# Metric Sweep Report

The sweep generated 96 extra deterministic specimens without rendering every map image. It tests whether the curated gallery is representative of the generator's metric space.

## Correlations With Score

| Metric | Pearson r |
| --- | ---: |
| network_length | 0.734 |
| branch_count | 0.712 |
| edge_complexity | 0.667 |
| tidal_swing_fraction | 0.431 |
| sinuosity | 0.368 |
| habitat_balance | 0.271 |
| mouth_width | -0.184 |

## Top Score Outliers

| Rank | Seed | Name | Score | Network | Swing | Branches | Mouth width |
| --- | ---: | --- | ---: | ---: | ---: | ---: | ---: |
| 1 | 42034 | Harrow bight | 5.3615 | 669.32 | 0.2681 | 7 | 0.8727 |
| 2 | 42011 | Lumen bar | 5.3047 | 671.19 | 0.2735 | 8 | 0.954 |
| 3 | 42027 | Morrow fen | 5.2612 | 676.05 | 0.233 | 8 | 0.8617 |
| 4 | 42021 | Tern leys | 5.2569 | 685.1 | 0.2754 | 8 | 0.8234 |
| 5 | 42050 | Glass strand | 5.256 | 638.05 | 0.2863 | 7 | 0.9168 |
| 6 | 42045 | Tern strand | 5.2147 | 632.3 | 0.2559 | 7 | 0.9437 |
| 7 | 42086 | Glass mouth | 5.1851 | 610.45 | 0.2633 | 7 | 1.0287 |
| 8 | 42023 | Harrow marsh | 5.1527 | 680.33 | 0.264 | 8 | 0.9692 |
| 9 | 42067 | Morrow wash | 5.1521 | 685.96 | 0.2651 | 8 | 0.8934 |
| 10 | 42075 | Wick fen | 5.1287 | 639.19 | 0.2638 | 7 | 0.8964 |

## Alternate Winners

- Longest network: Mica marsh seed 42081, network length 738.16.
- Largest tide swing: Glass strand seed 42050, swing 0.2863.

## Reading

The sweep makes the scoring function inspectable. If one metric dominates, the atlas is mostly selecting for that property. If the correlations are split, the top maps are winning through a mixture of shoreline shape, tide response, and channel structure.