Converting Mid- and High-Frequency Data into Longer-Horizon Alpha Factors
Summary
The document introduces an index-enhancement research report about building longer-horizon price and volume factors from mid- and high-frequency data. Its central motivation is a mismatch between data frequency and forecast horizon: as sampling frequency rises, the time-series autocorrelation of observations tends to fall, which can limit how far ahead a factor predicts. The report therefore frames frequency conversion, or downsampling, as an important step in extracting useful alpha from high-frequency inputs.
The proposed direction is to develop a general way to reduce the frequency of such data, then express alpha in a formulaic form so that many longer-period price and volume factors can be constructed in batches. The available text contains only the abstract and a link to the full report. It provides no equations, implementation details, empirical results, or evaluation of factor performance, so the specific downsampling method and its practical limits cannot be assessed from this excerpt.
Key ideas
- Higher-frequency observations tend to have lower time-series autocorrelation and may support shorter forecast horizons.
- Frequency conversion is presented as a way to make mid- and high-frequency information useful for longer-horizon alpha research.
- The report aims to define a general downsampling approach for such data.
- Formulaic alpha is proposed as a way to generate batches of longer-period price and volume factors.
- The excerpt provides no empirical evidence or methodological details beyond the abstract.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.