| name | forecasting-desk |
| description | Use when converting sports evidence into probability estimates, squad recommendations, risk-adjusted rankings, or market comparisons, especially when the answer should remain explainable and uncertainty-aware in any agent runtime. |
Forecasting Desk
Use this skill when the task is to turn evidence into a forecast or ranked choice.
Principles
- Forecasts are estimates, not facts.
- Every strong recommendation should map back to specific inputs.
- Confidence should fall when the evidence is stale, sparse, or contradictory.
- Explainability matters as much as the ranking itself.
Decision workflow
- List the decision options.
- State the key variables affecting each option.
- Convert evidence into directional effects.
- Estimate upside, downside, and fragility.
- Rank the options.
- Explain what new information would alter the ranking.
Useful output shapes
Probability summary
{
"option": "",
"probability": 0,
"confidence": 0,
"drivers": [],
"risks": [],
"change_triggers": []
}
Squad recommendation
{
"squad": [],
"captain": "",
"vice_captain": "",
"constraints_used": [],
"core_reasons": [],
"main_risks": [],
"alternatives": []
}
Guardrails
- Do not hide uncertainty behind confident prose.
- Avoid overfitting to one source or one narrative.
- Mark speculative edges clearly.
- If comparing with external markets, distinguish internal estimate from market-implied probability.
