A label carried into research
The BSP archive contains directional information associated with forecast events. Tradour models that information neutrally as a feature. A label such as Rise or Drop describes what the source recorded under its own convention. It does not automatically tell a user to buy, sell or reverse a position.
That distinction matters because research asks what historically followed a defined feature. Product advice would be a different claim with different evidence, legal boundaries and real-world implementation requirements.
Raw and displayed direction are not identical
Historical BSP emails applied a global inversion when presenting the model direction. Tradour therefore stores the original raw direction and the displayed direction as separate fields wherever the source allows. The transformation is explicit rather than buried in a chart or silently applied in presentation code.
A tested interpretation is another separate choice. Research can ask whether an outcome followed the displayed direction or moved opposite it. Neither interpretation is assumed to be inherently correct before the test is defined.
Strength and indicator count add context
Some records include a Rise/Drop strength and an indicator count. These can be studied as features, but a larger value is not automatically better. In the current historical archive, indicator count alone and strength alone do not improve monotonically.
A striking combination can appear when many cells are explored. That makes the wider search part of the evidence. The cell may be worth testing again, but its percentage should not be presented as a reliable future edge.