The Capall Audit
The audit is the evidence. It shows the model and market outputs, their settled outcomes, the data window and the confidence around each published statistic.
Incomplete samples remain marked as pending rather than being turned into confident claims.
What this page tracks
The record contains walk-forward model comparisons, each with live samples, outcomes and confidence intervals since the method was frozen:
- Pure-form view — how the CR ranking compares with settled outcomes.
- Market view — how the CMR ranking compares with settled outcomes.
- Model versus market — where the two rankings agree or differ.
- Data quality — where coverage, market formation or sample size limits the read.
Every number is read live from the settled record and rebuilt nightly. Cells below the minimum sample floor show "pending" rather than a misleading number.
What walk-forward means
Each model comparison is validated out-of-sample before it is published. The model is fitted on a rolling historical window, scored on the next slice, settled, then the window advances. No future data is used to train an earlier slice.
How to read the record
Ratings describe the model's view; CMR describes the market's view; price and movement provide context. None of these outputs is a command to take an action.