Test your own idea
Pick a slice of the dataset — sector, market-cap tier, gap size, EPS surprise size — and this runs the exact same methodology as the rest of the site against it: market-adjusted excess returns, a one-sample t-test, and an in-sample/out-of-sample split.
Important: this tool does not correct for multiple comparisons. If you try many different filters, some will look statistically significant purely by chance — that's not a flaw in the math, it's what happens when you test enough things. Treat any single significant result here as a hypothesis worth investigating further, not a proven edge, exactly the same standard this project holds its own segmentation findings to.
Horizon