Register
A falsifiable question, outcome, population, split, and failure criteria before results.
RESEARCH METHODS
Our models:
how they work.
What we check before a fantasy football study is published.
→
Provenance
Source, retrieval time, license, attribution, schema, and SHA-256 recorded.
Analyze
Deterministic code, frozen seed, temporal validation, baseline, uncertainty, artifacts.
Attack
Statistician and adversarial reviewer look for leakage, selection, and overclaiming.
Verify
Every number traces to an artifact; every source and license claim is checked.
Gate
Mechanical checks plus an editor decision. A blocker stops publication.
REQUIRED CHECKS
What we require
- Future data stays future. Forecasting studies use rolling or held-out periods.
- Start with a baseline. More complex models must improve on a simple relevant forecast.
- Show uncertainty. We do not publish point estimates alone.
- Show limitations. Caveats stay with the conclusion.
- Link the evidence. Claims map to tables, charts, code, and hashes.
- Keep review records. Prompts, findings, and decisions are recorded.
PUBLICATION CHECKLIST
What must pass before we publish.
We check the data, sample, validation, uncertainty, review, limitations, and citations.
✓Data provenance
✓Temporal validation
✓Sample threshold
✓Uncertainty
✓Adversarial review
✓Limitations
✓Citation trace
REVIEW TEAM
Who checks the work
Researcher
Checks the study against the registered question.
Statistician
Checks the model, baselines, uncertainty, and validation.
Code / methods
Checks the implementation, tests, and row counts.
Adversary
Looks for weak claims, alternate explanations, and reader misuse.
Citation checker
Checks sources, evidence paths, and usage rights.
Visualization reviewer
Checks chart values, labels, and accessibility.
Editor / publisher
Keeps the story useful, clear, and supported.
AI IN THE WORKFLOW
Research help, not the headline
GLM-5.3 helps design, challenge, source-check, and edit studies. Deterministic code produces the statistics, tables, charts, and publication checks.
- MODEL
- GLM-5.3 via OpenRouter
- CALCULATION
- Deterministic Python
- REVIEW OUTPUT
- Structured records
- PUBLIC EVIDENCE
- Per-study Research File
Review labels.
AI review is not independent human peer review. External human review will be labeled separately when it exists.
See published research files →Corrections.
When a result changes, we add a dated note and keep the original research record available.
Correction policy →