The feed is a schedule, not an event — and one field is written before the trade exists.

Three moments

Pre-entry, at-entry, and during — each tool has its own slot.

The regime read fires first and earliest. It has to, because it selects the branch, and the branch determines the management contract the trade will be held to for its whole life. The convergence grade is taken while the zone is still approaching, before the trigger candle. The continuation state is sampled at entry and again at the management checkpoints, because its job continues after the position opens. Only one of these three is a single reading; the others are a commitment and a series respectively, and treating all three as one entry-time snapshot loses what makes each useful.

Why the ordering carries the weight

A value recorded after the outcome is known has stopped being a measurement.

The reason the pre-entry commitment matters is not tidiness. A regime label chosen before the trade can be tested: group the trades by the label, compare how management actually performed in each group, and the label either predicted something or it did not. A regime label filled in afterwards cannot be tested at all, because the outcome has already contaminated it. Nobody assigns Trend to a trade that got stopped at break-even, and nobody notices themselves not doing it. The ordering is what protects the field from the trader who is honestly trying to be accurate.

The one that keeps moving

Continuation state is a series, and only its endpoints usually survive.

The continuation stack goes on producing readings for as long as the position is open, which raises a question the other two do not face: which of those readings enters the record. In practice the useful captures are the state at entry and the state at the checkpoints where management decisions actually get made. Everything between is real and not worth storing, because there is no decision attached to it. This is a deliberate thinning rather than a gap — the record wants the readings that accompanied a choice, so that the choice can later be examined alongside what it was looking at.

What a late write costs

One misplaced timestamp corrupts a whole slice, not one row.

The consequence of writing a field at the wrong moment is not confined to that trade. Every downstream analysis that groups by the field inherits the error, and the error is systematic rather than random — late-written labels drift toward whatever the outcomes were, so the resulting slices will show the labels predicting outcomes beautifully. A conditional expectancy table built on retrospectively-assigned regimes is not merely inaccurate; it is confidently wrong in the flattering direction, which is the hardest kind of wrong to catch by reading it.

• Retrospective labels produce slices that appear to validate the labelling.
• The error is invisible in the row and visible only in the discipline that produced it.
• Capturing at the moment of the decision is the only practical defence.

The reconstruction test

Ask whether the fields could be rebuilt from the chart alone. They could not.

A useful check on whether the schedule is being honoured: could someone reproduce these fields later by pulling up the chart and looking at where the trade happened. For the regime read the answer is roughly yes, which is exactly why it is tempting to defer. For the convergence grade taken before the trigger, and for the continuation state at a checkpoint that has since been overwritten, the answer is no — those readings existed at a moment and are gone. Fields that cannot be reconstructed are the ones the schedule exists to protect, and they are the ones worth being strict about.

The key idea

In an evidence system, when a value was written is part of the value.

The narrow set of fields carrying indicator output into the workbooks is only worth what the discipline of writing them is worth, and that discipline is mostly about timing rather than accuracy. A slightly imprecise reading committed before the outcome is far better evidence than a perfectly accurate one committed after, because only the first one could have been wrong. That asymmetry is the whole reason the feed has a schedule at all.