Do your models beat the market?

No, and we don't claim they do. We tested our own stock-picking models against strict, pre-registered standards. None of them passed. That's why we build checking tools instead of selling picks.

Details

• "Pre-registered" means the pass/fail bar was written down and locked before we saw any results, so it couldn't move afterwards.
• The tests were hypothetical backtests (simulations on historical data), not live trading.
• The checks in Backtest Auditor come from that process, which found real data errors along the way.
• We record calls from one simple model of our own, labeled "research control: failed pre-registered backtests", to test our recording system. It isn't there to suggest skill, and its calls aren't recommendations. See research control forecasters.

We haven't published figures from this research.