Skip to content
All library documents

Testing the 24-Hour Roll-Out Effect with a TradFi Perpetual Prototype

Article FMZ digest · Author: 发明者量化-小小梦

Summary

The article studies whether a predictable change in the displayed 24-hour return can influence later prices. As older price moves leave the rolling window, the displayed return changes mechanically even if the current price stays flat. The proposed signal estimates that change over the next 30 minutes, uses the current 24-hour move as an attention filter, and requires the shorter five-minute shock to agree in direction. A Rust prototype applies this idea to Binance TradFi perpetuals, with market scanning, candidate selection, maker order handling, position monitoring, and safeguards for uncertain order outcomes.

The cited crypto-perpetual research found a concentrated but fragile roll-out anomaly, with costs, funding, tail losses, and differences between long and short results limiting its usefulness. That evidence does not establish the effect in TradFi; the article presents its TradFi implementation as a hypothesis to test. It recommends event-level analysis, nearby-window placebo checks, direction-specific evaluation, and measurement of fills and costs. Simulated fills omit queue position, and the prototype has documented data-loading and order-state edge cases.

Key ideas

  • A rolling 24-hour return can change predictably as older price moves leave its calculation window.
  • The proposed signal trades the expected display change, while current return magnitude serves as an attention filter.
  • The TradFi perpetual implementation is a research prototype, and crypto-market evidence does not validate it in TradFi.
  • Evaluation should separate long and short events, test nearby window lengths, and include fill probability and trading costs.
  • The paper execution model omits queue position, while live operation must handle delayed or ambiguous order responses carefully.

Tags

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