Evidence first. Results next.

A forward record, not a story told after the fact.

Read the source.

Every Form 4, every new 10-K and 10-Q compared with the year before, and every merger agreement, read by AI each weekday. Source: SEC EDGAR only.

Write the rules first.

A test’s fixed rules select its picks before any results. Pretend purchases are recorded at the next market open, plus trading costs, and held for a set number of trading days.

Make a fair comparison.

Each pick is compared with the index of companies the same size — SPY, IJH or IWM — and with 20 random stocks of the same size. The score is excess return per hold, not the market’s return.

Keep the whole record.

Every test stays public. Open and closed picks, winners and losers, passed and failed tests. No backtests. The pass bar was written November 1, 2026: 60 scored picks, more than 1% per hold over same-size companies, and a t-statistic above 2.

The published research

Cohen, Malloy and Pomorski on routine and opportunistic insider trades.

Cohen, Malloy and Nguyen, Lazy Prices, on changes in corporate reporting language.

Published research informs the rules. It is not a promise that a forward test will pass.