Scouting knowledge

Each entity attribute has a true hidden value and, per player, a knowledge level or confidence interval. Scouting actions (assigning a scout, watching matches, spending time or money) raise knowledge for specific entities or regions; knowledge may decay. The display shows a masked value, a range, a noisy point estimate, or a star rating whose noise scales with the scout's judging skill and the knowledge level. State owned: (player, entity, attribute) knowledge and the current estimate. Update rule: knowledge increases with effort and observer skill; estimate = true value plus noise scaled by remaining uncertainty, or hidden until a threshold. Canonical forms: Football Manager attribute masking, regional scouting knowledge, and scout ability and potential star ratings; Out of the Park Baseball, where displayed ratings are your scouting staff's opinion and accuracy is a setting.

Recruitment becomes investment under uncertainty. The player decides where to send scouts, weighs a cheap unknown against a known star, and feels vindicated or burned when performance reveals the true value.

• representation of uncertainty: masked, range, noisy estimate, or rating
• scouting cost and speed per entity or region
• scout skill as the noise level of reports
• whether knowledge decays over time
• whether true values are ever fully revealed, for example after signing
• separate uncertainty for current ability and future potential

• Third-party editors and database tools reveal true attributes, collapsing the mechanic for players who use them.
• Cheap, complete scouting turns uncertainty into a delay rather than a decision.
• Uncertainty that persists after acquisition and long observation feels arbitrary.

• Information asymmetry creates market behavior: overpaying for hyped players and bargains in under-scouted regions.
• Uncertain potential couples with skill growth to make youth signings speculative bets.
• Rival AI clubs with better knowledge outbid the player, making scouting investment competitive.

• Can language-model scouts write reports whose hedges and biases track their numeric noise, so reading reports is itself a skill?
• Does Bayesian updating from observed performance, shown to the player, make uncertainty feel fairer than fixed masking?