By Keel Research Team · Updated May 12, 2026
Time-series and cross-sectional momentum applied to the Hyperliquid perp universe. Buy strength, hold winners, exit on reversal — the most studied edge in financial markets, adapted for HL's 24/7 + hourly-funding mechanics.
noindex is set in metadata).Trend following is the practice of holding positions in the direction of an asset’s prevailing movement — long assets going up, flat or short assets going down — and exiting when the trend reverses. It is one of the oldest and most consistently documented edges in financial markets, going back decades in commodities and equities and proven robust across asset classes, time horizons, and market regimes (with the important caveat that it underperforms in choppy, mean-reverting environments).
On crypto, trend strategies have produced some of the strongest Sharpe ratios on record. Crypto’s 24/7 volatility, regular regime shifts (bull → bear → recovery → expansion), and persistent momentum within regimes are ideal conditions for trend signals — when the model gets the regime right, the moves are large and persistent. The trick is staying out (or smaller) during the chop.
The strategy this page documents combines time-series trend (each asset’s own price relative to its moving average) with cross-sectional momentum (ranking the cohort) — both signals must agree before entering a position. Positions size by inverse-volatility and rebalance on the hourly HL funding cycle.
PriceDataLoader pulls hourly OHLCV across HL top-100 universe. TargetTimeframeResample aligns to the strategy's chosen bar (default 4h).
Two-part: time-series (price vs N-bar moving average) AND cross-sectional (cohort ROC percentile rank). Both must agree.
Inverse-volatility weights — bigger position in low-vol trends, smaller in high-vol. Portfolio-level vol target caps total exposure.
Funding-aware order placement; rebalance hourly. Exit when either trend signal flips OR portfolio drawdown breaches threshold.
Real backtest numbers populate this card once the source run is selected from Keel’s strategy registry. The page renders with noindex until then, so placeholder values never reach Google.
Equity curve renders here once the backtest is selected.
Works: persistent directional regimes — strong bull markets (long signals fire repeatedly, positions hold for weeks), structural bear markets (short signals work), and trending alt-cycles where rotation creates clear winners and losers. Historical Sharpes of 1–2+ on crypto trend strategies during these regimes are common.
Fails: chop. Sideways, low-momentum markets generate frequent false signals — the strategy buys breakouts that fail, sells breakdowns that recover, and bleeds via funding cost + slippage on each whipsaw. Regime detection (the /lab/regime page exists for this) is the standard overlay; many trend strategies reduce position size or stop entirely when breadth + volatility indicate “mixed” regime.
Tail risk: sudden regime flips. A trend strategy long going into an unexpected bear capitulation can give back months of gains in a few hours. Risk management — stops, position-size caps, max-drawdown circuit breakers — is the difference between a profitable trend strategy and one that blows up on the first major flip.
Open the strategy in your own Keel workspace. The pipeline, signals, sizing, and regime filters are all editable. Run your own backtest, optimize parameters, deploy live to Hyperliquid.
Keel is a Strategy OS for AI-assisted systematic trading on Hyperliquid. Backtest, optimize, and run live strategies across single-stock perps, indices, and crypto majors — realistic fees, slippage, and funding modeled.
Free to start — connect a Hyperliquid wallet when you’re ready to go live.
Trend following bets that assets which have been moving in a direction continue moving in that direction — buy strength, sell weakness, hold winners until they reverse. The simplest robust edge in financial markets, and one of the most studied.
HL's 24/7 markets + on-chain settlement + hourly funding mean trend signals can be evaluated continuously, not just on US market hours. The HL universe (~150 crypto perps + HIP-3 equity perps) gives enough breadth to build cross-sectional trend signals — long the top-quartile-momentum cohort, short or stay flat on the bottom-quartile cohort, rebalance hourly or daily.
Mean-reverting regimes. When markets chop sideways with no persistent direction, trend signals fire false breakouts and the strategy bleeds via whipsaws + funding cost. Risk-management overlays (volatility filter, regime detector, tighter stops in low-vol regimes) are how production trend strategies survive these periods. The /lab/regime indicator gives one input to this filter.
Several ways, all valid. Time-series: an asset's own moving-average crossover (e.g. 20/50 SMA), ROC vs an N-bar lookback, or trend-cleanliness (R² of log-price regression). Cross-sectional: ranking the cohort by ROC, MACD slope, or composite momentum signals and going long the top quartile. Production strategies usually combine both — only enter long if the asset is in its own uptrend AND in the cohort's top quartile.
Dual momentum is a structured variant of trend following (Antonacci-style): it requires both relative momentum (asset beating its peers) AND absolute momentum (asset beating cash / risk-free). Stricter than vanilla trend, fewer false positives, but lower opportunity set. See /strategies/dual-momentum for the dedicated walk-through.
Different edge: short high-funding perps, hedge directional exposure. Real backtest — Sharpe 2.17, +79.6% return.
Stricter cousin: relative + absolute momentum. Antonacci-style.
The live cohort this strategy reads — top-quartile ROC + MACD slope + rel vol on HL top-100.