DispatchAtlasResults Portal
The DispatchAtlas results portal — disclosure-filtered static exports of solver rankings, comparisons, and statistical evidence you can browse interactively.
The results portal reads disclosure-filtered static exports. Switch between the sortable solver leaderboard, the tabulated result rows, the interactive runtime-quality scatter, the table catalog, and the significance-and-rankings surface. Every view is build-inlined and works without a network request; every result row carries an explicit disclosure label, and the leaderboard sorts by any numeric column with the keyboard.
| Solver | Problem | Objective | Value | Feasible | Disclosure |
|---|---|---|---|---|---|
| earliest-start | smoke-job-shop-0 | makespan | 353.000 | yes | public |
| shortest-processing-time | smoke-job-shop-0 | makespan | 433.000 | yes | public |
| earliest-start | smoke-workflow-0 | makespan | 11.200 | yes | public |
| shortest-processing-time | smoke-workflow-0 | makespan | 11.200 | yes | public |
earliest-start | 2 | 2/2 | 11.200 | 182.100 |
shortest-processing-time | 2 | 2/2 | 11.200 | 222.100 |
Bundled demo results, not a comparative ranking — every value is computed directly from the preview records. A superiority claim stays closed until fair, equal-budget campaigns supply sufficient runs.
| Table | Role |
|---|---|
| Solver summary | summary |
| Pairwise comparisons | post-hoc |
| Average ranks | ranking |
| Benchmark characterization | characterization |
| Infeasible-row accounting | constraint |
| Effect sizes | post-hoc |
| Budget speedup | efficiency |
| Quality indicators | multi-objective |
| Mechanism ablation | ablation |
| Feature coverage | capability |
| Framework comparison | capability |
The full statistical surface prepares solver-summary tables, pairwise comparisons with multiple-comparison correction, omnibus rankings, critical-difference post-hoc views, confidence intervals, effect sizes, convergence and run-order trends, and runtime-versus-quality tradeoffs.
- Every comparison reports its correction metadata alongside the statistic.
- Every row carries a disclosure label, never a bare number.
- A superiority claim stays closed until paired counts clear the statistical power floor on a fair, equal-budget comparison.
Figure catalog — 17 figure types
Performance profile
Role: performance
Solver rankings
Role: ranking
Critical-difference diagram
Role: post-hoc
Runtime-quality trade-off
Role: efficiency
Convergence trajectory
Role: convergence
Exploration-exploitation balance
Role: search-dynamics
Benchmark characterization
Role: characterization
Pareto frontier
Role: multi-objective
Empirical cumulative distribution
Role: distribution
Run-order trend
Role: drift
Scalability curve
Role: scaling
Seed stability
Role: reliability
Robustness tail-risk
Role: reliability
Parameter sensitivity
Role: sensitivity
Mechanism ablation
Role: ablation
Feature-richness matrix
Role: capability
Distribution-distance calibration
Role: calibration
Final comparative claims require pilot review, statistical sufficiency, full campaign evidence, and disclosure checks. Until then, the page labels data as smoke-development preview material and keeps every superiority claim closed.
Navigation
- Analysis exports describes the statistical methods.
- Solver recommender ranks solvers by declared fit.
- Downloads lists the public bundles behind these views.