therefore set leverage = 1.0 with a 30% notional cap rather than the hypothesis's nominal 2x, because at this vol the 2x cap never binds -- a leverage>1 would be inert and trip the leverage_set_but_unused gate, and 1.0 produces byte-identical behavior since the cap is never reached. leverage is still referenced in sizing (as a non-amplifying cap) so there is no unused-leverage flag. Exits flip the position on a momentum sign change and apply a wide 4x ATR(24) catastrophe stop (entry price/ATR captured on entry, reconstructed from avg_px_open if state is lost) -- the stop is a backstop, not the primary risk lever. Sizing clears the Hyperliquid $10 min-notional and the ~0.09% RT taker cost is modest given multi-hour-to-day TSM holds. Uses the catalog HL symbol SOLUSD.HYPERLIQUID with dense ~2.5yr 1H history, single-instrument OHLCV (no funding/OI/IV/second-leg dependencies), filling the HYPERLIQUID venue and long_short direction gaps on the structurally-clean HL path. min_bars_required guarantees the momentum, vol, and ATR windows are all available before any trade.

Two independent disqualifiers. (1) DATA COVERAGE: the hypothesis claimed dense ~2.5-year 1H Hyperliquid SOL history, but the backtest runs only 2025-10-11 to 2026-05-22 (~221 days, 5,357 1H bars) — only ~7 months actually exist/loaded, far too short for a 3-window walk-forward and covering a single SOL regime. (2) DECISIVELY NEGATIVE SIGNAL: over that window the strategy loses -33.3% (CAGR -49.8%) with Sharpe -3.55 and the ENTIRE CI below zero (-6.46 to -1.00), PF 0.45, win rate 23.7%, expectancy -$154/trade, 23 consecutive losses, omega 0.36, probabilistic_sharpe 0.0065. The 1H time-series-momentum signal whipsaws badly on SOL — the 23.7% win rate from momentum-sign-flip exits shows the signal is chopped to pieces — and commission is 17.4% of gross. The vol-targeting differentiator DID work structurally (avg_position 30.6%, max DD 37% rather than the feared 81% blowup), but sizing a losing signal smaller does not create an edge. No lookback/target-vol setting turns a strongly negative, whipsaw-prone signal positive, and 7 months cannot support the walk-forward, so optimization would waste 2 hours. FAILURE PATTERN: short-horizon (1H) single-asset time-series momentum on a high-vol crypto major whipsaws (sub-25% win rate, PF <0.5) and bleeds ~17% of gross to fees; vol-targeting fixes the sizing-blowup tail risk but not the absent/negative edge, and the Hyperliquid SOL 1H window is too short (~7 months) to validate regardless.

Long-short single-instrument time-series momentum on SOLUSD.HYPERLIQUID perpetual futures, 1-hour bars, OHLCV-only. Direction is the sign of a 72-bar (~3-day) lookback return; position size is volatility-targeted: notional = (target_vol 25% / realized_annualized_vol) x equity, capped at 30% of equity, so exposure scales down in turbulent regimes and up in calm ones (the un-normalized-sizing blowup fix). It holds a single position, flips on a momentum sign change, and exits early on a wide 4x ATR(24) catastrophe stop. Two core parameters (mom_lookback, target_vol) plus a vol estimator window. Leverage 1.0.

Vol-targeting differentiator worked structurally: avg_position 30.6%, max DD 37% (CI high 54%) — no un-normalized-sizing blowup, the stated design goal

Clean single-instrument HL OHLCV path, no supplementary-data or multi-leg blockers; minimal 2-param design

Data window only ~7 months (2025-10 to 2026-05), NOT the claimed 2.5 years — too short for a meaningful 3-window walk-forward and covers a single SOL regime

Decisively negative: total return -33.3%, CAGR -49.8%, Sharpe -3.55 with entire CI below zero (-6.46 to -1.00), PF 0.45, win rate 23.7%, expectancy -$154/trade, 23 consecutive losses

1H TSM whipsaws on SOL: 23.7% win rate from momentum-sign-flip exits; commission 17.4% of gross compounds the loss

Both partial years negative (2025 -22.6%, 2026 -14.5%); omega 0.36, sortino -4.44

>0 to be viable

>730 for walk-forward

SolHyperliquidOneHourTimeSeriesMomentumVolTargetedLS aimed to harvest retail trend on Hyperliquid SOL with a minimal two-parameter 1H momentum signal whose key innovation was volatility-targeted sizing, explicitly to prevent the un-normalized-sizing blowups that had sunk prior strategies. The sizing fix worked exactly as designed — average position 30.6% and a 37% max drawdown rather than a catastrophic blowup — but it could not rescue an absent edge: over a far-too-short ~7-month window (not the claimed 2.5 years) the 1H signal whipsawed to a 23.7% win rate, profit factor 0.45, and a Sharpe of -3.55 with its entire CI below zero, bleeding 17% of gross to fees. The analyst abandoned it at the backtest-review gate on its first iteration, citing both the insufficient data coverage and the decisively negative, whipsaw-prone signal. The lesson recorded was that vol-targeting controls tail risk but does not manufacture an edge, and short-horizon single-asset momentum on a high-vol major is structurally chopped to pieces.

Short-horizon (1H) single-asset time-series momentum on a high-vol crypto major whipsaws (sub-25% win rate, PF <0.5) and bleeds ~17% of gross to fees — volatility targeting fixes the sizing-blowup tail risk but does not create an edge, since sizing a losing signal smaller still loses; and a venue-gap-driven hypothesis must verify the claimed data history actually exists before building on it.

The analyst issued an 'abandon' verdict at the pre-optimization backtest-review gate on two independent disqualifiers: the Hyperliquid SOL 1H data was only ~7 months (not the claimed ~2.5 years), far too short for a 3-window walk-forward and covering a single regime; and the 1H momentum signal whipsaws badly on SOL (23.7% win rate, PF 0.45) for a robustly negative edge that no parameter can turn positive.

A long-short, single-instrument time-series momentum (trend) strategy on SOLUSD.HYPERLIQUID 1H OHLCV-only perpetual futures that set direction by the sign of a single lookback return and set position SIZE by volatility targeting (notional scaled inversely to realized vol to hit ~25% annualized risk, capped at 30% of equity), flipping on momentum sign change with a wide 4x ATR catastrophe stop — the vol-targeting being the deliberate fix for un-normalized-sizing blowups.

Over only ~7 months (221 days, 219 trades) it was decisively negative: total return -33.3%, CAGR -49.8%, Sharpe -3.55 with the entire CI below zero (-6.46 to -1.00), profit factor 0.45, win rate 23.7%, expectancy -$154/trade, 23 consecutive losses, both partial years negative, with commission at 17.4% of gross — though the vol-targeting did work structurally (avg position 30.6%, max drawdown 37% rather than a feared 81% blowup).