Step 6: Focus

Purpose

The Focus step analyzes the grading results and ranks every test by its actual impact on the final score. Instead of just listing pass/fail, it answers the question: "Which failed tests cost the student the most points?" This ranking is essential for generating meaningful feedback that prioritizes the most impactful improvements.

How It Works

The FocusService traverses the ResultTree produced by the Grade step and calculates a diff_score for each test — the number of points (on the 0–100 final score scale) that the test deducted.

The calculation accounts for the full weight chain from root to test:

diff_score = (100 - test_score) × (test_weight / 100) × cumulative_multiplier

The cumulative_multiplier is the product of all parent weight factors from the category root down to the test's parent subject. When a node has both subjects and direct tests (split via subjects_weight), the multiplier is split accordingly between the two groups.

Tests are processed per category (base, bonus, penalty) and sorted by diff_score in descending order within each category.

Dependencies

Step	What It Needs
Grade	The ResultTree containing all test scores and the weight hierarchy

Input

Source	Data
Pipeline	StepName.GRADE → GradeStepResult.result_tree

Output

Field	Type	Description
data	Focus	Contains ranked test lists for base, penalty, and bonus categories
status	StepStatus.SUCCESS	On successful analysis

The Focus Data Structure

@dataclass
class Focus:
    base: List[FocusedTest]              # Tests from base category, sorted by impact
    penalty: Optional[List[FocusedTest]] # Tests from penalty category (if configured)
    bonus: Optional[List[FocusedTest]]   # Tests from bonus category (if configured)

@dataclass
class FocusedTest:
    test_result: TestResultNode  # Full test result with score, report, parameters
    diff_score: float            # Points deducted from final score (0-100 scale)

Example

Given a submission with these results:

Test	Score	Weight	Parent Multiplier	diff_score
Test A	50	30	1.0	15.0
Test B	90	20	1.0	2.0
Test C	0	10	0.5	5.0

The Focus base list would be: [Test A (15.0), Test C (5.0), Test B (2.0)]

This tells the student: "Fixing Test A would improve your score the most."

For a comprehensive explanation of the focus algorithm, data storage, and use cases, see the Focus Feature documentation.

Failure Scenarios

• Grade step result missing or malformed → exception caught, StepStatus.FAIL.
• Empty result tree → produces empty focus lists (not a failure).

Next Step

After identifying the most impactful errors, the pipeline proceeds to Step 7: Feedback to generate student-facing reports.

---

Source

autograder/steps/focus_step.py → FocusStep

autograder/services/focus_service.py → FocusService

autograder/models/dataclass/focus.py → Focus, FocusedTest