How a piece earns a number

Every piece goes through the same pipeline, and each step can refuse to continue.

Evidence classes

• CONFIRMED

  Measured directly and checked: for example, quotes verified against the download ledger by hash.

• ESTIMATED

  Produced by a model or a stated assumption. Useful, but only as good as the assumption.

• MISSING

  Not available, or not tested. We say so on screen instead of filling the gap.

Rules every piece follows

• Every number traces to data. Scene text can’t contain a typed number. Numbers come from checked claims, and a checker refuses the video if any don’t match.
• Gaps stay gaps. Missing or unusable observations are left out and counted, never smoothed over.
• No untested causes. If we haven’t tested why something happened, the video can show that it happened, not why.
• Choices are fixed in advance. Which day, window or market to feature is decided by a rule written before looking at the results.
• After-the-fact analysis is labelled. Diagnostics that weren’t pre-registered say so on screen.
• Failures are published too. A strategy that fails out of sample is a result.

What a piece is not

Pieces use backtests and historical data. They are not trading records, signals or recommendations, and nothing here is investment advice.