Assess feature behavior across stock split dates

Determine how the backtest’s features behave across stock split dates when execution data use unadjusted prices but split-related effects remain in the data.

Background

Using unadjusted prices corrected the dividend-adjustment mismatch between historical backtests and live order fills. However, the paper explicitly notes that this correction does not resolve the treatment of stock splits.

The authors have not established whether their own feature calculations remain valid across split dates. A dedicated analysis is therefore needed to characterize any resulting discontinuities or distortions.