APFF: Adaptive Factor Evaluation for Multi-Asset Perpetual Futures
Summary
The document describes APFF, a cross-sectional long-short strategy for USDT-margined perpetual contracts. It ranks a universe of contracts using five seed factor families: momentum, short-term reversal, funding rates, premium, and open interest. Raw signals are filtered, winsorized, and converted to cross-sectional ranks before being combined. The strategy scales exposure by volatility, limits individual positions, applies a limited BTC beta hedge, and preserves unused factor budget when data for some factors is unavailable.
APFF stores each decision and evaluates factors on matured forward samples, with candidates moving through observation, small-weight trial, active, or inactive states. The article details coverage, cost, stability, correlation, and multiple-testing checks, as well as order reconciliation, drawdown controls, and shadow, paper, and live modes. It reports engineering checks and simulated-exchange operation, but explicitly says these do not establish long-term returns or predictive skill. Historical point-in-time data, independent out-of-sample validation, and realistic cost calibration remain incomplete; paper results also differ from live execution.
Key ideas
- APFF combines ranked momentum, reversal, funding, premium, and open-interest signals across perpetual contracts.
- Factor weights and total exposure adapt to forward evidence, while unavailable factor budgets remain unused.
- New factors must pass separate observation and trial stages before gaining meaningful portfolio weight.
- Execution includes order-intent persistence, order and position reconciliation, turnover limits, and drawdown controls.
- The reported checks validate engineering behavior, not long-term profitability or factor predictive power.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.