How this works: This site uses a modified version of
a type of machine learning algorithm called a recommender system.
This is what, for example, Netflix might use to recommend new movies to you.
The algorithm is applied to data from 13F forms that large investors are required
to file with the SEC after each quarter, detailing what stocks they hold as of
the end of the quarter, as well as 13G and 13D forms that investors are
required to file when they buy more than 5 percent of a company.
How this was tested: I've tested how well this algorithm predicts
"unexpected" big buys, in which an investor goes from owning none or less than 1%
of a stock in one quarter, to 5% or more in the next quarter.
On a sample of a few thousand of these unexpected buys over the past few years,
I've found that a large fraction appear near the top of the predicted lists
for each investor (the lists from using "Find stocks a particular investor might like"
on the main page).
For example, about a third of the unexpected buys appear
in the top 35 predicted names for each investor.
Gurupredictor was created by Isaac Dimitrovsky, Ph.D. For inquiries, please use the form below: