Seven classifications: what the lab stamps on every trade.

The four that grade execution

Entry quality, opportunity quality, the zone, and the grade.

MAE quality classifies how much of the original stop risk the trade consumed, which is the cleanest read on entry timing and whether stops are sized for the noise they are meant to survive. MFE quality classifies whether the setup generated enough opportunity to justify the branch and the risk committed. The efficiency zone combines the pair into a single positional read, and the execution grade is the trade-level verdict that aggregates upward into the weekly rollup. These four are the visible product of the lab, and read on their own they describe a trade's life cycle competently.

The three that grade the record

Pips translation, branch integrity, and data QA are audits, not diagnostics.

The remaining three ask a different kind of question. The pips-translation check confirms that the R accounting reconciles with actual market distance — a mismatch means the risk arithmetic on that row is wrong, and every efficiency ratio derived from it is wrong with it. The branch integrity flag surfaces trades whose behaviour did not match their declared branch, which is a classification error rather than an execution one. Data QA status asks whether the row is complete and internally consistent enough to be read at all. None of the three says anything about how well the trade was managed. All three say whether the other four can be believed.

• An efficiency ratio computed from an unreconciled row is precise and meaningless.
• Branch integrity catches the case where a Normal trade was managed as a trend trade — the metrics look odd because the classification, not the execution, was wrong.
• Data QA status is read first for the same reason the regime engine's QA flag is: trust before content.

The branch integrity flag

The most interesting stamp, because it indicts the label rather than the trade.

Branch identity in MARS is declared at entry and is not supposed to drift — a trade is Normal, Trend Partial, Trend No-Partial, or Overflow, and each carries a different exit architecture and a different expected excursion profile. The integrity flag exists because declared identity and actual behaviour can diverge, and when they do the efficiency metrics become uninterpretable rather than merely poor. A Trend No-Partial trade exited at a fixed target will show excellent capture efficiency and terrible fat-tail participation, and neither number describes anything except a misclassification. The flag surfaces those rows so the classification can be examined instead of quietly absorbed into a branch's statistics, where it would slowly corrupt the branch-level verdict.

Why classify at all

Classifications aggregate; raw ratios do not.

The lab computes six or more ratios per trade, and at a live rhythm that is a volume of precision no weekly review can absorb row by row. Classifications solve it the way the gate ladder solves drawdown: by converting continuous values into states. States can be counted, and counting is what makes the weekly rollup possible — high-MAE counts, elite-MFE counts, breakeven trigger rates, and a week quality grade all aggregate from per-trade classifications rather than from raw numbers. An operator cannot usefully average forty capture-efficiency figures; they can immediately use the fact that six of forty landed in the poorest zone.

The key idea

A grading system that never grades its own inputs is grading fiction.

It would have been entirely natural to build four execution classifications and stop. The three audit stamps exist because the lab's arithmetic is only as good as the rows it runs on, and rows are typed, pasted, and transferred by hand. Checking the evidence before grading it is not overhead — it is the difference between a diagnostic and a decorative one.