Adaptive Factor Evaluation for a Multi-Asset Perpetual Futures Strategy

Summary

APFF is a framework for ranking a changing universe of crypto perpetual futures and combining interpretable signals into a long-short portfolio. Its seed factors cover momentum, short-term reversal, funding, premium, and open-interest behavior. Cross-sectional ranks put different signals on a common scale; volatility scaling, symbol caps, and a limited BTC hedge help manage exposures. Decisions and factor evaluations run on slower schedules than market-data collection.

The central research process records forward returns, waits for samples to mature, and updates factor budgets gradually. New candidates pass through observation and small-weight trials before gaining more capital. The article describes engineering checks and simulated runtime scenarios, but explicitly says these do not establish long-term returns or predictive power. It also identifies limitations, including incomplete historical point-in-time universe data, simulated premium proxies that differ from production data, and execution costs and risk controls that still need further work. Independent out-of-sample evidence remains necessary.

Key ideas

• APFF ranks multiple perpetual contracts by combining normalized cross-sectional scores from five seed factor families.
• Volatility scaling, per-symbol exposure caps, and a limited BTC hedge address risks that equal long and short notionals do not remove.
• Factor performance is evaluated using forward returns, with mature samples informing gradual budget changes.
• New factors move through observation and low-weight trials before receiving larger allocations.
• Engineering checks and simulated execution tests do not demonstrate long-term profitability or factor predictive power.

Tags

Cited by

• Strategies Perp-Led Selloff Absorption Panel: buy a Binance USD-M alt perp at the next 4H open when its 24h return is <= -2 sigma AND its perp-vs-index premium is simultaneously <= -1.5 sigma (the selling came through the derivative, not spot), exit on a FIXED 12h clock; long-only, 12% per leg, max 5 concurrent, flat ~93% of the time (10 pre-2021 liquid USD-M perps, 4H)
• Hypotheses Perp-Led Selloff Absorption Panel: buy a Binance USD-M alt perp at the next 4H open when its 24h return is <= -2 sigma AND its perp-vs-index premium is simultaneously <= -1.5 sigma (the selling came through the derivative, not spot), exit on a FIXED 12h clock; long-only, 12% per leg, max 5 concurrent, flat ~93% of the time (10 pre-2021 liquid USD-M perps, 4H)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.