Investigating Differences Between Factor Platform and Replicated Returns

Summary

This forum post raises a factor-research reproducibility problem: a user reports that returns from reproducing four factors differ substantially from the platform’s published results. The suggested comparison holds the factor and backtest period constant, uses a portfolio of 500 stocks, and checks both ascending and descending score rankings. The author proposes using any platform factor for a trial replication and gives an illustrative comparison between a platform annualized return and a lower reproduced return that they would consider close enough.

The post provides a checklist for aligning a replication, but it does not identify the cause of the discrepancy or describe data cleaning, universe construction, rebalancing, transaction costs, weighting, or other backtest settings. It reports no independently verified replication outcome. Its value is as a research question about validating factor backtests and checking whether implementation choices match the platform, rather than as a factor strategy or a resolved analysis.

Key ideas

• The author reports large return differences when reproducing four platform factors.
• The proposed comparison aligns the factor and backtest period and holds 500 stocks.
• Both ascending and descending score rankings are suggested for checking the result.
• The post does not determine which implementation or data choices explain the discrepancy.
• The example return comparison is a target for acceptable replication, not a verified result.

Tags

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