Phase 3 · Predictive Analytics
See the risk before it costs you a tournament.
Three models running on your historical data: who's about to leave, which fixture is going to break, and where demand is about to spike. Real numbers. Clear actions. No dashboards that nobody opens.
Models trained on 2.847 events · 1.126 referees
Updated 18 min ago
Gradient Boosted Trees · AUC 0.89
Referees at risk of leaving
Ranked by 90-day churn probability
12 flagged
High Risk
7
+2 vs last month
Medium Risk
14
unchanged
Retained by Action
23
+18%
Model AUC
0.89
validation set
Upcoming matches with no-show risk
Next 14 days · sorted by probability
Weather · Travel · History · Workload
Critical (>60%)
3
action needed
Elevated (30-60%)
5
watch list
On Track
42
of 50 fixtures
Avg Backup Time
2.4h
down from 9h
Referee demand · next 30 days
Forecast vs. current capacity · 95% confidence band
LightGBM · MAPE 6.2%
Shortage Days
6
bookings at risk
Peak Date
May 18
Ibercup finals
Forecast Demand
2.140
+12% vs 2025
Current Capacity
1.806
-334 gap
Forecasted demand
95% confidence band
Current capacity
Shortage days