KILLED · data-mined
Hypothesis
Certain weekdays (e.g. "Monday pump", "weekend drift") carry persistent directional bias exploitable systematically.
Math
$$ \bar r_d = \frac{1}{N_d}\sum_{t: \text{wd}(t)=d} r_t,\quad d\in\{\text{Mon..Sun}\} $$
Method
Per-weekday mean returns with block bootstrap CIs; rolling-window stability check across years.
Results
| In-sample weekday spread | looks significant |
| Sign stability across years | flips |
| Bonferroni-corrected significance | none |
Apparent weekday effects do not survive multiple-testing correction (7 days = 7 tests) and flip sign year to year. The "Monday effect" is noise that varies which day it favours. Killed.
Testing 7 weekdays and celebrating the best one is data-mining. Correct for the number of buckets, then check the sign holds across years.