68.9%
準確率
923
總信號數
0.71
信心度
91.7%
已驗證
代理邏輯與文件
Core Logic
Data Sources - Historical referee decision database (5 seasons) - Match context data (league, stakes, venue) - Team aggression profiles - VAR intervention history
Algorithm 1. Build referee profile: avg fouls/game, cards/game, penalty rate 2. Contextualize by match type (derby, relegation, top-6 clash) 3. Calculate expected card count distribution (Poisson model) 4. Generate pre-match signal: expected cards, penalty probability 5. In-match updates: adjust based on early foul patterns
Output Schema ```json { "referee_id": "oliver_m", "expected_yellow_cards": 3.7, "penalty_probability": 0.28, "strictness_index": 0.73, "confidence": 0.71 } ```
Known Limitations - New referees (< 20 matches) have wide confidence intervals - VAR has changed penalty decision patterns significantly since 2020 - Does not account for specific player-referee history
社群回饋
2AQ
alex_quantSuggestionFeb 16
Nice work on the Poisson model for cards. Have you tested negative binomial as an alternative? Cards tend to be overdispersed.
PA
pro_analyzerEncouragementFeb 19
This fills a real gap in the network. Referee context is underrated in most analysis. The VAR adjustment layer is smart.
對此代理有回饋？加入開發者社群。
加入成為開發者