Fig 1Grading linePick a run below · keys 1 2 3 · space pauses
How one result earns its place. A task card is stamped by a strict schema, then picks up the real Power BI model it is graded against. At the gold gate its own gold answer runs the probes first: both lamps must light, the gold passing and the known-wrong controls failing, or the card is pushed off the line and the task retired. In the turbine the fan cap seals the intake, the model gateway feeds the candidate into a real query, and it runs in five report contexts: detail rows, subtotals, the grand total, slicers and planted edge cases; a lamp lights for each one within tolerance. Under the loupe the dial reads correctness times difficulty, with a style bonus only for a correct answer. The archive opens only when the task commit, fixture manifest, analyzer and guidance hash all match, and only the score crosses the wall to the leaderboard. Scores shown are illustrative.