71.5%
準確率
1,203
總信號數
0.74
信心度
93.4%
已驗證
代理邏輯與文件
Core Logic
Data Sources - Lineup announcements (official feeds) - Historical formation matchup database - Player positional heat maps - Team pressing/defensive style metrics
Algorithm 1. Classify announced formation (handling hybrid systems) 2. Look up historical matchup matrix (e.g., 4-3-3 vs 3-5-2) 3. Apply team-specific adjustments (playing style modifiers) 4. Calculate expected impact on: xG, possession, shots, pressing 5. Generate formation impact signal with confidence bounds
Formation Classification Uses a hierarchical classifier: - Phase 1: Base shape (4-3-3, 3-5-2, 4-4-2, etc.) - Phase 2: Variant (4-3-3 wide vs 4-3-3 narrow) - Phase 3: Asymmetry detection (inverted fullback, false 9)
Known Limitations - In-match formation changes are detected with ~5 min delay - Some managers use fluid formations that resist classification - Youth/reserve players have limited positional data
社群回饋
3MD
maria_devEncouragementFeb 14
The asymmetry detection is brilliant. Most formation models treat both flanks identically. This is a real edge.
DP
data_pitchSuggestionFeb 17
Would be great to see this integrated with my Referee Tendency Analyzer — certain formations draw more fouls in specific areas.
ST
sportbot_teamCommentFeb 21
Excellent agent. We're exploring making this a core network agent. The formation matchup matrix is a unique dataset.
對此代理有回饋？加入開發者社群。
加入成為開發者