WeaveMark study report

Crowd Factory Puzzle Ablation Summary

A browser puzzle game about steering autonomous crowds through factory automation, belts, machines, crates, spatial pushing rules, hazards, and readable level constraints.

At a glance

Net score

+11

Sum of -3..+3 criteria.

Strongest control

[C2] Matched-prose no-expand control

Game `@expand` study

Treatment

[T] Expanded WeaveMark treatment

Focused expansion evidence where source concepts are already concrete.

Semantic trace

workers -> belts/machines -> crate pushing -> hazards -> readable levels

Propagation path inspected.

Variant metrics

Columns are grouped into authoring size, final output size, and semantic-information proxies.

VariantRoleSourceVariablesOutputLeverageFact unitsDensityYield
[C1] Compact no-expand control
[C1] Compact no-expand control
Compact source that names the puzzle inspirations without expansion.9001,17113.01x7059.8777.8
[C2] Matched-prose no-expand control
[C2] Matched-prose no-expand control
Manual prose expansion without using `@expand`.18201,3147.22x89.2567.9490.4
[T] Expanded WeaveMark treatment
[T] Expanded WeaveMark treatment
Uses `@expand mode: intention` to unpack concept labels into mechanics.19103,33417.46x244.7573.41,281.4
Metric definitions and scoring legendOpen for exact meanings.

Source words

Words in the local study source for a variant; this is the local authoring burden.

Variable words

Words in a variant's explicit input payload, when a template or refinement uses one.

Output words

Words in the saved compiled final artifact.

Leverage

Output words divided by local source words; larger means more final artifact per local word, not quality by itself.

Fact units

Novelty-weighted semantic fact units extracted from the output by deterministic rules.

Density

Discounted fact units per 1,000 output words; higher means a more information-dense output.

Yield

Discounted fact units per 1,000 local source words; higher means more semantic material per local authoring word.

Treatment-control comparison

Metric[C2] Matched-prose no-expand control[T] Expanded WeaveMark treatmentDirection
Source-only leverage7.2217.46win
Discounted fact units89.25244.75win
Information density67.973.4win
Information yield490.41,281.4win

Contrastive gain/loss scores

Scores compare [T] Expanded WeaveMark treatment against [C2] Matched-prose no-expand control on the -3..+3 scale.

Primary scores are blind* using hybrid-derived-metrics-and-masked-review: anonymous absolute 1..7 scores were frozen before reveal, then converted to the -3..+3 treatment-control scale. Hybrid blind* scoring uses derived metrics for mechanical criteria and masked source/output review for criteria that require actual reading. The masked review is less blind because domain content, source syntax, or style can leak, but this is necessary to avoid replacing readability and integration judgments with weak length/density proxies.

CriterionBlind* scoreEvidence
Authoring leverage+3derived-evidence method. Blind* absolute scores: 7 for [T] versus 1 for the strongest control.
Information yield+3derived-evidence method. Blind* absolute scores: 7 for [T] versus 1 for the strongest control.
Grounded expressiveness+2masked-source-output review method. Blind* absolute scores: 7 for [T] versus 3 for the strongest control.
Input readability-1masked-source review method. Blind* absolute scores: 5 for [T] versus 6 for the strongest control.
Output readability+1masked-output review method. Blind* absolute scores: 6 for [T] versus 5 for the strongest control.
Constraint integration+2masked-source-output review method. Blind* absolute scores: 6 for [T] versus 3 for the strongest control.
Reusable abstraction quality+1masked-source review method. Blind* absolute scores: 6 for [T] versus 4 for the strongest control.
Total+11Net contrastive gain/loss.
Metric definitions and scoring legendOpen for exact meanings.

Contrastive score

A -3..+3 judgment comparing [T] against the strongest listed control for each criterion.

Total score

The sum of the seven contrastive criterion scores for one study.

Score color

Green means [T] is better, red means worse, amber means similar; intensity follows magnitude.

Gains and failures

Gains

  • [T] Expanded WeaveMark treatment wins source-only leverage: 17.46 versus 7.22 for [C2] Matched-prose no-expand control.
  • [T] Expanded WeaveMark treatment wins discounted fact units: 244.75 versus 89.25 for [C2] Matched-prose no-expand control.
  • [T] Expanded WeaveMark treatment wins information density: 73.4 versus 67.9 for [C2] Matched-prose no-expand control.
  • [T] Expanded WeaveMark treatment wins information yield: 1,281.4 versus 490.4 for [C2] Matched-prose no-expand control.
  • `@expand` keeps the source modular and still frames a coherent factory-crowd puzzle.
  • The treatment wins the matched-prose control on the deterministic leverage, fact-unit, density, and yield proxies.
  • It captures integration framing across autonomous workers, belts, machines, crates, hazards, and levels.

Failures and caveats

  • The source concepts are concrete enough that manual prose can unpack them very effectively.
  • The matched-prose control remains an important fairness baseline even when the treatment wins the proxy metrics.
  • This is saved-output semantic evidence; behavioral proof still requires downstream implementation/use.

Conclusion

A positive `@expand` result: useful for clarity and framing, and currently ahead of matched prose on deterministic proxy metrics.