Quadrants, transition risks and event classes
PriorityX exposes three related concepts. Keep them separate when interpreting charts or filtering saved results:
| Concept | Produced by | Values |
|---|---|---|
| Score quadrant | Fitting / movement tracking | Q1–Q4, from score signs |
| Saved transition risk | extract_transitions() | critical, high, medium, low, stable |
| Displayed event class | Timeline / AutoViz preparation using movement | 1 Crisis, 2 Investigate, 3 Monitor, 4 Low |
A red Q1 point in a matrix and a red Crisis event use separate rules. Quadrants are score positions, not calendar quarters, ranks or verified current status. See methodology for score scales and inclusion filters.
Saved transition risks
For cross-quadrant changes, extraction uses these mappings:
| Transition | Saved risk_level |
|---|---|
| Q2/Q3/Q4 → Q1 | critical |
| Q3/Q4 → Q2 | high |
| Q4 → Q3 | medium |
| Q1 → Q2, Q1 → Q4, Q2 → Q3 | low |
| Other cross-quadrant changes | stable |
The default focus_risk_increasing=True removes cross-quadrant low and stable
rows. Pass False to retain them. When the quadrant is unchanged, extraction
checks a positive y_delta > 1.0 first (*Y, critical), then positive
x_delta > 1.0 (*X, high). If both exceed 1.0, that extraction branch records
*Y because Y has precedence. It does not generate every possible display spike.
Filter saved risks with their actual column:
critical_transitions = transitions[transitions["risk_level"] == "critical"]
These rows are not necessarily the same as Crisis events after display preparation.
Displayed event classes
When full movement history and cumulative counts are available, the renderer compares an event’s score position/count with its preceding saved movement row. The first matching rule wins:
| Class | Rule |
|---|---|
| 1 Crisis | abs(delta_x) > 0.40 or abs(delta_y) > 0.40, or count growth > 500% with a count increase > 50 |
| 2 Investigate | Either absolute score delta > 0.15; otherwise a destination exactly Q1 with both coordinates outside the boundary band; otherwise growth > 100% with count increase >= 5 |
| 3 Monitor | Either coordinate has absolute value <= 0.10, or the destination is exactly Q1, after the earlier rules fail |
| 4 Low | No earlier rule matches |
The comparisons are strict where shown: exactly 0.40 is not a Crisis score
trigger. Extreme negative score moves also satisfy the absolute-delta rule.
Counts are changes in cumulative event rows, not a percentage change in the
single-period count; a zero preceding cumulative count gives a zero percentage.
A saved destination suffix is passed through, so Q1*Y is not the exact Q1
string used by the destination-specific branches.
The display falls back to Investigate when the matching current row, preceding
history or cumulative counts are unavailable. Preserve history and its original
chronological row order before display filtering; do not send a YTD-only movement
extract for class calculation. Plotting does not add a priority column to the
caller’s transitions. AutoViz’s prepared events expose the class separately.
Spike suffixes
During display classification, the absolute score-spike tests above give *X,
*Y or *XY. A count-only Crisis uses the compatibility *Y marker. If a saved
to_quadrant already contains a suffix, the display preserves it rather than
replacing it. Therefore suffixes can originate from extraction or rendering;
the extraction and display thresholds are not interchangeable.
In compact notation, 3→②*XY reads source Q3 → destination Q2, with both-axis
spike indication. Marker date/coordinates remain at the event. Color identifies
the displayed event class independently of the destination digit.
Threshold interpretation
These are fixed operational heuristics in the implementation. The repository does not establish a percentile, sigma equivalence, calibrated detection accuracy or universal response action for them. Count-model log effects and standardized continuous-metric effects use different scales. Review historical performance and false-positive tolerance for your own data and workflow.
The display classifier is priorityx.tracking.transitions.classify_priority.
The driver-analysis helper priorityx.tracking.drivers.classify_priority uses
Critical for numeric class 1 and a count_delta argument; timeline displays use
Crisis. These labels are algorithmic classifications, not outcome verification.