From the abstract:

>Measured from the day a trade became public, the median purchase did −2.4% against the equal-weighted index over six months (90% range −3.9% to +0.1%), and the median stock members sold did −1.7%: what they bought did no better than what they sold. Measured from the members' own trade dates, purchases did −1.5%. No group of purchases (by party, size, account or disclosure delay) did clearly better than the index.

Not what you would have guessed, given that the topic was inspiring enough to make one person ask AI to vibe code a website to present the info. But negative results are results too and deserve to be reported.