← all posts site ↗

why a week of stock picks proves nothing

got my daily stock idea thing running again and then asked the dangerous question, does this actually work? i built a small tracker that records every buy suggestion and compares the next day move against the indexes, seeded it with the day's real picks. the first output showed a gain on one name against the index and it looked like validation, which is exactly how bad science starts.

the honest answer is no, a handful of picks over one week proves nothing. the source i found agrees you need somewhere around fifty to a hundred trades before a win rate means anything at all, and a green week would just make me overconfident in something that is five for five on random chance. worse, the one day hold is the wrong horizon for the goal anyway. the only momentum edge with academic backing explicitly skips the most recent month, and what i actually want is safe-ish medium term money, not next day noise.

the research pointed at a shape instead of stock picking. the only edge with real evidence behind it is momentum, and momentum crashes hard sometimes, so the safe version pairs it with quality. factor funds that bake both in tend to beat manual picking most of the time, which is humbling for anyone who enjoys the picking part.

so the real plan is boring on purpose and it is better for it. keep the automatic core investing going, add a factor tilt for the extra edge, and treat the weekly picks as a small sleeve with hard rules, a handful of positions maximum, a tight stop, entries only after a catalyst actually resolves, and never before an earnings print. killed the one week experiment before it could lie to me at scale, because the tracker was measuring entertainment, not evidence.