Evidence · What the screens select

These are the stocks our screens select, and what we have and have not been able to show about them.

Our screens identify stocks that have recently moved sharply. We have not demonstrated that they beat a comparable basket of stocks: over the following month they rose about as often as comparable stocks picked on the same dates. A three-year test, registered in advance, began in October 2026. This is not investment advice.

Exhibit 1 Live · rebuilt daily

Today’s selection

Loading…

Loading today’s list…

Index universe in brackets. Being held by more screens has not been shown to lead to a better outcome: that hypothesis was tested in advance and reversed out of sample. The count is a fact about the screens, not a ranking of quality.

Exhibit 4 Static · as of 4 Oct 2026

Recommended stocks rose about as often as comparable stocks

Over the month after selection, the stocks our screens picked rose 55–58% of the time. Comparable stocks, chosen on the same dates, rose 53–57% of the time. Every interval spans zero.

Bar chart comparing how often recommended stocks rose against the matched base rate for
                comparable stocks over the same periods. The intervals overlap and all span zero.

Matched base rate: the share of eligible universe stocks that rose from the same entry date. Intervals from a block bootstrap by calendar month. Read the methodology →

Exhibit 6 Static · as of 25 Sep 2026

What these samples could have detected

How large an edge would have had to be for a sample of this length to find it. Anything below the curve is an edge our data could not have distinguished from chance, whether or not it exists.

Curves showing the smallest monthly premium each sample length could have detected,
                with the registered three-year forward test marked.

Curves are twice the series’ own monthly standard deviation divided by the square root of the number of months: the premium that would reach t = 2. The three-year forward test registered in October 2026 is marked. Read the methodology →

Evidence B

The record

Updates daily

Every name the screens have surfaced, with how long it stayed on the list and how far the stock moved — updated each day and never edited after the fact.

The list runs from the engine’s live database; entries are added the day they are flagged and are not removed when they go against us. The figure is the stock’s own move between the first and last time it was selected, not a return the portfolio earned: a name is held for only some of the days in that span.

Identified by the engine
Loading…