Research built to be disproven
Every idea here is tested against the past before it touches the present — and tested honestly. Backtests use only what was knowable on the day. Results are measured against matched comparisons. A result that cannot survive an attempt to break it is not a result.
Luck is subtracted first
Try enough ideas and some will look brilliant by accident. Our headline numbers are penalized for every variant that was tried along the way, so a strategy cannot ride a lucky draw into the book. What remains after the discount is what we believe.
Capital is earned
Before any strategy trades, it passes through a set of adversarial gates designed to kill it — on data quality, on statistical power, on the honesty of its construction. Most ideas die there. The gates have no override.
Nothing is owed a place
Every strategy is held to the record that admitted it. When the record stops holding, the strategy is removed without ceremony. The book has no tenure.