Lab
Where we test before anything can touch the public record
This page explains the research. It is not the local lab interface.
What we are testing
Whether an AI shortlist beats the quantitative leader, and whether a second ranking deserves a seat. Nothing is promoted until live sessions exist and the result is labelled honestly.
How we check a change
Replay a past close with only that day’s data. Enter at the next open. Compare against a control. Do not overwrite the public files. A clean job is a log line, not proof.
What is available
Public experiments appear only when they correspond to real studies. If the evidence is thin, we say so. Exploratory runs are not out-of-sample results.
Does the AI shortlist beat the quantitative #1 on a 20-session horizon? Live sessions are still too few. Not proven.
A second ranking on the same tape. Exploratory — not a validation and not out of sample.