By Keel Research Team · Updated May 12, 2026
Antonacci-style dual momentum applied to Hyperliquid perpetuals: long only when the asset is beating both its peer cohort (relative momentum) AND cash (absolute momentum). Fewer signals than vanilla trend following, materially better drawdowns.
noindex is set in metadata).Dual momentum was formalized by Gary Antonacci in 2014 as a refinement of vanilla momentum strategies. It introduces a second filter on top of standard relative-momentum ranking: absolute momentum. A buy signal requires both conditions:
If either fails, the strategy goes flat. The combination is stricter than vanilla trend following, which produces signals based on price action alone — dual momentum demands the asset be both better than peers and better than nothing.
The practical effect: fewer signals, lower turnover, and meaningfully smaller drawdowns during regime transitions. Bull markets give back some upside (the strategy is sometimes flat when vanilla trend would have been long); bear and chop regimes outperform substantially because the absolute-momentum filter pulls the strategy out before drawdowns get deep.
PriceDataLoader pulls historical OHLCV across HL top-100 universe. Lookback windows: 3-month and 6-month rolling returns.
Rank each asset by 6-month ROC vs the cohort. Top quartile qualifies as a relative-momentum winner.
Filter the relative-momentum cohort by absolute return > 0 over the lookback. Assets that pass both filters become candidates.
Equal-weight (or vol-weighted) top-N qualifiers. Rebalance hourly to manage funding; flat positions in any asset that drops out of either filter.
Real backtest numbers populate once the source run is selected. Page renders with noindex until then.
Equity curve renders here once the backtest is selected.
Works: regime transitions. Dual momentum’s edge over vanilla trend following is largest exactly when markets turn — the absolute-momentum filter pulls the strategy to cash before drawdowns get severe. Bear markets and chop regimes are where the strategy materially outperforms.
Underperforms: raging bull markets. When everything is up and the trend is one-directional, the absolute-momentum filter rarely fires, so dual momentum stays largely fully invested — but the relative-momentum filter still keeps positions concentrated in top-cohort names, which can mean missing the broader rally. Vanilla trend following would have caught more of the rally.
Tail risk: sudden gap moves between rebalance windows. If the absolute-momentum filter checks daily and the market drops 30% intraday, the strategy can still hold the position until next rebalance. Hourly rebalancing on HL mitigates this vs Antonacci’s original monthly cadence on equity ETFs.
Open the strategy in your own Keel workspace. The pipeline, lookback windows, sizing, and rebalance cadence are editable. Run your own backtest, optimize, deploy live.
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.
Dual momentum is a strategy formalized by Gary Antonacci in 2014. It requires two conditions for a buy signal: (1) relative momentum — the asset must be in the top quartile (or similar) of its peer cohort over a lookback window, AND (2) absolute momentum — the asset must have outperformed cash / risk-free over the same window. Both must agree. Fewer false positives than vanilla trend following at the cost of fewer signals overall.
Relative momentum alone can keep you long during a broad bear market (you're long the "least-bad" asset, still losing). Absolute momentum alone can buy a struggling asset just because its 6-month return is barely positive. Combining filters out both failure modes: stay long only when the asset is beating peers AND beating cash. The strategy goes flat (or to cash equivalent) when either condition fails.
Hyperliquid is on-chain, so "cash" is just USDC sitting unallocated. The absolute-momentum filter compares the candidate asset's return against zero (or a stablecoin yield baseline). If the asset hasn't outperformed staying flat in USDC, the strategy skips it — even if it's the best of a bad cohort.
Dual momentum is a stricter, more disciplined variant of trend following. It produces fewer signals, smaller drawdowns, and lower turnover — usually at the cost of slightly lower upside in strong bull markets where vanilla trend would have been fully exposed. The trade-off is consistency: dual momentum tends to underperform in extended bull regimes but materially outperforms in bear/regime-switching environments.
The Keel implementation: every 4h (or daily), rank the top-100 HL universe by 3-month or 6-month ROC. Filter by absolute momentum (return > 0 over the window). Hold equal-weighted (or vol-weighted) positions in the top-N qualifiers; flat on the rest. Hourly rebalance to manage funding costs. See the pipeline below for the full component graph.
Vanilla trend — fewer filters, more signals. The strategy dual momentum builds on.
Different edge: short high-funding perps, market-neutral. Real backtest — Sharpe 2.17.
Live cohort momentum ranking — input to the relative-momentum filter.