signal-platform/.kiro/specs/rr-scanner-target-quality/tasks.md

Tasks

1. Add quality score helper function

• 1.1 Create _compute_quality_score(rr, strength, distance, entry_price, *, w_rr=0.35, w_strength=0.35, w_proximity=0.30, rr_cap=10.0) -> float function in app/services/rr_scanner_service.py that computes a weighted sum of normalized R:R, normalized strength, and normalized proximity
• 1.2 Implement normalization: norm_rr = min(rr / rr_cap, 1.0), norm_strength = strength / 100.0, norm_proximity = 1.0 - min(distance / entry_price, 1.0)
• 1.3 Return w_rr * norm_rr + w_strength * norm_strength + w_proximity * norm_proximity

2. Replace long setup selection logic

• 2.1 In scan_ticker, replace the long setup loop that tracks best_rr / best_target with a loop that computes quality_score for each candidate via _compute_quality_score and tracks best_quality / best_candidate_rr / best_candidate_target
• 2.2 Keep the rr >= rr_threshold filter — only candidates meeting the threshold are scored
• 2.3 Store the selected candidate's actual R:R ratio (not the quality score) in TradeSetup.rr_ratio

3. Replace short setup selection logic

• 3.1 Apply the same quality-score selection change to the short setup loop, mirroring the long setup changes
• 3.2 Ensure distance is computed as entry_price - lv.price_level for short candidates

4. Write unit tests for _compute_quality_score

• 4.1 Create tests/unit/test_rr_scanner_quality_score.py with tests for known inputs verifying the formula output
• 4.2 Test edge cases: strength=0, strength=100, distance=0, rr at cap, rr above cap
• 4.3 Test that all normalized components stay in 0–1 range

5. Write exploratory bug-condition tests (run on unfixed code to confirm bug)

• 5.1 [PBT-exploration] Create tests/unit/test_rr_scanner_bug_exploration.py with a property test that generates multiple S/R levels with varying strengths and distances, calls scan_ticker, and asserts the selected target is NOT always the most distant level — expected to FAIL on unfixed code, confirming the bug

6. Write fix-checking tests

• 6.1 [PBT-fix] Create tests/unit/test_rr_scanner_fix_check.py with a property test that generates multiple candidate S/R levels meeting the R:R threshold, calls scan_ticker on fixed code, and asserts the selected target has the highest quality score among all candidates

7. Write preservation tests

• 7.1 [PBT-preservation] Create tests/unit/test_rr_scanner_preservation.py with a property test that generates zero-candidate and single-candidate scenarios and asserts the fixed function produces the same output as the original (no setup for zero candidates, same setup for single candidate)
• 7.2 Add unit test verifying that when no S/R levels exist, no setup is produced (unchanged)
• 7.3 Add unit test verifying that when only one candidate meets threshold, it is selected (unchanged)
• 7.4 Add unit test verifying get_trade_setups sorting is unchanged (R:R desc, composite desc)

8. Integration test

• 8.1 Add integration test in tests/unit/test_rr_scanner_integration.py that mocks DB with multiple S/R levels of varying quality, runs scan_ticker, and verifies the full flow: quality-based selection, correct TradeSetup fields, database persistence